2018-06-28 19:23:57 +00:00
|
|
|
// Copyright (c) 2011-present, Facebook, Inc. All rights reserved.
|
|
|
|
|
// This source code is licensed under both the GPLv2 (found in the
|
|
|
|
|
// COPYING file in the root directory) and Apache 2.0 License
|
|
|
|
|
// (found in the LICENSE.Apache file in the root directory).
|
|
|
|
|
//
|
|
|
|
|
// Copyright (c) 2011 The LevelDB Authors. All rights reserved.
|
|
|
|
|
// Use of this source code is governed by a BSD-style license that can be
|
|
|
|
|
// found in the LICENSE file. See the AUTHORS file for names of contributors.
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
|
2023-10-25 18:59:09 +00:00
|
|
|
#include <memory>
|
|
|
|
|
|
2018-06-28 19:23:57 +00:00
|
|
|
#include "db/db_test_util.h"
|
2021-01-26 18:58:31 +00:00
|
|
|
#include "file/sst_file_manager_impl.h"
|
2018-06-28 19:23:57 +00:00
|
|
|
#include "port/stack_trace.h"
|
2020-03-04 20:30:34 +00:00
|
|
|
#include "rocksdb/io_status.h"
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
#include "rocksdb/sst_file_manager.h"
|
2019-05-30 18:21:38 +00:00
|
|
|
#include "test_util/sync_point.h"
|
2023-10-25 18:59:09 +00:00
|
|
|
#include "test_util/testharness.h"
|
2020-07-09 21:33:42 +00:00
|
|
|
#include "util/random.h"
|
|
|
|
|
#include "utilities/fault_injection_env.h"
|
|
|
|
|
#include "utilities/fault_injection_fs.h"
|
2018-06-28 19:23:57 +00:00
|
|
|
|
2020-02-20 20:07:53 +00:00
|
|
|
namespace ROCKSDB_NAMESPACE {
|
2018-06-28 19:23:57 +00:00
|
|
|
|
2020-03-04 20:30:34 +00:00
|
|
|
class DBErrorHandlingFSTest : public DBTestBase {
|
2018-06-28 19:23:57 +00:00
|
|
|
public:
|
2020-08-18 01:41:20 +00:00
|
|
|
DBErrorHandlingFSTest()
|
2021-07-23 15:37:27 +00:00
|
|
|
: DBTestBase("db_error_handling_fs_test", /*env_do_fsync=*/true) {
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
fault_fs_.reset(new FaultInjectionTestFS(env_->GetFileSystem()));
|
|
|
|
|
fault_env_.reset(new CompositeEnvWrapper(env_, fault_fs_));
|
|
|
|
|
}
|
2020-01-30 18:53:46 +00:00
|
|
|
|
2024-05-14 23:44:32 +00:00
|
|
|
~DBErrorHandlingFSTest() {
|
|
|
|
|
// Before destroying fault_env_
|
|
|
|
|
SyncPoint::GetInstance()->DisableProcessing();
|
|
|
|
|
SyncPoint::GetInstance()->LoadDependency({});
|
|
|
|
|
SyncPoint::GetInstance()->ClearAllCallBacks();
|
|
|
|
|
Close();
|
|
|
|
|
}
|
|
|
|
|
|
2020-01-30 18:53:46 +00:00
|
|
|
std::string GetManifestNameFromLiveFiles() {
|
|
|
|
|
std::vector<std::string> live_files;
|
|
|
|
|
uint64_t manifest_size;
|
|
|
|
|
|
2020-10-02 23:39:17 +00:00
|
|
|
Status s = dbfull()->GetLiveFiles(live_files, &manifest_size, false);
|
|
|
|
|
if (!s.ok()) {
|
|
|
|
|
return "";
|
|
|
|
|
}
|
2020-01-30 18:53:46 +00:00
|
|
|
for (auto& file : live_files) {
|
|
|
|
|
uint64_t num = 0;
|
|
|
|
|
FileType type;
|
|
|
|
|
if (ParseFileName(file, &num, &type) && type == kDescriptorFile) {
|
|
|
|
|
return file;
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
return "";
|
|
|
|
|
}
|
2018-06-28 19:23:57 +00:00
|
|
|
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
std::shared_ptr<FaultInjectionTestFS> fault_fs_;
|
|
|
|
|
std::unique_ptr<Env> fault_env_;
|
2018-06-28 19:23:57 +00:00
|
|
|
};
|
|
|
|
|
|
2020-03-04 20:30:34 +00:00
|
|
|
class ErrorHandlerFSListener : public EventListener {
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
public:
|
2020-03-04 20:30:34 +00:00
|
|
|
ErrorHandlerFSListener()
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
: mutex_(),
|
|
|
|
|
cv_(&mutex_),
|
|
|
|
|
no_auto_recovery_(false),
|
|
|
|
|
recovery_complete_(false),
|
|
|
|
|
file_creation_started_(false),
|
|
|
|
|
override_bg_error_(false),
|
|
|
|
|
file_count_(0),
|
2020-03-04 20:30:34 +00:00
|
|
|
fault_fs_(nullptr) {}
|
2020-10-02 23:39:17 +00:00
|
|
|
~ErrorHandlerFSListener() {
|
|
|
|
|
file_creation_error_.PermitUncheckedError();
|
|
|
|
|
bg_error_.PermitUncheckedError();
|
2021-12-08 22:22:41 +00:00
|
|
|
new_bg_error_.PermitUncheckedError();
|
2020-10-02 23:39:17 +00:00
|
|
|
}
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
|
2018-09-17 20:08:13 +00:00
|
|
|
void OnTableFileCreationStarted(
|
|
|
|
|
const TableFileCreationBriefInfo& /*ti*/) override {
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
InstrumentedMutexLock l(&mutex_);
|
|
|
|
|
file_creation_started_ = true;
|
|
|
|
|
if (file_count_ > 0) {
|
|
|
|
|
if (--file_count_ == 0) {
|
2020-03-04 20:30:34 +00:00
|
|
|
fault_fs_->SetFilesystemActive(false, file_creation_error_);
|
|
|
|
|
file_creation_error_ = IOStatus::OK();
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
cv_.SignalAll();
|
|
|
|
|
}
|
|
|
|
|
|
2020-10-02 23:39:17 +00:00
|
|
|
void OnErrorRecoveryBegin(BackgroundErrorReason /*reason*/, Status bg_error,
|
|
|
|
|
bool* auto_recovery) override {
|
|
|
|
|
bg_error.PermitUncheckedError();
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
if (*auto_recovery && no_auto_recovery_) {
|
|
|
|
|
*auto_recovery = false;
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
2021-12-08 22:22:41 +00:00
|
|
|
void OnErrorRecoveryEnd(const BackgroundErrorRecoveryInfo& info) override {
|
|
|
|
|
InstrumentedMutexLock l(&mutex_);
|
|
|
|
|
recovery_complete_ = true;
|
|
|
|
|
cv_.SignalAll();
|
|
|
|
|
new_bg_error_ = info.new_bg_error;
|
|
|
|
|
}
|
|
|
|
|
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
bool WaitForRecovery(uint64_t /*abs_time_us*/) {
|
|
|
|
|
InstrumentedMutexLock l(&mutex_);
|
|
|
|
|
while (!recovery_complete_) {
|
|
|
|
|
cv_.Wait(/*abs_time_us*/);
|
|
|
|
|
}
|
|
|
|
|
if (recovery_complete_) {
|
|
|
|
|
recovery_complete_ = false;
|
|
|
|
|
return true;
|
|
|
|
|
}
|
|
|
|
|
return false;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
void WaitForTableFileCreationStarted(uint64_t /*abs_time_us*/) {
|
|
|
|
|
InstrumentedMutexLock l(&mutex_);
|
|
|
|
|
while (!file_creation_started_) {
|
|
|
|
|
cv_.Wait(/*abs_time_us*/);
|
|
|
|
|
}
|
|
|
|
|
file_creation_started_ = false;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
void OnBackgroundError(BackgroundErrorReason /*reason*/,
|
|
|
|
|
Status* bg_error) override {
|
|
|
|
|
if (override_bg_error_) {
|
|
|
|
|
*bg_error = bg_error_;
|
|
|
|
|
override_bg_error_ = false;
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
void EnableAutoRecovery(bool enable = true) { no_auto_recovery_ = !enable; }
|
|
|
|
|
|
|
|
|
|
void OverrideBGError(Status bg_err) {
|
|
|
|
|
bg_error_ = bg_err;
|
|
|
|
|
override_bg_error_ = true;
|
|
|
|
|
}
|
|
|
|
|
|
2020-03-04 20:30:34 +00:00
|
|
|
void InjectFileCreationError(FaultInjectionTestFS* fs, int file_count,
|
|
|
|
|
IOStatus io_s) {
|
|
|
|
|
fault_fs_ = fs;
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
file_count_ = file_count;
|
2020-03-04 20:30:34 +00:00
|
|
|
file_creation_error_ = io_s;
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
}
|
|
|
|
|
|
2021-12-08 22:22:41 +00:00
|
|
|
Status new_bg_error() { return new_bg_error_; }
|
|
|
|
|
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
private:
|
|
|
|
|
InstrumentedMutex mutex_;
|
|
|
|
|
InstrumentedCondVar cv_;
|
|
|
|
|
bool no_auto_recovery_;
|
|
|
|
|
bool recovery_complete_;
|
|
|
|
|
bool file_creation_started_;
|
|
|
|
|
bool override_bg_error_;
|
|
|
|
|
int file_count_;
|
2020-03-04 20:30:34 +00:00
|
|
|
IOStatus file_creation_error_;
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
Status bg_error_;
|
2021-12-08 22:22:41 +00:00
|
|
|
Status new_bg_error_;
|
2020-03-04 20:30:34 +00:00
|
|
|
FaultInjectionTestFS* fault_fs_;
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
};
|
|
|
|
|
|
2023-10-31 23:13:36 +00:00
|
|
|
TEST_F(DBErrorHandlingFSTest, FlushWriteError) {
|
|
|
|
|
std::shared_ptr<ErrorHandlerFSListener> listener =
|
|
|
|
|
std::make_shared<ErrorHandlerFSListener>();
|
2018-06-28 19:23:57 +00:00
|
|
|
Options options = GetDefaultOptions();
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
options.env = fault_env_.get();
|
2018-06-28 19:23:57 +00:00
|
|
|
options.create_if_missing = true;
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
options.listeners.emplace_back(listener);
|
2021-03-18 05:36:42 +00:00
|
|
|
options.statistics = CreateDBStatistics();
|
2018-06-28 19:23:57 +00:00
|
|
|
Status s;
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
|
|
|
|
|
listener->EnableAutoRecovery(false);
|
2018-06-28 19:23:57 +00:00
|
|
|
DestroyAndReopen(options);
|
|
|
|
|
|
2020-10-02 23:39:17 +00:00
|
|
|
ASSERT_OK(Put(Key(0), "val"));
|
2020-03-04 20:30:34 +00:00
|
|
|
SyncPoint::GetInstance()->SetCallBack("FlushJob::Start", [&](void*) {
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
fault_fs_->SetFilesystemActive(false, IOStatus::NoSpace("Out of space"));
|
2018-06-28 19:23:57 +00:00
|
|
|
});
|
|
|
|
|
SyncPoint::GetInstance()->EnableProcessing();
|
|
|
|
|
s = Flush();
|
2020-02-20 20:07:53 +00:00
|
|
|
ASSERT_EQ(s.severity(), ROCKSDB_NAMESPACE::Status::Severity::kHardError);
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
SyncPoint::GetInstance()->DisableProcessing();
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
fault_fs_->SetFilesystemActive(true);
|
2018-06-28 19:23:57 +00:00
|
|
|
s = dbfull()->Resume();
|
2020-12-24 00:54:05 +00:00
|
|
|
ASSERT_OK(s);
|
2021-03-18 05:36:42 +00:00
|
|
|
ASSERT_EQ(1, options.statistics->getAndResetTickerCount(
|
|
|
|
|
ERROR_HANDLER_BG_ERROR_COUNT));
|
|
|
|
|
ASSERT_EQ(1, options.statistics->getAndResetTickerCount(
|
|
|
|
|
ERROR_HANDLER_BG_IO_ERROR_COUNT));
|
|
|
|
|
ASSERT_EQ(0, options.statistics->getAndResetTickerCount(
|
|
|
|
|
ERROR_HANDLER_BG_RETRYABLE_IO_ERROR_COUNT));
|
|
|
|
|
ASSERT_EQ(0, options.statistics->getAndResetTickerCount(
|
|
|
|
|
ERROR_HANDLER_AUTORESUME_COUNT));
|
|
|
|
|
ASSERT_EQ(0, options.statistics->getAndResetTickerCount(
|
|
|
|
|
ERROR_HANDLER_AUTORESUME_RETRY_TOTAL_COUNT));
|
|
|
|
|
ASSERT_EQ(0, options.statistics->getAndResetTickerCount(
|
|
|
|
|
ERROR_HANDLER_AUTORESUME_SUCCESS_COUNT));
|
2018-06-28 19:23:57 +00:00
|
|
|
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
Reopen(options);
|
|
|
|
|
ASSERT_EQ("val", Get(Key(0)));
|
2018-06-28 19:23:57 +00:00
|
|
|
Destroy(options);
|
|
|
|
|
}
|
|
|
|
|
|
All the NoSpace() errors will be handled by regular SetBGError and RecoverFromNoSpace() (#8376)
Summary:
In the current logic, any IO Error with retryable flag == true will be handled by the special logic and in most cases, StartRecoverFromRetryableBGIOError will be called to do the auto resume. If the NoSpace error with retryable flag is set during WAL write, it is mapped as a hard error, which will trigger the auto recovery. During the recover process, if write continues and append to the WAL, the write process sees that bg_error is set to HardError and it calls WriteStatusCheck(), which calls SetBGError() with Status (not IOStatus). This will redirect to the regular SetBGError interface, in which recovery_error_ will be set to the corresponding error. With the recovery_error_ set, the auto resume thread created in StartRecoverFromRetryableBGIOError will keep failing as long as user keeps trying to write.
To fix this issue. All the NoSpace error (no matter retryable flag is set or not) will be redirect to the regular SetBGError, and RecoverFromNoSpace() will do the recovery job which calls SstFileManager::StartErrorRecovery().
Pull Request resolved: https://github.com/facebook/rocksdb/pull/8376
Test Plan: make check and added the new testing case
Reviewed By: anand1976
Differential Revision: D29071828
Pulled By: zhichao-cao
fbshipit-source-id: 7171d7e14cc4620fdab49b7eff7a2fe9a89942c2
2021-06-11 21:47:40 +00:00
|
|
|
// All the NoSpace IOError will be handled as the regular BG Error no matter the
|
|
|
|
|
// retryable flag is set of not. So the auto resume for retryable IO Error will
|
|
|
|
|
// not be triggered. Also, it is mapped as hard error.
|
2023-10-31 23:13:36 +00:00
|
|
|
TEST_F(DBErrorHandlingFSTest, FlushWriteNoSpaceError) {
|
|
|
|
|
std::shared_ptr<ErrorHandlerFSListener> listener =
|
|
|
|
|
std::make_shared<ErrorHandlerFSListener>();
|
All the NoSpace() errors will be handled by regular SetBGError and RecoverFromNoSpace() (#8376)
Summary:
In the current logic, any IO Error with retryable flag == true will be handled by the special logic and in most cases, StartRecoverFromRetryableBGIOError will be called to do the auto resume. If the NoSpace error with retryable flag is set during WAL write, it is mapped as a hard error, which will trigger the auto recovery. During the recover process, if write continues and append to the WAL, the write process sees that bg_error is set to HardError and it calls WriteStatusCheck(), which calls SetBGError() with Status (not IOStatus). This will redirect to the regular SetBGError interface, in which recovery_error_ will be set to the corresponding error. With the recovery_error_ set, the auto resume thread created in StartRecoverFromRetryableBGIOError will keep failing as long as user keeps trying to write.
To fix this issue. All the NoSpace error (no matter retryable flag is set or not) will be redirect to the regular SetBGError, and RecoverFromNoSpace() will do the recovery job which calls SstFileManager::StartErrorRecovery().
Pull Request resolved: https://github.com/facebook/rocksdb/pull/8376
Test Plan: make check and added the new testing case
Reviewed By: anand1976
Differential Revision: D29071828
Pulled By: zhichao-cao
fbshipit-source-id: 7171d7e14cc4620fdab49b7eff7a2fe9a89942c2
2021-06-11 21:47:40 +00:00
|
|
|
Options options = GetDefaultOptions();
|
|
|
|
|
options.env = fault_env_.get();
|
|
|
|
|
options.create_if_missing = true;
|
|
|
|
|
options.listeners.emplace_back(listener);
|
|
|
|
|
options.max_bgerror_resume_count = 2;
|
|
|
|
|
options.bgerror_resume_retry_interval = 100000; // 0.1 second
|
|
|
|
|
options.statistics = CreateDBStatistics();
|
|
|
|
|
Status s;
|
|
|
|
|
|
|
|
|
|
listener->EnableAutoRecovery(false);
|
|
|
|
|
DestroyAndReopen(options);
|
|
|
|
|
|
|
|
|
|
IOStatus error_msg = IOStatus::NoSpace("Retryable IO Error");
|
|
|
|
|
error_msg.SetRetryable(true);
|
|
|
|
|
|
|
|
|
|
ASSERT_OK(Put(Key(1), "val1"));
|
|
|
|
|
SyncPoint::GetInstance()->SetCallBack(
|
|
|
|
|
"BuildTable:BeforeFinishBuildTable",
|
|
|
|
|
[&](void*) { fault_fs_->SetFilesystemActive(false, error_msg); });
|
|
|
|
|
SyncPoint::GetInstance()->EnableProcessing();
|
|
|
|
|
s = Flush();
|
|
|
|
|
ASSERT_EQ(s.severity(), ROCKSDB_NAMESPACE::Status::Severity::kHardError);
|
|
|
|
|
SyncPoint::GetInstance()->DisableProcessing();
|
|
|
|
|
fault_fs_->SetFilesystemActive(true);
|
|
|
|
|
s = dbfull()->Resume();
|
|
|
|
|
ASSERT_OK(s);
|
|
|
|
|
ASSERT_EQ(1, options.statistics->getAndResetTickerCount(
|
|
|
|
|
ERROR_HANDLER_BG_ERROR_COUNT));
|
|
|
|
|
ASSERT_EQ(1, options.statistics->getAndResetTickerCount(
|
|
|
|
|
ERROR_HANDLER_BG_IO_ERROR_COUNT));
|
|
|
|
|
ASSERT_EQ(0, options.statistics->getAndResetTickerCount(
|
|
|
|
|
ERROR_HANDLER_BG_RETRYABLE_IO_ERROR_COUNT));
|
|
|
|
|
ASSERT_EQ(0, options.statistics->getAndResetTickerCount(
|
|
|
|
|
ERROR_HANDLER_AUTORESUME_COUNT));
|
|
|
|
|
ASSERT_EQ(0, options.statistics->getAndResetTickerCount(
|
|
|
|
|
ERROR_HANDLER_AUTORESUME_RETRY_TOTAL_COUNT));
|
|
|
|
|
ASSERT_EQ(0, options.statistics->getAndResetTickerCount(
|
|
|
|
|
ERROR_HANDLER_AUTORESUME_SUCCESS_COUNT));
|
|
|
|
|
Destroy(options);
|
|
|
|
|
}
|
|
|
|
|
|
2023-10-31 23:13:36 +00:00
|
|
|
TEST_F(DBErrorHandlingFSTest, FlushWriteRetryableError) {
|
|
|
|
|
std::shared_ptr<ErrorHandlerFSListener> listener =
|
|
|
|
|
std::make_shared<ErrorHandlerFSListener>();
|
Pass IOStatus to write path and set retryable IO Error as hard error in BG jobs (#6487)
Summary:
In the current code base, we use Status to get and store the returned status from the call. Specifically, for IO related functions, the current Status cannot reflect the IO Error details such as error scope, error retryable attribute, and others. With the implementation of https://github.com/facebook/rocksdb/issues/5761, we have the new Wrapper for IO, which returns IOStatus instead of Status. However, the IOStatus is purged at the lower level of write path and transferred to Status.
The first job of this PR is to pass the IOStatus to the write path (flush, WAL write, and Compaction). The second job is to identify the Retryable IO Error as HardError, and set the bg_error_ as HardError. In this case, the DB Instance becomes read only. User is informed of the Status and need to take actions to deal with it (e.g., call db->Resume()).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6487
Test Plan: Added the testing case to error_handler_fs_test. Pass make asan_check
Reviewed By: anand1976
Differential Revision: D20685017
Pulled By: zhichao-cao
fbshipit-source-id: ff85f042896243abcd6ef37877834e26f36b6eb0
2020-03-27 23:03:05 +00:00
|
|
|
Options options = GetDefaultOptions();
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
options.env = fault_env_.get();
|
Pass IOStatus to write path and set retryable IO Error as hard error in BG jobs (#6487)
Summary:
In the current code base, we use Status to get and store the returned status from the call. Specifically, for IO related functions, the current Status cannot reflect the IO Error details such as error scope, error retryable attribute, and others. With the implementation of https://github.com/facebook/rocksdb/issues/5761, we have the new Wrapper for IO, which returns IOStatus instead of Status. However, the IOStatus is purged at the lower level of write path and transferred to Status.
The first job of this PR is to pass the IOStatus to the write path (flush, WAL write, and Compaction). The second job is to identify the Retryable IO Error as HardError, and set the bg_error_ as HardError. In this case, the DB Instance becomes read only. User is informed of the Status and need to take actions to deal with it (e.g., call db->Resume()).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6487
Test Plan: Added the testing case to error_handler_fs_test. Pass make asan_check
Reviewed By: anand1976
Differential Revision: D20685017
Pulled By: zhichao-cao
fbshipit-source-id: ff85f042896243abcd6ef37877834e26f36b6eb0
2020-03-27 23:03:05 +00:00
|
|
|
options.create_if_missing = true;
|
|
|
|
|
options.listeners.emplace_back(listener);
|
2020-07-15 18:02:44 +00:00
|
|
|
options.max_bgerror_resume_count = 0;
|
2021-03-18 05:36:42 +00:00
|
|
|
options.statistics = CreateDBStatistics();
|
Pass IOStatus to write path and set retryable IO Error as hard error in BG jobs (#6487)
Summary:
In the current code base, we use Status to get and store the returned status from the call. Specifically, for IO related functions, the current Status cannot reflect the IO Error details such as error scope, error retryable attribute, and others. With the implementation of https://github.com/facebook/rocksdb/issues/5761, we have the new Wrapper for IO, which returns IOStatus instead of Status. However, the IOStatus is purged at the lower level of write path and transferred to Status.
The first job of this PR is to pass the IOStatus to the write path (flush, WAL write, and Compaction). The second job is to identify the Retryable IO Error as HardError, and set the bg_error_ as HardError. In this case, the DB Instance becomes read only. User is informed of the Status and need to take actions to deal with it (e.g., call db->Resume()).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6487
Test Plan: Added the testing case to error_handler_fs_test. Pass make asan_check
Reviewed By: anand1976
Differential Revision: D20685017
Pulled By: zhichao-cao
fbshipit-source-id: ff85f042896243abcd6ef37877834e26f36b6eb0
2020-03-27 23:03:05 +00:00
|
|
|
Status s;
|
|
|
|
|
|
|
|
|
|
listener->EnableAutoRecovery(false);
|
|
|
|
|
DestroyAndReopen(options);
|
|
|
|
|
|
|
|
|
|
IOStatus error_msg = IOStatus::IOError("Retryable IO Error");
|
|
|
|
|
error_msg.SetRetryable(true);
|
|
|
|
|
|
2020-10-02 23:39:17 +00:00
|
|
|
ASSERT_OK(Put(Key(1), "val1"));
|
Pass IOStatus to write path and set retryable IO Error as hard error in BG jobs (#6487)
Summary:
In the current code base, we use Status to get and store the returned status from the call. Specifically, for IO related functions, the current Status cannot reflect the IO Error details such as error scope, error retryable attribute, and others. With the implementation of https://github.com/facebook/rocksdb/issues/5761, we have the new Wrapper for IO, which returns IOStatus instead of Status. However, the IOStatus is purged at the lower level of write path and transferred to Status.
The first job of this PR is to pass the IOStatus to the write path (flush, WAL write, and Compaction). The second job is to identify the Retryable IO Error as HardError, and set the bg_error_ as HardError. In this case, the DB Instance becomes read only. User is informed of the Status and need to take actions to deal with it (e.g., call db->Resume()).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6487
Test Plan: Added the testing case to error_handler_fs_test. Pass make asan_check
Reviewed By: anand1976
Differential Revision: D20685017
Pulled By: zhichao-cao
fbshipit-source-id: ff85f042896243abcd6ef37877834e26f36b6eb0
2020-03-27 23:03:05 +00:00
|
|
|
SyncPoint::GetInstance()->SetCallBack(
|
|
|
|
|
"BuildTable:BeforeFinishBuildTable",
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
[&](void*) { fault_fs_->SetFilesystemActive(false, error_msg); });
|
Pass IOStatus to write path and set retryable IO Error as hard error in BG jobs (#6487)
Summary:
In the current code base, we use Status to get and store the returned status from the call. Specifically, for IO related functions, the current Status cannot reflect the IO Error details such as error scope, error retryable attribute, and others. With the implementation of https://github.com/facebook/rocksdb/issues/5761, we have the new Wrapper for IO, which returns IOStatus instead of Status. However, the IOStatus is purged at the lower level of write path and transferred to Status.
The first job of this PR is to pass the IOStatus to the write path (flush, WAL write, and Compaction). The second job is to identify the Retryable IO Error as HardError, and set the bg_error_ as HardError. In this case, the DB Instance becomes read only. User is informed of the Status and need to take actions to deal with it (e.g., call db->Resume()).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6487
Test Plan: Added the testing case to error_handler_fs_test. Pass make asan_check
Reviewed By: anand1976
Differential Revision: D20685017
Pulled By: zhichao-cao
fbshipit-source-id: ff85f042896243abcd6ef37877834e26f36b6eb0
2020-03-27 23:03:05 +00:00
|
|
|
SyncPoint::GetInstance()->EnableProcessing();
|
|
|
|
|
s = Flush();
|
2021-03-26 04:41:44 +00:00
|
|
|
ASSERT_EQ(s.severity(), ROCKSDB_NAMESPACE::Status::Severity::kSoftError);
|
Pass IOStatus to write path and set retryable IO Error as hard error in BG jobs (#6487)
Summary:
In the current code base, we use Status to get and store the returned status from the call. Specifically, for IO related functions, the current Status cannot reflect the IO Error details such as error scope, error retryable attribute, and others. With the implementation of https://github.com/facebook/rocksdb/issues/5761, we have the new Wrapper for IO, which returns IOStatus instead of Status. However, the IOStatus is purged at the lower level of write path and transferred to Status.
The first job of this PR is to pass the IOStatus to the write path (flush, WAL write, and Compaction). The second job is to identify the Retryable IO Error as HardError, and set the bg_error_ as HardError. In this case, the DB Instance becomes read only. User is informed of the Status and need to take actions to deal with it (e.g., call db->Resume()).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6487
Test Plan: Added the testing case to error_handler_fs_test. Pass make asan_check
Reviewed By: anand1976
Differential Revision: D20685017
Pulled By: zhichao-cao
fbshipit-source-id: ff85f042896243abcd6ef37877834e26f36b6eb0
2020-03-27 23:03:05 +00:00
|
|
|
SyncPoint::GetInstance()->DisableProcessing();
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
fault_fs_->SetFilesystemActive(true);
|
Pass IOStatus to write path and set retryable IO Error as hard error in BG jobs (#6487)
Summary:
In the current code base, we use Status to get and store the returned status from the call. Specifically, for IO related functions, the current Status cannot reflect the IO Error details such as error scope, error retryable attribute, and others. With the implementation of https://github.com/facebook/rocksdb/issues/5761, we have the new Wrapper for IO, which returns IOStatus instead of Status. However, the IOStatus is purged at the lower level of write path and transferred to Status.
The first job of this PR is to pass the IOStatus to the write path (flush, WAL write, and Compaction). The second job is to identify the Retryable IO Error as HardError, and set the bg_error_ as HardError. In this case, the DB Instance becomes read only. User is informed of the Status and need to take actions to deal with it (e.g., call db->Resume()).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6487
Test Plan: Added the testing case to error_handler_fs_test. Pass make asan_check
Reviewed By: anand1976
Differential Revision: D20685017
Pulled By: zhichao-cao
fbshipit-source-id: ff85f042896243abcd6ef37877834e26f36b6eb0
2020-03-27 23:03:05 +00:00
|
|
|
s = dbfull()->Resume();
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
ASSERT_OK(s);
|
2021-03-18 05:36:42 +00:00
|
|
|
ASSERT_EQ(1, options.statistics->getAndResetTickerCount(
|
|
|
|
|
ERROR_HANDLER_BG_ERROR_COUNT));
|
|
|
|
|
ASSERT_EQ(1, options.statistics->getAndResetTickerCount(
|
|
|
|
|
ERROR_HANDLER_BG_IO_ERROR_COUNT));
|
|
|
|
|
ASSERT_EQ(1, options.statistics->getAndResetTickerCount(
|
|
|
|
|
ERROR_HANDLER_BG_RETRYABLE_IO_ERROR_COUNT));
|
|
|
|
|
ASSERT_EQ(0, options.statistics->getAndResetTickerCount(
|
|
|
|
|
ERROR_HANDLER_AUTORESUME_COUNT));
|
|
|
|
|
ASSERT_EQ(0, options.statistics->getAndResetTickerCount(
|
|
|
|
|
ERROR_HANDLER_AUTORESUME_RETRY_TOTAL_COUNT));
|
|
|
|
|
ASSERT_EQ(0, options.statistics->getAndResetTickerCount(
|
|
|
|
|
ERROR_HANDLER_AUTORESUME_SUCCESS_COUNT));
|
Pass IOStatus to write path and set retryable IO Error as hard error in BG jobs (#6487)
Summary:
In the current code base, we use Status to get and store the returned status from the call. Specifically, for IO related functions, the current Status cannot reflect the IO Error details such as error scope, error retryable attribute, and others. With the implementation of https://github.com/facebook/rocksdb/issues/5761, we have the new Wrapper for IO, which returns IOStatus instead of Status. However, the IOStatus is purged at the lower level of write path and transferred to Status.
The first job of this PR is to pass the IOStatus to the write path (flush, WAL write, and Compaction). The second job is to identify the Retryable IO Error as HardError, and set the bg_error_ as HardError. In this case, the DB Instance becomes read only. User is informed of the Status and need to take actions to deal with it (e.g., call db->Resume()).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6487
Test Plan: Added the testing case to error_handler_fs_test. Pass make asan_check
Reviewed By: anand1976
Differential Revision: D20685017
Pulled By: zhichao-cao
fbshipit-source-id: ff85f042896243abcd6ef37877834e26f36b6eb0
2020-03-27 23:03:05 +00:00
|
|
|
Reopen(options);
|
|
|
|
|
ASSERT_EQ("val1", Get(Key(1)));
|
|
|
|
|
|
2020-10-02 23:39:17 +00:00
|
|
|
ASSERT_OK(Put(Key(2), "val2"));
|
Pass IOStatus to write path and set retryable IO Error as hard error in BG jobs (#6487)
Summary:
In the current code base, we use Status to get and store the returned status from the call. Specifically, for IO related functions, the current Status cannot reflect the IO Error details such as error scope, error retryable attribute, and others. With the implementation of https://github.com/facebook/rocksdb/issues/5761, we have the new Wrapper for IO, which returns IOStatus instead of Status. However, the IOStatus is purged at the lower level of write path and transferred to Status.
The first job of this PR is to pass the IOStatus to the write path (flush, WAL write, and Compaction). The second job is to identify the Retryable IO Error as HardError, and set the bg_error_ as HardError. In this case, the DB Instance becomes read only. User is informed of the Status and need to take actions to deal with it (e.g., call db->Resume()).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6487
Test Plan: Added the testing case to error_handler_fs_test. Pass make asan_check
Reviewed By: anand1976
Differential Revision: D20685017
Pulled By: zhichao-cao
fbshipit-source-id: ff85f042896243abcd6ef37877834e26f36b6eb0
2020-03-27 23:03:05 +00:00
|
|
|
SyncPoint::GetInstance()->SetCallBack(
|
|
|
|
|
"BuildTable:BeforeSyncTable",
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
[&](void*) { fault_fs_->SetFilesystemActive(false, error_msg); });
|
Pass IOStatus to write path and set retryable IO Error as hard error in BG jobs (#6487)
Summary:
In the current code base, we use Status to get and store the returned status from the call. Specifically, for IO related functions, the current Status cannot reflect the IO Error details such as error scope, error retryable attribute, and others. With the implementation of https://github.com/facebook/rocksdb/issues/5761, we have the new Wrapper for IO, which returns IOStatus instead of Status. However, the IOStatus is purged at the lower level of write path and transferred to Status.
The first job of this PR is to pass the IOStatus to the write path (flush, WAL write, and Compaction). The second job is to identify the Retryable IO Error as HardError, and set the bg_error_ as HardError. In this case, the DB Instance becomes read only. User is informed of the Status and need to take actions to deal with it (e.g., call db->Resume()).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6487
Test Plan: Added the testing case to error_handler_fs_test. Pass make asan_check
Reviewed By: anand1976
Differential Revision: D20685017
Pulled By: zhichao-cao
fbshipit-source-id: ff85f042896243abcd6ef37877834e26f36b6eb0
2020-03-27 23:03:05 +00:00
|
|
|
SyncPoint::GetInstance()->EnableProcessing();
|
|
|
|
|
s = Flush();
|
2021-03-26 04:41:44 +00:00
|
|
|
ASSERT_EQ(s.severity(), ROCKSDB_NAMESPACE::Status::Severity::kSoftError);
|
Pass IOStatus to write path and set retryable IO Error as hard error in BG jobs (#6487)
Summary:
In the current code base, we use Status to get and store the returned status from the call. Specifically, for IO related functions, the current Status cannot reflect the IO Error details such as error scope, error retryable attribute, and others. With the implementation of https://github.com/facebook/rocksdb/issues/5761, we have the new Wrapper for IO, which returns IOStatus instead of Status. However, the IOStatus is purged at the lower level of write path and transferred to Status.
The first job of this PR is to pass the IOStatus to the write path (flush, WAL write, and Compaction). The second job is to identify the Retryable IO Error as HardError, and set the bg_error_ as HardError. In this case, the DB Instance becomes read only. User is informed of the Status and need to take actions to deal with it (e.g., call db->Resume()).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6487
Test Plan: Added the testing case to error_handler_fs_test. Pass make asan_check
Reviewed By: anand1976
Differential Revision: D20685017
Pulled By: zhichao-cao
fbshipit-source-id: ff85f042896243abcd6ef37877834e26f36b6eb0
2020-03-27 23:03:05 +00:00
|
|
|
SyncPoint::GetInstance()->DisableProcessing();
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
fault_fs_->SetFilesystemActive(true);
|
Pass IOStatus to write path and set retryable IO Error as hard error in BG jobs (#6487)
Summary:
In the current code base, we use Status to get and store the returned status from the call. Specifically, for IO related functions, the current Status cannot reflect the IO Error details such as error scope, error retryable attribute, and others. With the implementation of https://github.com/facebook/rocksdb/issues/5761, we have the new Wrapper for IO, which returns IOStatus instead of Status. However, the IOStatus is purged at the lower level of write path and transferred to Status.
The first job of this PR is to pass the IOStatus to the write path (flush, WAL write, and Compaction). The second job is to identify the Retryable IO Error as HardError, and set the bg_error_ as HardError. In this case, the DB Instance becomes read only. User is informed of the Status and need to take actions to deal with it (e.g., call db->Resume()).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6487
Test Plan: Added the testing case to error_handler_fs_test. Pass make asan_check
Reviewed By: anand1976
Differential Revision: D20685017
Pulled By: zhichao-cao
fbshipit-source-id: ff85f042896243abcd6ef37877834e26f36b6eb0
2020-03-27 23:03:05 +00:00
|
|
|
s = dbfull()->Resume();
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
ASSERT_OK(s);
|
Pass IOStatus to write path and set retryable IO Error as hard error in BG jobs (#6487)
Summary:
In the current code base, we use Status to get and store the returned status from the call. Specifically, for IO related functions, the current Status cannot reflect the IO Error details such as error scope, error retryable attribute, and others. With the implementation of https://github.com/facebook/rocksdb/issues/5761, we have the new Wrapper for IO, which returns IOStatus instead of Status. However, the IOStatus is purged at the lower level of write path and transferred to Status.
The first job of this PR is to pass the IOStatus to the write path (flush, WAL write, and Compaction). The second job is to identify the Retryable IO Error as HardError, and set the bg_error_ as HardError. In this case, the DB Instance becomes read only. User is informed of the Status and need to take actions to deal with it (e.g., call db->Resume()).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6487
Test Plan: Added the testing case to error_handler_fs_test. Pass make asan_check
Reviewed By: anand1976
Differential Revision: D20685017
Pulled By: zhichao-cao
fbshipit-source-id: ff85f042896243abcd6ef37877834e26f36b6eb0
2020-03-27 23:03:05 +00:00
|
|
|
Reopen(options);
|
|
|
|
|
ASSERT_EQ("val2", Get(Key(2)));
|
|
|
|
|
|
2020-10-02 23:39:17 +00:00
|
|
|
ASSERT_OK(Put(Key(3), "val3"));
|
Pass IOStatus to write path and set retryable IO Error as hard error in BG jobs (#6487)
Summary:
In the current code base, we use Status to get and store the returned status from the call. Specifically, for IO related functions, the current Status cannot reflect the IO Error details such as error scope, error retryable attribute, and others. With the implementation of https://github.com/facebook/rocksdb/issues/5761, we have the new Wrapper for IO, which returns IOStatus instead of Status. However, the IOStatus is purged at the lower level of write path and transferred to Status.
The first job of this PR is to pass the IOStatus to the write path (flush, WAL write, and Compaction). The second job is to identify the Retryable IO Error as HardError, and set the bg_error_ as HardError. In this case, the DB Instance becomes read only. User is informed of the Status and need to take actions to deal with it (e.g., call db->Resume()).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6487
Test Plan: Added the testing case to error_handler_fs_test. Pass make asan_check
Reviewed By: anand1976
Differential Revision: D20685017
Pulled By: zhichao-cao
fbshipit-source-id: ff85f042896243abcd6ef37877834e26f36b6eb0
2020-03-27 23:03:05 +00:00
|
|
|
SyncPoint::GetInstance()->SetCallBack(
|
|
|
|
|
"BuildTable:BeforeCloseTableFile",
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
[&](void*) { fault_fs_->SetFilesystemActive(false, error_msg); });
|
Pass IOStatus to write path and set retryable IO Error as hard error in BG jobs (#6487)
Summary:
In the current code base, we use Status to get and store the returned status from the call. Specifically, for IO related functions, the current Status cannot reflect the IO Error details such as error scope, error retryable attribute, and others. With the implementation of https://github.com/facebook/rocksdb/issues/5761, we have the new Wrapper for IO, which returns IOStatus instead of Status. However, the IOStatus is purged at the lower level of write path and transferred to Status.
The first job of this PR is to pass the IOStatus to the write path (flush, WAL write, and Compaction). The second job is to identify the Retryable IO Error as HardError, and set the bg_error_ as HardError. In this case, the DB Instance becomes read only. User is informed of the Status and need to take actions to deal with it (e.g., call db->Resume()).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6487
Test Plan: Added the testing case to error_handler_fs_test. Pass make asan_check
Reviewed By: anand1976
Differential Revision: D20685017
Pulled By: zhichao-cao
fbshipit-source-id: ff85f042896243abcd6ef37877834e26f36b6eb0
2020-03-27 23:03:05 +00:00
|
|
|
SyncPoint::GetInstance()->EnableProcessing();
|
|
|
|
|
s = Flush();
|
2021-03-26 04:41:44 +00:00
|
|
|
ASSERT_EQ(s.severity(), ROCKSDB_NAMESPACE::Status::Severity::kSoftError);
|
Pass IOStatus to write path and set retryable IO Error as hard error in BG jobs (#6487)
Summary:
In the current code base, we use Status to get and store the returned status from the call. Specifically, for IO related functions, the current Status cannot reflect the IO Error details such as error scope, error retryable attribute, and others. With the implementation of https://github.com/facebook/rocksdb/issues/5761, we have the new Wrapper for IO, which returns IOStatus instead of Status. However, the IOStatus is purged at the lower level of write path and transferred to Status.
The first job of this PR is to pass the IOStatus to the write path (flush, WAL write, and Compaction). The second job is to identify the Retryable IO Error as HardError, and set the bg_error_ as HardError. In this case, the DB Instance becomes read only. User is informed of the Status and need to take actions to deal with it (e.g., call db->Resume()).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6487
Test Plan: Added the testing case to error_handler_fs_test. Pass make asan_check
Reviewed By: anand1976
Differential Revision: D20685017
Pulled By: zhichao-cao
fbshipit-source-id: ff85f042896243abcd6ef37877834e26f36b6eb0
2020-03-27 23:03:05 +00:00
|
|
|
SyncPoint::GetInstance()->DisableProcessing();
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
fault_fs_->SetFilesystemActive(true);
|
Pass IOStatus to write path and set retryable IO Error as hard error in BG jobs (#6487)
Summary:
In the current code base, we use Status to get and store the returned status from the call. Specifically, for IO related functions, the current Status cannot reflect the IO Error details such as error scope, error retryable attribute, and others. With the implementation of https://github.com/facebook/rocksdb/issues/5761, we have the new Wrapper for IO, which returns IOStatus instead of Status. However, the IOStatus is purged at the lower level of write path and transferred to Status.
The first job of this PR is to pass the IOStatus to the write path (flush, WAL write, and Compaction). The second job is to identify the Retryable IO Error as HardError, and set the bg_error_ as HardError. In this case, the DB Instance becomes read only. User is informed of the Status and need to take actions to deal with it (e.g., call db->Resume()).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6487
Test Plan: Added the testing case to error_handler_fs_test. Pass make asan_check
Reviewed By: anand1976
Differential Revision: D20685017
Pulled By: zhichao-cao
fbshipit-source-id: ff85f042896243abcd6ef37877834e26f36b6eb0
2020-03-27 23:03:05 +00:00
|
|
|
s = dbfull()->Resume();
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
ASSERT_OK(s);
|
Pass IOStatus to write path and set retryable IO Error as hard error in BG jobs (#6487)
Summary:
In the current code base, we use Status to get and store the returned status from the call. Specifically, for IO related functions, the current Status cannot reflect the IO Error details such as error scope, error retryable attribute, and others. With the implementation of https://github.com/facebook/rocksdb/issues/5761, we have the new Wrapper for IO, which returns IOStatus instead of Status. However, the IOStatus is purged at the lower level of write path and transferred to Status.
The first job of this PR is to pass the IOStatus to the write path (flush, WAL write, and Compaction). The second job is to identify the Retryable IO Error as HardError, and set the bg_error_ as HardError. In this case, the DB Instance becomes read only. User is informed of the Status and need to take actions to deal with it (e.g., call db->Resume()).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6487
Test Plan: Added the testing case to error_handler_fs_test. Pass make asan_check
Reviewed By: anand1976
Differential Revision: D20685017
Pulled By: zhichao-cao
fbshipit-source-id: ff85f042896243abcd6ef37877834e26f36b6eb0
2020-03-27 23:03:05 +00:00
|
|
|
Reopen(options);
|
|
|
|
|
ASSERT_EQ("val3", Get(Key(3)));
|
|
|
|
|
|
|
|
|
|
Destroy(options);
|
|
|
|
|
}
|
|
|
|
|
|
2023-10-31 23:13:36 +00:00
|
|
|
TEST_F(DBErrorHandlingFSTest, FlushWriteFileScopeError) {
|
|
|
|
|
std::shared_ptr<ErrorHandlerFSListener> listener =
|
|
|
|
|
std::make_shared<ErrorHandlerFSListener>();
|
2021-01-08 00:30:06 +00:00
|
|
|
Options options = GetDefaultOptions();
|
|
|
|
|
options.env = fault_env_.get();
|
|
|
|
|
options.create_if_missing = true;
|
|
|
|
|
options.listeners.emplace_back(listener);
|
|
|
|
|
options.max_bgerror_resume_count = 0;
|
|
|
|
|
Status s;
|
|
|
|
|
|
|
|
|
|
listener->EnableAutoRecovery(false);
|
|
|
|
|
DestroyAndReopen(options);
|
|
|
|
|
|
|
|
|
|
IOStatus error_msg = IOStatus::IOError("File Scope Data Loss Error");
|
|
|
|
|
error_msg.SetDataLoss(true);
|
|
|
|
|
error_msg.SetScope(
|
|
|
|
|
ROCKSDB_NAMESPACE::IOStatus::IOErrorScope::kIOErrorScopeFile);
|
|
|
|
|
error_msg.SetRetryable(false);
|
|
|
|
|
|
|
|
|
|
ASSERT_OK(Put(Key(1), "val1"));
|
|
|
|
|
SyncPoint::GetInstance()->SetCallBack(
|
|
|
|
|
"BuildTable:BeforeFinishBuildTable",
|
|
|
|
|
[&](void*) { fault_fs_->SetFilesystemActive(false, error_msg); });
|
|
|
|
|
SyncPoint::GetInstance()->EnableProcessing();
|
|
|
|
|
s = Flush();
|
2021-03-26 04:41:44 +00:00
|
|
|
ASSERT_EQ(s.severity(), ROCKSDB_NAMESPACE::Status::Severity::kSoftError);
|
2021-01-08 00:30:06 +00:00
|
|
|
SyncPoint::GetInstance()->DisableProcessing();
|
|
|
|
|
fault_fs_->SetFilesystemActive(true);
|
|
|
|
|
s = dbfull()->Resume();
|
|
|
|
|
ASSERT_OK(s);
|
|
|
|
|
Reopen(options);
|
|
|
|
|
ASSERT_EQ("val1", Get(Key(1)));
|
|
|
|
|
|
|
|
|
|
ASSERT_OK(Put(Key(2), "val2"));
|
|
|
|
|
SyncPoint::GetInstance()->SetCallBack(
|
|
|
|
|
"BuildTable:BeforeSyncTable",
|
|
|
|
|
[&](void*) { fault_fs_->SetFilesystemActive(false, error_msg); });
|
|
|
|
|
SyncPoint::GetInstance()->EnableProcessing();
|
|
|
|
|
s = Flush();
|
2021-03-26 04:41:44 +00:00
|
|
|
ASSERT_EQ(s.severity(), ROCKSDB_NAMESPACE::Status::Severity::kSoftError);
|
2021-01-08 00:30:06 +00:00
|
|
|
SyncPoint::GetInstance()->DisableProcessing();
|
|
|
|
|
fault_fs_->SetFilesystemActive(true);
|
|
|
|
|
s = dbfull()->Resume();
|
|
|
|
|
ASSERT_OK(s);
|
|
|
|
|
Reopen(options);
|
|
|
|
|
ASSERT_EQ("val2", Get(Key(2)));
|
|
|
|
|
|
|
|
|
|
ASSERT_OK(Put(Key(3), "val3"));
|
|
|
|
|
SyncPoint::GetInstance()->SetCallBack(
|
|
|
|
|
"BuildTable:BeforeCloseTableFile",
|
|
|
|
|
[&](void*) { fault_fs_->SetFilesystemActive(false, error_msg); });
|
|
|
|
|
SyncPoint::GetInstance()->EnableProcessing();
|
|
|
|
|
s = Flush();
|
2021-03-26 04:41:44 +00:00
|
|
|
ASSERT_EQ(s.severity(), ROCKSDB_NAMESPACE::Status::Severity::kSoftError);
|
2021-01-08 00:30:06 +00:00
|
|
|
SyncPoint::GetInstance()->DisableProcessing();
|
|
|
|
|
fault_fs_->SetFilesystemActive(true);
|
|
|
|
|
s = dbfull()->Resume();
|
|
|
|
|
ASSERT_OK(s);
|
|
|
|
|
Reopen(options);
|
|
|
|
|
ASSERT_EQ("val3", Get(Key(3)));
|
|
|
|
|
|
|
|
|
|
// not file scope, but retyrable set
|
|
|
|
|
error_msg.SetDataLoss(false);
|
|
|
|
|
error_msg.SetScope(
|
|
|
|
|
ROCKSDB_NAMESPACE::IOStatus::IOErrorScope::kIOErrorScopeFileSystem);
|
|
|
|
|
error_msg.SetRetryable(true);
|
|
|
|
|
|
|
|
|
|
ASSERT_OK(Put(Key(3), "val3"));
|
|
|
|
|
SyncPoint::GetInstance()->SetCallBack(
|
|
|
|
|
"BuildTable:BeforeCloseTableFile",
|
|
|
|
|
[&](void*) { fault_fs_->SetFilesystemActive(false, error_msg); });
|
|
|
|
|
SyncPoint::GetInstance()->EnableProcessing();
|
|
|
|
|
s = Flush();
|
2021-03-26 04:41:44 +00:00
|
|
|
ASSERT_EQ(s.severity(), ROCKSDB_NAMESPACE::Status::Severity::kSoftError);
|
2021-01-08 00:30:06 +00:00
|
|
|
SyncPoint::GetInstance()->DisableProcessing();
|
|
|
|
|
fault_fs_->SetFilesystemActive(true);
|
|
|
|
|
s = dbfull()->Resume();
|
|
|
|
|
ASSERT_OK(s);
|
|
|
|
|
Reopen(options);
|
|
|
|
|
ASSERT_EQ("val3", Get(Key(3)));
|
|
|
|
|
|
|
|
|
|
Destroy(options);
|
|
|
|
|
}
|
|
|
|
|
|
2023-10-31 23:13:36 +00:00
|
|
|
TEST_F(DBErrorHandlingFSTest, FlushWALWriteRetryableError) {
|
|
|
|
|
std::shared_ptr<ErrorHandlerFSListener> listener =
|
|
|
|
|
std::make_shared<ErrorHandlerFSListener>();
|
2021-03-18 21:31:30 +00:00
|
|
|
Options options = GetDefaultOptions();
|
|
|
|
|
options.env = fault_env_.get();
|
|
|
|
|
options.create_if_missing = true;
|
|
|
|
|
options.listeners.emplace_back(listener);
|
|
|
|
|
options.max_bgerror_resume_count = 0;
|
|
|
|
|
Status s;
|
|
|
|
|
|
|
|
|
|
IOStatus error_msg = IOStatus::IOError("Retryable IO Error");
|
|
|
|
|
error_msg.SetRetryable(true);
|
|
|
|
|
|
|
|
|
|
listener->EnableAutoRecovery(false);
|
|
|
|
|
SyncPoint::GetInstance()->SetCallBack(
|
2024-05-30 21:53:13 +00:00
|
|
|
"DBImpl::SyncClosedWals:Start",
|
2021-03-18 21:31:30 +00:00
|
|
|
[&](void*) { fault_fs_->SetFilesystemActive(false, error_msg); });
|
|
|
|
|
SyncPoint::GetInstance()->EnableProcessing();
|
|
|
|
|
|
|
|
|
|
CreateAndReopenWithCF({"pikachu, sdfsdfsdf"}, options);
|
|
|
|
|
|
|
|
|
|
WriteOptions wo = WriteOptions();
|
|
|
|
|
wo.disableWAL = false;
|
|
|
|
|
ASSERT_OK(Put(Key(1), "val1", wo));
|
|
|
|
|
|
|
|
|
|
s = Flush();
|
|
|
|
|
ASSERT_EQ(s.severity(), ROCKSDB_NAMESPACE::Status::Severity::kHardError);
|
|
|
|
|
SyncPoint::GetInstance()->DisableProcessing();
|
|
|
|
|
fault_fs_->SetFilesystemActive(true);
|
|
|
|
|
auto cfh = dbfull()->GetColumnFamilyHandle(1);
|
|
|
|
|
s = dbfull()->DropColumnFamily(cfh);
|
|
|
|
|
|
|
|
|
|
s = dbfull()->Resume();
|
|
|
|
|
ASSERT_OK(s);
|
|
|
|
|
ASSERT_EQ("val1", Get(Key(1)));
|
|
|
|
|
ASSERT_OK(Put(Key(3), "val3", wo));
|
|
|
|
|
ASSERT_EQ("val3", Get(Key(3)));
|
|
|
|
|
s = Flush();
|
|
|
|
|
ASSERT_OK(s);
|
|
|
|
|
ASSERT_EQ("val3", Get(Key(3)));
|
|
|
|
|
|
|
|
|
|
Destroy(options);
|
|
|
|
|
}
|
|
|
|
|
|
2023-10-31 23:13:36 +00:00
|
|
|
TEST_F(DBErrorHandlingFSTest, FlushWALAtomicWriteRetryableError) {
|
|
|
|
|
std::shared_ptr<ErrorHandlerFSListener> listener =
|
|
|
|
|
std::make_shared<ErrorHandlerFSListener>();
|
2021-03-18 21:31:30 +00:00
|
|
|
Options options = GetDefaultOptions();
|
|
|
|
|
options.env = fault_env_.get();
|
|
|
|
|
options.create_if_missing = true;
|
|
|
|
|
options.listeners.emplace_back(listener);
|
|
|
|
|
options.max_bgerror_resume_count = 0;
|
|
|
|
|
options.atomic_flush = true;
|
|
|
|
|
Status s;
|
|
|
|
|
|
|
|
|
|
IOStatus error_msg = IOStatus::IOError("Retryable IO Error");
|
|
|
|
|
error_msg.SetRetryable(true);
|
|
|
|
|
|
|
|
|
|
listener->EnableAutoRecovery(false);
|
|
|
|
|
SyncPoint::GetInstance()->SetCallBack(
|
2024-05-30 21:53:13 +00:00
|
|
|
"DBImpl::SyncClosedWals:Start",
|
2021-03-18 21:31:30 +00:00
|
|
|
[&](void*) { fault_fs_->SetFilesystemActive(false, error_msg); });
|
|
|
|
|
SyncPoint::GetInstance()->EnableProcessing();
|
|
|
|
|
|
|
|
|
|
CreateAndReopenWithCF({"pikachu, sdfsdfsdf"}, options);
|
|
|
|
|
|
|
|
|
|
WriteOptions wo = WriteOptions();
|
|
|
|
|
wo.disableWAL = false;
|
|
|
|
|
ASSERT_OK(Put(Key(1), "val1", wo));
|
|
|
|
|
|
|
|
|
|
s = Flush();
|
|
|
|
|
ASSERT_EQ(s.severity(), ROCKSDB_NAMESPACE::Status::Severity::kHardError);
|
|
|
|
|
SyncPoint::GetInstance()->DisableProcessing();
|
|
|
|
|
fault_fs_->SetFilesystemActive(true);
|
|
|
|
|
auto cfh = dbfull()->GetColumnFamilyHandle(1);
|
|
|
|
|
s = dbfull()->DropColumnFamily(cfh);
|
|
|
|
|
|
|
|
|
|
s = dbfull()->Resume();
|
|
|
|
|
ASSERT_OK(s);
|
|
|
|
|
ASSERT_EQ("val1", Get(Key(1)));
|
|
|
|
|
ASSERT_OK(Put(Key(3), "val3", wo));
|
|
|
|
|
ASSERT_EQ("val3", Get(Key(3)));
|
|
|
|
|
s = Flush();
|
|
|
|
|
ASSERT_OK(s);
|
|
|
|
|
ASSERT_EQ("val3", Get(Key(3)));
|
|
|
|
|
|
|
|
|
|
Destroy(options);
|
|
|
|
|
}
|
|
|
|
|
|
2021-11-08 23:47:53 +00:00
|
|
|
// The flush error is injected before we finish the table build
|
2023-10-31 23:13:36 +00:00
|
|
|
TEST_F(DBErrorHandlingFSTest, FlushWritNoWALRetryableError1) {
|
|
|
|
|
std::shared_ptr<ErrorHandlerFSListener> listener =
|
|
|
|
|
std::make_shared<ErrorHandlerFSListener>();
|
2020-09-18 03:22:35 +00:00
|
|
|
Options options = GetDefaultOptions();
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
options.env = fault_env_.get();
|
2020-09-18 03:22:35 +00:00
|
|
|
options.create_if_missing = true;
|
|
|
|
|
options.listeners.emplace_back(listener);
|
|
|
|
|
options.max_bgerror_resume_count = 0;
|
2021-03-18 05:36:42 +00:00
|
|
|
options.statistics = CreateDBStatistics();
|
2020-09-18 03:22:35 +00:00
|
|
|
Status s;
|
|
|
|
|
|
|
|
|
|
listener->EnableAutoRecovery(false);
|
|
|
|
|
DestroyAndReopen(options);
|
|
|
|
|
|
|
|
|
|
IOStatus error_msg = IOStatus::IOError("Retryable IO Error");
|
|
|
|
|
error_msg.SetRetryable(true);
|
|
|
|
|
|
|
|
|
|
WriteOptions wo = WriteOptions();
|
|
|
|
|
wo.disableWAL = true;
|
2020-10-02 23:39:17 +00:00
|
|
|
ASSERT_OK(Put(Key(1), "val1", wo));
|
2020-09-18 03:22:35 +00:00
|
|
|
SyncPoint::GetInstance()->SetCallBack(
|
|
|
|
|
"BuildTable:BeforeFinishBuildTable",
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
[&](void*) { fault_fs_->SetFilesystemActive(false, error_msg); });
|
2020-09-18 03:22:35 +00:00
|
|
|
SyncPoint::GetInstance()->EnableProcessing();
|
|
|
|
|
s = Flush();
|
2020-10-02 23:39:17 +00:00
|
|
|
ASSERT_OK(Put(Key(2), "val2", wo));
|
2020-09-18 03:22:35 +00:00
|
|
|
ASSERT_EQ(s.severity(), ROCKSDB_NAMESPACE::Status::Severity::kSoftError);
|
|
|
|
|
ASSERT_EQ("val2", Get(Key(2)));
|
|
|
|
|
SyncPoint::GetInstance()->DisableProcessing();
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
fault_fs_->SetFilesystemActive(true);
|
2020-09-18 03:22:35 +00:00
|
|
|
s = dbfull()->Resume();
|
2020-12-24 00:54:05 +00:00
|
|
|
ASSERT_OK(s);
|
2020-09-18 03:22:35 +00:00
|
|
|
ASSERT_EQ("val1", Get(Key(1)));
|
|
|
|
|
ASSERT_EQ("val2", Get(Key(2)));
|
2020-10-02 23:39:17 +00:00
|
|
|
ASSERT_OK(Put(Key(3), "val3", wo));
|
2020-09-18 03:22:35 +00:00
|
|
|
ASSERT_EQ("val3", Get(Key(3)));
|
|
|
|
|
s = Flush();
|
|
|
|
|
ASSERT_OK(s);
|
|
|
|
|
ASSERT_EQ("val3", Get(Key(3)));
|
2021-03-18 05:36:42 +00:00
|
|
|
ASSERT_EQ(1, options.statistics->getAndResetTickerCount(
|
|
|
|
|
ERROR_HANDLER_BG_ERROR_COUNT));
|
|
|
|
|
ASSERT_EQ(1, options.statistics->getAndResetTickerCount(
|
|
|
|
|
ERROR_HANDLER_BG_IO_ERROR_COUNT));
|
|
|
|
|
ASSERT_EQ(1, options.statistics->getAndResetTickerCount(
|
|
|
|
|
ERROR_HANDLER_BG_RETRYABLE_IO_ERROR_COUNT));
|
|
|
|
|
ASSERT_EQ(0, options.statistics->getAndResetTickerCount(
|
|
|
|
|
ERROR_HANDLER_AUTORESUME_COUNT));
|
|
|
|
|
ASSERT_EQ(0, options.statistics->getAndResetTickerCount(
|
|
|
|
|
ERROR_HANDLER_AUTORESUME_RETRY_TOTAL_COUNT));
|
|
|
|
|
ASSERT_EQ(0, options.statistics->getAndResetTickerCount(
|
|
|
|
|
ERROR_HANDLER_AUTORESUME_SUCCESS_COUNT));
|
2020-09-18 03:22:35 +00:00
|
|
|
|
2020-09-20 00:56:39 +00:00
|
|
|
Destroy(options);
|
|
|
|
|
}
|
|
|
|
|
|
2021-11-08 23:47:53 +00:00
|
|
|
// The retryable IO error is injected before we sync table
|
2023-10-31 23:13:36 +00:00
|
|
|
TEST_F(DBErrorHandlingFSTest, FlushWriteNoWALRetryableError2) {
|
|
|
|
|
std::shared_ptr<ErrorHandlerFSListener> listener =
|
|
|
|
|
std::make_shared<ErrorHandlerFSListener>();
|
2020-09-20 00:56:39 +00:00
|
|
|
Options options = GetDefaultOptions();
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
options.env = fault_env_.get();
|
2020-09-20 00:56:39 +00:00
|
|
|
options.create_if_missing = true;
|
|
|
|
|
options.listeners.emplace_back(listener);
|
|
|
|
|
options.max_bgerror_resume_count = 0;
|
|
|
|
|
Status s;
|
|
|
|
|
|
|
|
|
|
listener->EnableAutoRecovery(false);
|
|
|
|
|
DestroyAndReopen(options);
|
|
|
|
|
|
|
|
|
|
IOStatus error_msg = IOStatus::IOError("Retryable IO Error");
|
|
|
|
|
error_msg.SetRetryable(true);
|
|
|
|
|
|
|
|
|
|
WriteOptions wo = WriteOptions();
|
|
|
|
|
wo.disableWAL = true;
|
|
|
|
|
|
2020-10-02 23:39:17 +00:00
|
|
|
ASSERT_OK(Put(Key(1), "val1", wo));
|
2020-09-18 03:22:35 +00:00
|
|
|
SyncPoint::GetInstance()->SetCallBack(
|
2020-09-20 00:56:39 +00:00
|
|
|
"BuildTable:BeforeSyncTable",
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
[&](void*) { fault_fs_->SetFilesystemActive(false, error_msg); });
|
2020-09-18 03:22:35 +00:00
|
|
|
SyncPoint::GetInstance()->EnableProcessing();
|
|
|
|
|
s = Flush();
|
2020-10-02 23:39:17 +00:00
|
|
|
ASSERT_OK(Put(Key(2), "val2", wo));
|
2020-09-18 03:22:35 +00:00
|
|
|
ASSERT_EQ(s.severity(), ROCKSDB_NAMESPACE::Status::Severity::kSoftError);
|
2020-09-20 00:56:39 +00:00
|
|
|
ASSERT_EQ("val2", Get(Key(2)));
|
2020-09-18 03:22:35 +00:00
|
|
|
SyncPoint::GetInstance()->DisableProcessing();
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
fault_fs_->SetFilesystemActive(true);
|
2020-09-18 03:22:35 +00:00
|
|
|
s = dbfull()->Resume();
|
2020-12-24 00:54:05 +00:00
|
|
|
ASSERT_OK(s);
|
2020-09-20 00:56:39 +00:00
|
|
|
ASSERT_EQ("val1", Get(Key(1)));
|
|
|
|
|
ASSERT_EQ("val2", Get(Key(2)));
|
2020-10-02 23:39:17 +00:00
|
|
|
ASSERT_OK(Put(Key(3), "val3", wo));
|
2020-09-20 00:56:39 +00:00
|
|
|
ASSERT_EQ("val3", Get(Key(3)));
|
2020-09-18 03:22:35 +00:00
|
|
|
s = Flush();
|
|
|
|
|
ASSERT_OK(s);
|
2020-09-20 00:56:39 +00:00
|
|
|
ASSERT_EQ("val3", Get(Key(3)));
|
2020-09-18 03:22:35 +00:00
|
|
|
|
2020-09-20 00:56:39 +00:00
|
|
|
Destroy(options);
|
|
|
|
|
}
|
|
|
|
|
|
2021-11-08 23:47:53 +00:00
|
|
|
// The retryable IO error is injected before we close the table file
|
2023-10-31 23:13:36 +00:00
|
|
|
TEST_F(DBErrorHandlingFSTest, FlushWriteNoWALRetryableError3) {
|
|
|
|
|
std::shared_ptr<ErrorHandlerFSListener> listener =
|
|
|
|
|
std::make_shared<ErrorHandlerFSListener>();
|
2020-09-20 00:56:39 +00:00
|
|
|
Options options = GetDefaultOptions();
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
options.env = fault_env_.get();
|
2020-09-20 00:56:39 +00:00
|
|
|
options.create_if_missing = true;
|
|
|
|
|
options.listeners.emplace_back(listener);
|
|
|
|
|
options.max_bgerror_resume_count = 0;
|
|
|
|
|
Status s;
|
|
|
|
|
|
|
|
|
|
listener->EnableAutoRecovery(false);
|
|
|
|
|
DestroyAndReopen(options);
|
|
|
|
|
|
|
|
|
|
IOStatus error_msg = IOStatus::IOError("Retryable IO Error");
|
|
|
|
|
error_msg.SetRetryable(true);
|
|
|
|
|
|
|
|
|
|
WriteOptions wo = WriteOptions();
|
|
|
|
|
wo.disableWAL = true;
|
|
|
|
|
|
2020-10-02 23:39:17 +00:00
|
|
|
ASSERT_OK(Put(Key(1), "val1", wo));
|
2020-09-18 03:22:35 +00:00
|
|
|
SyncPoint::GetInstance()->SetCallBack(
|
2020-09-20 00:56:39 +00:00
|
|
|
"BuildTable:BeforeCloseTableFile",
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
[&](void*) { fault_fs_->SetFilesystemActive(false, error_msg); });
|
2020-09-18 03:22:35 +00:00
|
|
|
SyncPoint::GetInstance()->EnableProcessing();
|
|
|
|
|
s = Flush();
|
2020-10-02 23:39:17 +00:00
|
|
|
ASSERT_OK(Put(Key(2), "val2", wo));
|
2020-09-18 03:22:35 +00:00
|
|
|
ASSERT_EQ(s.severity(), ROCKSDB_NAMESPACE::Status::Severity::kSoftError);
|
2020-09-20 00:56:39 +00:00
|
|
|
ASSERT_EQ("val2", Get(Key(2)));
|
2020-09-18 03:22:35 +00:00
|
|
|
SyncPoint::GetInstance()->DisableProcessing();
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
fault_fs_->SetFilesystemActive(true);
|
2020-09-18 03:22:35 +00:00
|
|
|
s = dbfull()->Resume();
|
2020-12-24 00:54:05 +00:00
|
|
|
ASSERT_OK(s);
|
2020-09-20 00:56:39 +00:00
|
|
|
ASSERT_EQ("val1", Get(Key(1)));
|
|
|
|
|
ASSERT_EQ("val2", Get(Key(2)));
|
2020-10-02 23:39:17 +00:00
|
|
|
ASSERT_OK(Put(Key(3), "val3", wo));
|
2020-09-20 00:56:39 +00:00
|
|
|
ASSERT_EQ("val3", Get(Key(3)));
|
2020-09-18 03:22:35 +00:00
|
|
|
s = Flush();
|
|
|
|
|
ASSERT_OK(s);
|
2020-09-20 00:56:39 +00:00
|
|
|
ASSERT_EQ("val3", Get(Key(3)));
|
|
|
|
|
|
2020-09-18 03:22:35 +00:00
|
|
|
Destroy(options);
|
|
|
|
|
}
|
|
|
|
|
|
2020-03-04 20:30:34 +00:00
|
|
|
TEST_F(DBErrorHandlingFSTest, ManifestWriteError) {
|
2023-10-31 23:13:36 +00:00
|
|
|
std::shared_ptr<ErrorHandlerFSListener> listener =
|
|
|
|
|
std::make_shared<ErrorHandlerFSListener>();
|
2020-01-30 18:53:46 +00:00
|
|
|
Options options = GetDefaultOptions();
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
options.env = fault_env_.get();
|
2020-01-30 18:53:46 +00:00
|
|
|
options.create_if_missing = true;
|
|
|
|
|
options.listeners.emplace_back(listener);
|
|
|
|
|
Status s;
|
|
|
|
|
std::string old_manifest;
|
|
|
|
|
std::string new_manifest;
|
|
|
|
|
|
|
|
|
|
listener->EnableAutoRecovery(false);
|
|
|
|
|
DestroyAndReopen(options);
|
|
|
|
|
old_manifest = GetManifestNameFromLiveFiles();
|
|
|
|
|
|
2020-10-02 23:39:17 +00:00
|
|
|
ASSERT_OK(Put(Key(0), "val"));
|
|
|
|
|
ASSERT_OK(Flush());
|
|
|
|
|
ASSERT_OK(Put(Key(1), "val"));
|
2020-01-30 18:53:46 +00:00
|
|
|
SyncPoint::GetInstance()->SetCallBack(
|
2020-03-04 20:30:34 +00:00
|
|
|
"VersionSet::LogAndApply:WriteManifest", [&](void*) {
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
fault_fs_->SetFilesystemActive(false,
|
|
|
|
|
IOStatus::NoSpace("Out of space"));
|
2020-03-04 20:30:34 +00:00
|
|
|
});
|
2020-01-30 18:53:46 +00:00
|
|
|
SyncPoint::GetInstance()->EnableProcessing();
|
|
|
|
|
s = Flush();
|
2020-02-20 20:07:53 +00:00
|
|
|
ASSERT_EQ(s.severity(), ROCKSDB_NAMESPACE::Status::Severity::kHardError);
|
2023-11-11 16:11:11 +00:00
|
|
|
ASSERT_FALSE(dbfull()->TEST_GetFilesToQuarantine().empty());
|
2020-01-30 18:53:46 +00:00
|
|
|
SyncPoint::GetInstance()->ClearAllCallBacks();
|
|
|
|
|
SyncPoint::GetInstance()->DisableProcessing();
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
fault_fs_->SetFilesystemActive(true);
|
2020-01-30 18:53:46 +00:00
|
|
|
s = dbfull()->Resume();
|
2020-12-24 00:54:05 +00:00
|
|
|
ASSERT_OK(s);
|
2020-01-30 18:53:46 +00:00
|
|
|
|
|
|
|
|
new_manifest = GetManifestNameFromLiveFiles();
|
|
|
|
|
ASSERT_NE(new_manifest, old_manifest);
|
2023-11-11 16:11:11 +00:00
|
|
|
ASSERT_TRUE(dbfull()->TEST_GetFilesToQuarantine().empty());
|
2020-01-30 18:53:46 +00:00
|
|
|
|
|
|
|
|
Reopen(options);
|
|
|
|
|
ASSERT_EQ("val", Get(Key(0)));
|
|
|
|
|
ASSERT_EQ("val", Get(Key(1)));
|
|
|
|
|
Close();
|
|
|
|
|
}
|
|
|
|
|
|
Pass IOStatus to write path and set retryable IO Error as hard error in BG jobs (#6487)
Summary:
In the current code base, we use Status to get and store the returned status from the call. Specifically, for IO related functions, the current Status cannot reflect the IO Error details such as error scope, error retryable attribute, and others. With the implementation of https://github.com/facebook/rocksdb/issues/5761, we have the new Wrapper for IO, which returns IOStatus instead of Status. However, the IOStatus is purged at the lower level of write path and transferred to Status.
The first job of this PR is to pass the IOStatus to the write path (flush, WAL write, and Compaction). The second job is to identify the Retryable IO Error as HardError, and set the bg_error_ as HardError. In this case, the DB Instance becomes read only. User is informed of the Status and need to take actions to deal with it (e.g., call db->Resume()).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6487
Test Plan: Added the testing case to error_handler_fs_test. Pass make asan_check
Reviewed By: anand1976
Differential Revision: D20685017
Pulled By: zhichao-cao
fbshipit-source-id: ff85f042896243abcd6ef37877834e26f36b6eb0
2020-03-27 23:03:05 +00:00
|
|
|
TEST_F(DBErrorHandlingFSTest, ManifestWriteRetryableError) {
|
2023-10-31 23:13:36 +00:00
|
|
|
std::shared_ptr<ErrorHandlerFSListener> listener =
|
|
|
|
|
std::make_shared<ErrorHandlerFSListener>();
|
Pass IOStatus to write path and set retryable IO Error as hard error in BG jobs (#6487)
Summary:
In the current code base, we use Status to get and store the returned status from the call. Specifically, for IO related functions, the current Status cannot reflect the IO Error details such as error scope, error retryable attribute, and others. With the implementation of https://github.com/facebook/rocksdb/issues/5761, we have the new Wrapper for IO, which returns IOStatus instead of Status. However, the IOStatus is purged at the lower level of write path and transferred to Status.
The first job of this PR is to pass the IOStatus to the write path (flush, WAL write, and Compaction). The second job is to identify the Retryable IO Error as HardError, and set the bg_error_ as HardError. In this case, the DB Instance becomes read only. User is informed of the Status and need to take actions to deal with it (e.g., call db->Resume()).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6487
Test Plan: Added the testing case to error_handler_fs_test. Pass make asan_check
Reviewed By: anand1976
Differential Revision: D20685017
Pulled By: zhichao-cao
fbshipit-source-id: ff85f042896243abcd6ef37877834e26f36b6eb0
2020-03-27 23:03:05 +00:00
|
|
|
Options options = GetDefaultOptions();
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
options.env = fault_env_.get();
|
Pass IOStatus to write path and set retryable IO Error as hard error in BG jobs (#6487)
Summary:
In the current code base, we use Status to get and store the returned status from the call. Specifically, for IO related functions, the current Status cannot reflect the IO Error details such as error scope, error retryable attribute, and others. With the implementation of https://github.com/facebook/rocksdb/issues/5761, we have the new Wrapper for IO, which returns IOStatus instead of Status. However, the IOStatus is purged at the lower level of write path and transferred to Status.
The first job of this PR is to pass the IOStatus to the write path (flush, WAL write, and Compaction). The second job is to identify the Retryable IO Error as HardError, and set the bg_error_ as HardError. In this case, the DB Instance becomes read only. User is informed of the Status and need to take actions to deal with it (e.g., call db->Resume()).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6487
Test Plan: Added the testing case to error_handler_fs_test. Pass make asan_check
Reviewed By: anand1976
Differential Revision: D20685017
Pulled By: zhichao-cao
fbshipit-source-id: ff85f042896243abcd6ef37877834e26f36b6eb0
2020-03-27 23:03:05 +00:00
|
|
|
options.create_if_missing = true;
|
|
|
|
|
options.listeners.emplace_back(listener);
|
2020-07-15 18:02:44 +00:00
|
|
|
options.max_bgerror_resume_count = 0;
|
Pass IOStatus to write path and set retryable IO Error as hard error in BG jobs (#6487)
Summary:
In the current code base, we use Status to get and store the returned status from the call. Specifically, for IO related functions, the current Status cannot reflect the IO Error details such as error scope, error retryable attribute, and others. With the implementation of https://github.com/facebook/rocksdb/issues/5761, we have the new Wrapper for IO, which returns IOStatus instead of Status. However, the IOStatus is purged at the lower level of write path and transferred to Status.
The first job of this PR is to pass the IOStatus to the write path (flush, WAL write, and Compaction). The second job is to identify the Retryable IO Error as HardError, and set the bg_error_ as HardError. In this case, the DB Instance becomes read only. User is informed of the Status and need to take actions to deal with it (e.g., call db->Resume()).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6487
Test Plan: Added the testing case to error_handler_fs_test. Pass make asan_check
Reviewed By: anand1976
Differential Revision: D20685017
Pulled By: zhichao-cao
fbshipit-source-id: ff85f042896243abcd6ef37877834e26f36b6eb0
2020-03-27 23:03:05 +00:00
|
|
|
Status s;
|
|
|
|
|
std::string old_manifest;
|
|
|
|
|
std::string new_manifest;
|
|
|
|
|
|
|
|
|
|
listener->EnableAutoRecovery(false);
|
|
|
|
|
DestroyAndReopen(options);
|
|
|
|
|
old_manifest = GetManifestNameFromLiveFiles();
|
|
|
|
|
|
|
|
|
|
IOStatus error_msg = IOStatus::IOError("Retryable IO Error");
|
|
|
|
|
error_msg.SetRetryable(true);
|
|
|
|
|
|
2020-10-02 23:39:17 +00:00
|
|
|
ASSERT_OK(Put(Key(0), "val"));
|
|
|
|
|
ASSERT_OK(Flush());
|
|
|
|
|
ASSERT_OK(Put(Key(1), "val"));
|
Pass IOStatus to write path and set retryable IO Error as hard error in BG jobs (#6487)
Summary:
In the current code base, we use Status to get and store the returned status from the call. Specifically, for IO related functions, the current Status cannot reflect the IO Error details such as error scope, error retryable attribute, and others. With the implementation of https://github.com/facebook/rocksdb/issues/5761, we have the new Wrapper for IO, which returns IOStatus instead of Status. However, the IOStatus is purged at the lower level of write path and transferred to Status.
The first job of this PR is to pass the IOStatus to the write path (flush, WAL write, and Compaction). The second job is to identify the Retryable IO Error as HardError, and set the bg_error_ as HardError. In this case, the DB Instance becomes read only. User is informed of the Status and need to take actions to deal with it (e.g., call db->Resume()).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6487
Test Plan: Added the testing case to error_handler_fs_test. Pass make asan_check
Reviewed By: anand1976
Differential Revision: D20685017
Pulled By: zhichao-cao
fbshipit-source-id: ff85f042896243abcd6ef37877834e26f36b6eb0
2020-03-27 23:03:05 +00:00
|
|
|
SyncPoint::GetInstance()->SetCallBack(
|
|
|
|
|
"VersionSet::LogAndApply:WriteManifest",
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
[&](void*) { fault_fs_->SetFilesystemActive(false, error_msg); });
|
Pass IOStatus to write path and set retryable IO Error as hard error in BG jobs (#6487)
Summary:
In the current code base, we use Status to get and store the returned status from the call. Specifically, for IO related functions, the current Status cannot reflect the IO Error details such as error scope, error retryable attribute, and others. With the implementation of https://github.com/facebook/rocksdb/issues/5761, we have the new Wrapper for IO, which returns IOStatus instead of Status. However, the IOStatus is purged at the lower level of write path and transferred to Status.
The first job of this PR is to pass the IOStatus to the write path (flush, WAL write, and Compaction). The second job is to identify the Retryable IO Error as HardError, and set the bg_error_ as HardError. In this case, the DB Instance becomes read only. User is informed of the Status and need to take actions to deal with it (e.g., call db->Resume()).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6487
Test Plan: Added the testing case to error_handler_fs_test. Pass make asan_check
Reviewed By: anand1976
Differential Revision: D20685017
Pulled By: zhichao-cao
fbshipit-source-id: ff85f042896243abcd6ef37877834e26f36b6eb0
2020-03-27 23:03:05 +00:00
|
|
|
SyncPoint::GetInstance()->EnableProcessing();
|
|
|
|
|
s = Flush();
|
2021-03-26 04:41:44 +00:00
|
|
|
ASSERT_EQ(s.severity(), ROCKSDB_NAMESPACE::Status::Severity::kSoftError);
|
2023-11-11 16:11:11 +00:00
|
|
|
ASSERT_FALSE(dbfull()->TEST_GetFilesToQuarantine().empty());
|
Pass IOStatus to write path and set retryable IO Error as hard error in BG jobs (#6487)
Summary:
In the current code base, we use Status to get and store the returned status from the call. Specifically, for IO related functions, the current Status cannot reflect the IO Error details such as error scope, error retryable attribute, and others. With the implementation of https://github.com/facebook/rocksdb/issues/5761, we have the new Wrapper for IO, which returns IOStatus instead of Status. However, the IOStatus is purged at the lower level of write path and transferred to Status.
The first job of this PR is to pass the IOStatus to the write path (flush, WAL write, and Compaction). The second job is to identify the Retryable IO Error as HardError, and set the bg_error_ as HardError. In this case, the DB Instance becomes read only. User is informed of the Status and need to take actions to deal with it (e.g., call db->Resume()).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6487
Test Plan: Added the testing case to error_handler_fs_test. Pass make asan_check
Reviewed By: anand1976
Differential Revision: D20685017
Pulled By: zhichao-cao
fbshipit-source-id: ff85f042896243abcd6ef37877834e26f36b6eb0
2020-03-27 23:03:05 +00:00
|
|
|
SyncPoint::GetInstance()->ClearAllCallBacks();
|
|
|
|
|
SyncPoint::GetInstance()->DisableProcessing();
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
fault_fs_->SetFilesystemActive(true);
|
Pass IOStatus to write path and set retryable IO Error as hard error in BG jobs (#6487)
Summary:
In the current code base, we use Status to get and store the returned status from the call. Specifically, for IO related functions, the current Status cannot reflect the IO Error details such as error scope, error retryable attribute, and others. With the implementation of https://github.com/facebook/rocksdb/issues/5761, we have the new Wrapper for IO, which returns IOStatus instead of Status. However, the IOStatus is purged at the lower level of write path and transferred to Status.
The first job of this PR is to pass the IOStatus to the write path (flush, WAL write, and Compaction). The second job is to identify the Retryable IO Error as HardError, and set the bg_error_ as HardError. In this case, the DB Instance becomes read only. User is informed of the Status and need to take actions to deal with it (e.g., call db->Resume()).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6487
Test Plan: Added the testing case to error_handler_fs_test. Pass make asan_check
Reviewed By: anand1976
Differential Revision: D20685017
Pulled By: zhichao-cao
fbshipit-source-id: ff85f042896243abcd6ef37877834e26f36b6eb0
2020-03-27 23:03:05 +00:00
|
|
|
s = dbfull()->Resume();
|
2020-12-24 00:54:05 +00:00
|
|
|
ASSERT_OK(s);
|
Pass IOStatus to write path and set retryable IO Error as hard error in BG jobs (#6487)
Summary:
In the current code base, we use Status to get and store the returned status from the call. Specifically, for IO related functions, the current Status cannot reflect the IO Error details such as error scope, error retryable attribute, and others. With the implementation of https://github.com/facebook/rocksdb/issues/5761, we have the new Wrapper for IO, which returns IOStatus instead of Status. However, the IOStatus is purged at the lower level of write path and transferred to Status.
The first job of this PR is to pass the IOStatus to the write path (flush, WAL write, and Compaction). The second job is to identify the Retryable IO Error as HardError, and set the bg_error_ as HardError. In this case, the DB Instance becomes read only. User is informed of the Status and need to take actions to deal with it (e.g., call db->Resume()).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6487
Test Plan: Added the testing case to error_handler_fs_test. Pass make asan_check
Reviewed By: anand1976
Differential Revision: D20685017
Pulled By: zhichao-cao
fbshipit-source-id: ff85f042896243abcd6ef37877834e26f36b6eb0
2020-03-27 23:03:05 +00:00
|
|
|
|
|
|
|
|
new_manifest = GetManifestNameFromLiveFiles();
|
|
|
|
|
ASSERT_NE(new_manifest, old_manifest);
|
2023-11-11 16:11:11 +00:00
|
|
|
ASSERT_TRUE(dbfull()->TEST_GetFilesToQuarantine().empty());
|
Pass IOStatus to write path and set retryable IO Error as hard error in BG jobs (#6487)
Summary:
In the current code base, we use Status to get and store the returned status from the call. Specifically, for IO related functions, the current Status cannot reflect the IO Error details such as error scope, error retryable attribute, and others. With the implementation of https://github.com/facebook/rocksdb/issues/5761, we have the new Wrapper for IO, which returns IOStatus instead of Status. However, the IOStatus is purged at the lower level of write path and transferred to Status.
The first job of this PR is to pass the IOStatus to the write path (flush, WAL write, and Compaction). The second job is to identify the Retryable IO Error as HardError, and set the bg_error_ as HardError. In this case, the DB Instance becomes read only. User is informed of the Status and need to take actions to deal with it (e.g., call db->Resume()).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6487
Test Plan: Added the testing case to error_handler_fs_test. Pass make asan_check
Reviewed By: anand1976
Differential Revision: D20685017
Pulled By: zhichao-cao
fbshipit-source-id: ff85f042896243abcd6ef37877834e26f36b6eb0
2020-03-27 23:03:05 +00:00
|
|
|
|
|
|
|
|
Reopen(options);
|
|
|
|
|
ASSERT_EQ("val", Get(Key(0)));
|
|
|
|
|
ASSERT_EQ("val", Get(Key(1)));
|
|
|
|
|
Close();
|
|
|
|
|
}
|
|
|
|
|
|
2021-01-08 00:30:06 +00:00
|
|
|
TEST_F(DBErrorHandlingFSTest, ManifestWriteFileScopeError) {
|
2023-10-31 23:13:36 +00:00
|
|
|
std::shared_ptr<ErrorHandlerFSListener> listener =
|
|
|
|
|
std::make_shared<ErrorHandlerFSListener>();
|
2021-01-08 00:30:06 +00:00
|
|
|
Options options = GetDefaultOptions();
|
|
|
|
|
options.env = fault_env_.get();
|
|
|
|
|
options.create_if_missing = true;
|
|
|
|
|
options.listeners.emplace_back(listener);
|
|
|
|
|
options.max_bgerror_resume_count = 0;
|
|
|
|
|
Status s;
|
|
|
|
|
std::string old_manifest;
|
|
|
|
|
std::string new_manifest;
|
|
|
|
|
|
|
|
|
|
listener->EnableAutoRecovery(false);
|
|
|
|
|
DestroyAndReopen(options);
|
|
|
|
|
old_manifest = GetManifestNameFromLiveFiles();
|
|
|
|
|
|
|
|
|
|
IOStatus error_msg = IOStatus::IOError("File Scope Data Loss Error");
|
|
|
|
|
error_msg.SetDataLoss(true);
|
|
|
|
|
error_msg.SetScope(
|
|
|
|
|
ROCKSDB_NAMESPACE::IOStatus::IOErrorScope::kIOErrorScopeFile);
|
|
|
|
|
error_msg.SetRetryable(false);
|
|
|
|
|
|
|
|
|
|
ASSERT_OK(Put(Key(0), "val"));
|
|
|
|
|
ASSERT_OK(Flush());
|
|
|
|
|
ASSERT_OK(Put(Key(1), "val"));
|
|
|
|
|
SyncPoint::GetInstance()->SetCallBack(
|
|
|
|
|
"VersionSet::LogAndApply:WriteManifest",
|
|
|
|
|
[&](void*) { fault_fs_->SetFilesystemActive(false, error_msg); });
|
|
|
|
|
SyncPoint::GetInstance()->EnableProcessing();
|
|
|
|
|
s = Flush();
|
2023-11-11 16:11:11 +00:00
|
|
|
ASSERT_FALSE(dbfull()->TEST_GetFilesToQuarantine().empty());
|
2021-03-26 04:41:44 +00:00
|
|
|
ASSERT_EQ(s.severity(), ROCKSDB_NAMESPACE::Status::Severity::kSoftError);
|
2021-01-08 00:30:06 +00:00
|
|
|
SyncPoint::GetInstance()->ClearAllCallBacks();
|
|
|
|
|
SyncPoint::GetInstance()->DisableProcessing();
|
|
|
|
|
fault_fs_->SetFilesystemActive(true);
|
|
|
|
|
s = dbfull()->Resume();
|
|
|
|
|
ASSERT_OK(s);
|
|
|
|
|
|
|
|
|
|
new_manifest = GetManifestNameFromLiveFiles();
|
|
|
|
|
ASSERT_NE(new_manifest, old_manifest);
|
2023-11-11 16:11:11 +00:00
|
|
|
ASSERT_TRUE(dbfull()->TEST_GetFilesToQuarantine().empty());
|
2021-01-08 00:30:06 +00:00
|
|
|
|
|
|
|
|
Reopen(options);
|
|
|
|
|
ASSERT_EQ("val", Get(Key(0)));
|
|
|
|
|
ASSERT_EQ("val", Get(Key(1)));
|
|
|
|
|
Close();
|
|
|
|
|
}
|
|
|
|
|
|
2020-12-03 02:22:05 +00:00
|
|
|
TEST_F(DBErrorHandlingFSTest, ManifestWriteNoWALRetryableError) {
|
2023-10-31 23:13:36 +00:00
|
|
|
std::shared_ptr<ErrorHandlerFSListener> listener =
|
|
|
|
|
std::make_shared<ErrorHandlerFSListener>();
|
2020-12-03 02:22:05 +00:00
|
|
|
Options options = GetDefaultOptions();
|
|
|
|
|
options.env = fault_env_.get();
|
|
|
|
|
options.create_if_missing = true;
|
|
|
|
|
options.listeners.emplace_back(listener);
|
|
|
|
|
options.max_bgerror_resume_count = 0;
|
|
|
|
|
Status s;
|
|
|
|
|
std::string old_manifest;
|
|
|
|
|
std::string new_manifest;
|
|
|
|
|
|
|
|
|
|
listener->EnableAutoRecovery(false);
|
|
|
|
|
DestroyAndReopen(options);
|
|
|
|
|
old_manifest = GetManifestNameFromLiveFiles();
|
|
|
|
|
|
|
|
|
|
IOStatus error_msg = IOStatus::IOError("Retryable IO Error");
|
|
|
|
|
error_msg.SetRetryable(true);
|
|
|
|
|
|
|
|
|
|
WriteOptions wo = WriteOptions();
|
|
|
|
|
wo.disableWAL = true;
|
|
|
|
|
ASSERT_OK(Put(Key(0), "val", wo));
|
|
|
|
|
ASSERT_OK(Flush());
|
|
|
|
|
ASSERT_OK(Put(Key(1), "val", wo));
|
|
|
|
|
SyncPoint::GetInstance()->SetCallBack(
|
|
|
|
|
"VersionSet::LogAndApply:WriteManifest",
|
|
|
|
|
[&](void*) { fault_fs_->SetFilesystemActive(false, error_msg); });
|
|
|
|
|
SyncPoint::GetInstance()->EnableProcessing();
|
|
|
|
|
s = Flush();
|
|
|
|
|
ASSERT_EQ(s.severity(), ROCKSDB_NAMESPACE::Status::Severity::kSoftError);
|
2023-11-11 16:11:11 +00:00
|
|
|
ASSERT_FALSE(dbfull()->TEST_GetFilesToQuarantine().empty());
|
2020-12-03 02:22:05 +00:00
|
|
|
SyncPoint::GetInstance()->ClearAllCallBacks();
|
|
|
|
|
SyncPoint::GetInstance()->DisableProcessing();
|
|
|
|
|
fault_fs_->SetFilesystemActive(true);
|
|
|
|
|
s = dbfull()->Resume();
|
2020-12-24 00:54:05 +00:00
|
|
|
ASSERT_OK(s);
|
2020-12-03 02:22:05 +00:00
|
|
|
|
|
|
|
|
new_manifest = GetManifestNameFromLiveFiles();
|
|
|
|
|
ASSERT_NE(new_manifest, old_manifest);
|
2023-11-11 16:11:11 +00:00
|
|
|
ASSERT_TRUE(dbfull()->TEST_GetFilesToQuarantine().empty());
|
2020-12-03 02:22:05 +00:00
|
|
|
|
|
|
|
|
Reopen(options);
|
|
|
|
|
ASSERT_EQ("val", Get(Key(0)));
|
|
|
|
|
ASSERT_EQ("val", Get(Key(1)));
|
|
|
|
|
Close();
|
|
|
|
|
}
|
|
|
|
|
|
2020-03-04 20:30:34 +00:00
|
|
|
TEST_F(DBErrorHandlingFSTest, DoubleManifestWriteError) {
|
2023-10-31 23:13:36 +00:00
|
|
|
std::shared_ptr<ErrorHandlerFSListener> listener =
|
|
|
|
|
std::make_shared<ErrorHandlerFSListener>();
|
2020-01-30 18:53:46 +00:00
|
|
|
Options options = GetDefaultOptions();
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
options.env = fault_env_.get();
|
2020-01-30 18:53:46 +00:00
|
|
|
options.create_if_missing = true;
|
|
|
|
|
options.listeners.emplace_back(listener);
|
|
|
|
|
Status s;
|
|
|
|
|
std::string old_manifest;
|
|
|
|
|
std::string new_manifest;
|
|
|
|
|
|
|
|
|
|
listener->EnableAutoRecovery(false);
|
|
|
|
|
DestroyAndReopen(options);
|
|
|
|
|
old_manifest = GetManifestNameFromLiveFiles();
|
|
|
|
|
|
2020-10-02 23:39:17 +00:00
|
|
|
ASSERT_OK(Put(Key(0), "val"));
|
|
|
|
|
ASSERT_OK(Flush());
|
|
|
|
|
ASSERT_OK(Put(Key(1), "val"));
|
2020-01-30 18:53:46 +00:00
|
|
|
SyncPoint::GetInstance()->SetCallBack(
|
2020-03-04 20:30:34 +00:00
|
|
|
"VersionSet::LogAndApply:WriteManifest", [&](void*) {
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
fault_fs_->SetFilesystemActive(false,
|
|
|
|
|
IOStatus::NoSpace("Out of space"));
|
2020-03-04 20:30:34 +00:00
|
|
|
});
|
2020-01-30 18:53:46 +00:00
|
|
|
SyncPoint::GetInstance()->EnableProcessing();
|
|
|
|
|
s = Flush();
|
2024-06-27 01:17:05 +00:00
|
|
|
ASSERT_TRUE(s.IsNoSpace());
|
|
|
|
|
ASSERT_EQ(dbfull()->TEST_GetBGError().severity(),
|
|
|
|
|
ROCKSDB_NAMESPACE::Status::Severity::kHardError);
|
2023-11-11 16:11:11 +00:00
|
|
|
ASSERT_FALSE(dbfull()->TEST_GetFilesToQuarantine().empty());
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
fault_fs_->SetFilesystemActive(true);
|
2020-01-30 18:53:46 +00:00
|
|
|
|
|
|
|
|
// This Resume() will attempt to create a new manifest file and fail again
|
|
|
|
|
s = dbfull()->Resume();
|
2024-06-27 01:17:05 +00:00
|
|
|
ASSERT_TRUE(s.IsNoSpace());
|
|
|
|
|
ASSERT_EQ(dbfull()->TEST_GetBGError().severity(),
|
|
|
|
|
ROCKSDB_NAMESPACE::Status::Severity::kHardError);
|
2023-11-11 16:11:11 +00:00
|
|
|
ASSERT_FALSE(dbfull()->TEST_GetFilesToQuarantine().empty());
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
fault_fs_->SetFilesystemActive(true);
|
2020-01-30 18:53:46 +00:00
|
|
|
SyncPoint::GetInstance()->ClearAllCallBacks();
|
|
|
|
|
SyncPoint::GetInstance()->DisableProcessing();
|
|
|
|
|
|
|
|
|
|
// A successful Resume() will create a new manifest file
|
|
|
|
|
s = dbfull()->Resume();
|
2020-12-24 00:54:05 +00:00
|
|
|
ASSERT_OK(s);
|
2020-01-30 18:53:46 +00:00
|
|
|
|
|
|
|
|
new_manifest = GetManifestNameFromLiveFiles();
|
|
|
|
|
ASSERT_NE(new_manifest, old_manifest);
|
2023-11-11 16:11:11 +00:00
|
|
|
ASSERT_TRUE(dbfull()->TEST_GetFilesToQuarantine().empty());
|
2020-01-30 18:53:46 +00:00
|
|
|
|
|
|
|
|
Reopen(options);
|
|
|
|
|
ASSERT_EQ("val", Get(Key(0)));
|
|
|
|
|
ASSERT_EQ("val", Get(Key(1)));
|
|
|
|
|
Close();
|
|
|
|
|
}
|
|
|
|
|
|
2020-03-04 20:30:34 +00:00
|
|
|
TEST_F(DBErrorHandlingFSTest, CompactionManifestWriteError) {
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
if (mem_env_ != nullptr) {
|
|
|
|
|
ROCKSDB_GTEST_SKIP("Test requires non-mock environment");
|
|
|
|
|
return;
|
|
|
|
|
}
|
2023-10-31 23:13:36 +00:00
|
|
|
std::shared_ptr<ErrorHandlerFSListener> listener =
|
|
|
|
|
std::make_shared<ErrorHandlerFSListener>();
|
2020-01-30 18:53:46 +00:00
|
|
|
Options options = GetDefaultOptions();
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
options.env = fault_env_.get();
|
2020-01-30 18:53:46 +00:00
|
|
|
options.create_if_missing = true;
|
|
|
|
|
options.level0_file_num_compaction_trigger = 2;
|
|
|
|
|
options.listeners.emplace_back(listener);
|
|
|
|
|
Status s;
|
|
|
|
|
std::string old_manifest;
|
|
|
|
|
std::string new_manifest;
|
|
|
|
|
std::atomic<bool> fail_manifest(false);
|
|
|
|
|
DestroyAndReopen(options);
|
|
|
|
|
old_manifest = GetManifestNameFromLiveFiles();
|
|
|
|
|
|
2020-10-02 23:39:17 +00:00
|
|
|
ASSERT_OK(Put(Key(0), "val"));
|
|
|
|
|
ASSERT_OK(Put(Key(2), "val"));
|
2020-01-30 18:53:46 +00:00
|
|
|
s = Flush();
|
2020-12-24 00:54:05 +00:00
|
|
|
ASSERT_OK(s);
|
2020-01-30 18:53:46 +00:00
|
|
|
|
2020-02-20 20:07:53 +00:00
|
|
|
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->LoadDependency(
|
2020-01-30 18:53:46 +00:00
|
|
|
// Wait for flush of 2nd L0 file before starting compaction
|
|
|
|
|
{{"DBImpl::FlushMemTable:FlushMemTableFinished",
|
|
|
|
|
"BackgroundCallCompaction:0"},
|
2020-02-20 20:07:53 +00:00
|
|
|
// Wait for compaction to detect manifest write error
|
|
|
|
|
{"BackgroundCallCompaction:1", "CompactionManifestWriteError:0"},
|
|
|
|
|
// Make compaction thread wait for error to be cleared
|
2020-01-30 18:53:46 +00:00
|
|
|
{"CompactionManifestWriteError:1",
|
|
|
|
|
"DBImpl::BackgroundCallCompaction:FoundObsoleteFiles"},
|
2020-02-20 20:07:53 +00:00
|
|
|
// Wait for DB instance to clear bg_error before calling
|
|
|
|
|
// TEST_WaitForCompact
|
|
|
|
|
{"SstFileManagerImpl::ErrorCleared", "CompactionManifestWriteError:2"}});
|
2020-01-30 18:53:46 +00:00
|
|
|
// trigger manifest write failure in compaction thread
|
2020-02-20 20:07:53 +00:00
|
|
|
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->SetCallBack(
|
|
|
|
|
"BackgroundCallCompaction:0", [&](void*) { fail_manifest.store(true); });
|
|
|
|
|
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->SetCallBack(
|
|
|
|
|
"VersionSet::LogAndApply:WriteManifest", [&](void*) {
|
|
|
|
|
if (fail_manifest.load()) {
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
fault_fs_->SetFilesystemActive(false,
|
|
|
|
|
IOStatus::NoSpace("Out of space"));
|
2020-02-20 20:07:53 +00:00
|
|
|
}
|
2020-01-30 18:53:46 +00:00
|
|
|
});
|
2020-02-20 20:07:53 +00:00
|
|
|
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->EnableProcessing();
|
2020-01-30 18:53:46 +00:00
|
|
|
|
2020-10-02 23:39:17 +00:00
|
|
|
ASSERT_OK(Put(Key(1), "val"));
|
2020-01-30 18:53:46 +00:00
|
|
|
// This Flush will trigger a compaction, which will fail when appending to
|
|
|
|
|
// the manifest
|
|
|
|
|
s = Flush();
|
2020-12-24 00:54:05 +00:00
|
|
|
ASSERT_OK(s);
|
2020-01-30 18:53:46 +00:00
|
|
|
|
|
|
|
|
TEST_SYNC_POINT("CompactionManifestWriteError:0");
|
|
|
|
|
// Clear all errors so when the compaction is retried, it will succeed
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
fault_fs_->SetFilesystemActive(true);
|
2020-02-20 20:07:53 +00:00
|
|
|
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->ClearAllCallBacks();
|
2020-01-30 18:53:46 +00:00
|
|
|
TEST_SYNC_POINT("CompactionManifestWriteError:1");
|
2023-11-11 16:11:11 +00:00
|
|
|
ASSERT_FALSE(dbfull()->TEST_GetFilesToQuarantine().empty());
|
2020-01-30 18:53:46 +00:00
|
|
|
TEST_SYNC_POINT("CompactionManifestWriteError:2");
|
|
|
|
|
|
|
|
|
|
s = dbfull()->TEST_WaitForCompact();
|
2020-02-20 20:07:53 +00:00
|
|
|
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->DisableProcessing();
|
2020-12-24 00:54:05 +00:00
|
|
|
ASSERT_OK(s);
|
2020-01-30 18:53:46 +00:00
|
|
|
|
|
|
|
|
new_manifest = GetManifestNameFromLiveFiles();
|
|
|
|
|
ASSERT_NE(new_manifest, old_manifest);
|
2023-11-11 16:11:11 +00:00
|
|
|
ASSERT_TRUE(dbfull()->TEST_GetFilesToQuarantine().empty());
|
2020-01-30 18:53:46 +00:00
|
|
|
Reopen(options);
|
|
|
|
|
ASSERT_EQ("val", Get(Key(0)));
|
|
|
|
|
ASSERT_EQ("val", Get(Key(1)));
|
|
|
|
|
ASSERT_EQ("val", Get(Key(2)));
|
|
|
|
|
Close();
|
|
|
|
|
}
|
|
|
|
|
|
Pass IOStatus to write path and set retryable IO Error as hard error in BG jobs (#6487)
Summary:
In the current code base, we use Status to get and store the returned status from the call. Specifically, for IO related functions, the current Status cannot reflect the IO Error details such as error scope, error retryable attribute, and others. With the implementation of https://github.com/facebook/rocksdb/issues/5761, we have the new Wrapper for IO, which returns IOStatus instead of Status. However, the IOStatus is purged at the lower level of write path and transferred to Status.
The first job of this PR is to pass the IOStatus to the write path (flush, WAL write, and Compaction). The second job is to identify the Retryable IO Error as HardError, and set the bg_error_ as HardError. In this case, the DB Instance becomes read only. User is informed of the Status and need to take actions to deal with it (e.g., call db->Resume()).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6487
Test Plan: Added the testing case to error_handler_fs_test. Pass make asan_check
Reviewed By: anand1976
Differential Revision: D20685017
Pulled By: zhichao-cao
fbshipit-source-id: ff85f042896243abcd6ef37877834e26f36b6eb0
2020-03-27 23:03:05 +00:00
|
|
|
TEST_F(DBErrorHandlingFSTest, CompactionManifestWriteRetryableError) {
|
2023-10-31 23:13:36 +00:00
|
|
|
std::shared_ptr<ErrorHandlerFSListener> listener =
|
|
|
|
|
std::make_shared<ErrorHandlerFSListener>();
|
Pass IOStatus to write path and set retryable IO Error as hard error in BG jobs (#6487)
Summary:
In the current code base, we use Status to get and store the returned status from the call. Specifically, for IO related functions, the current Status cannot reflect the IO Error details such as error scope, error retryable attribute, and others. With the implementation of https://github.com/facebook/rocksdb/issues/5761, we have the new Wrapper for IO, which returns IOStatus instead of Status. However, the IOStatus is purged at the lower level of write path and transferred to Status.
The first job of this PR is to pass the IOStatus to the write path (flush, WAL write, and Compaction). The second job is to identify the Retryable IO Error as HardError, and set the bg_error_ as HardError. In this case, the DB Instance becomes read only. User is informed of the Status and need to take actions to deal with it (e.g., call db->Resume()).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6487
Test Plan: Added the testing case to error_handler_fs_test. Pass make asan_check
Reviewed By: anand1976
Differential Revision: D20685017
Pulled By: zhichao-cao
fbshipit-source-id: ff85f042896243abcd6ef37877834e26f36b6eb0
2020-03-27 23:03:05 +00:00
|
|
|
Options options = GetDefaultOptions();
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
options.env = fault_env_.get();
|
Pass IOStatus to write path and set retryable IO Error as hard error in BG jobs (#6487)
Summary:
In the current code base, we use Status to get and store the returned status from the call. Specifically, for IO related functions, the current Status cannot reflect the IO Error details such as error scope, error retryable attribute, and others. With the implementation of https://github.com/facebook/rocksdb/issues/5761, we have the new Wrapper for IO, which returns IOStatus instead of Status. However, the IOStatus is purged at the lower level of write path and transferred to Status.
The first job of this PR is to pass the IOStatus to the write path (flush, WAL write, and Compaction). The second job is to identify the Retryable IO Error as HardError, and set the bg_error_ as HardError. In this case, the DB Instance becomes read only. User is informed of the Status and need to take actions to deal with it (e.g., call db->Resume()).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6487
Test Plan: Added the testing case to error_handler_fs_test. Pass make asan_check
Reviewed By: anand1976
Differential Revision: D20685017
Pulled By: zhichao-cao
fbshipit-source-id: ff85f042896243abcd6ef37877834e26f36b6eb0
2020-03-27 23:03:05 +00:00
|
|
|
options.create_if_missing = true;
|
|
|
|
|
options.level0_file_num_compaction_trigger = 2;
|
|
|
|
|
options.listeners.emplace_back(listener);
|
2020-07-15 18:02:44 +00:00
|
|
|
options.max_bgerror_resume_count = 0;
|
Pass IOStatus to write path and set retryable IO Error as hard error in BG jobs (#6487)
Summary:
In the current code base, we use Status to get and store the returned status from the call. Specifically, for IO related functions, the current Status cannot reflect the IO Error details such as error scope, error retryable attribute, and others. With the implementation of https://github.com/facebook/rocksdb/issues/5761, we have the new Wrapper for IO, which returns IOStatus instead of Status. However, the IOStatus is purged at the lower level of write path and transferred to Status.
The first job of this PR is to pass the IOStatus to the write path (flush, WAL write, and Compaction). The second job is to identify the Retryable IO Error as HardError, and set the bg_error_ as HardError. In this case, the DB Instance becomes read only. User is informed of the Status and need to take actions to deal with it (e.g., call db->Resume()).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6487
Test Plan: Added the testing case to error_handler_fs_test. Pass make asan_check
Reviewed By: anand1976
Differential Revision: D20685017
Pulled By: zhichao-cao
fbshipit-source-id: ff85f042896243abcd6ef37877834e26f36b6eb0
2020-03-27 23:03:05 +00:00
|
|
|
Status s;
|
|
|
|
|
std::string old_manifest;
|
|
|
|
|
std::string new_manifest;
|
2020-04-03 01:06:26 +00:00
|
|
|
std::atomic<bool> fail_manifest(false);
|
Pass IOStatus to write path and set retryable IO Error as hard error in BG jobs (#6487)
Summary:
In the current code base, we use Status to get and store the returned status from the call. Specifically, for IO related functions, the current Status cannot reflect the IO Error details such as error scope, error retryable attribute, and others. With the implementation of https://github.com/facebook/rocksdb/issues/5761, we have the new Wrapper for IO, which returns IOStatus instead of Status. However, the IOStatus is purged at the lower level of write path and transferred to Status.
The first job of this PR is to pass the IOStatus to the write path (flush, WAL write, and Compaction). The second job is to identify the Retryable IO Error as HardError, and set the bg_error_ as HardError. In this case, the DB Instance becomes read only. User is informed of the Status and need to take actions to deal with it (e.g., call db->Resume()).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6487
Test Plan: Added the testing case to error_handler_fs_test. Pass make asan_check
Reviewed By: anand1976
Differential Revision: D20685017
Pulled By: zhichao-cao
fbshipit-source-id: ff85f042896243abcd6ef37877834e26f36b6eb0
2020-03-27 23:03:05 +00:00
|
|
|
DestroyAndReopen(options);
|
|
|
|
|
old_manifest = GetManifestNameFromLiveFiles();
|
|
|
|
|
|
|
|
|
|
IOStatus error_msg = IOStatus::IOError("Retryable IO Error");
|
|
|
|
|
error_msg.SetRetryable(true);
|
|
|
|
|
|
2020-10-02 23:39:17 +00:00
|
|
|
ASSERT_OK(Put(Key(0), "val"));
|
|
|
|
|
ASSERT_OK(Put(Key(2), "val"));
|
Pass IOStatus to write path and set retryable IO Error as hard error in BG jobs (#6487)
Summary:
In the current code base, we use Status to get and store the returned status from the call. Specifically, for IO related functions, the current Status cannot reflect the IO Error details such as error scope, error retryable attribute, and others. With the implementation of https://github.com/facebook/rocksdb/issues/5761, we have the new Wrapper for IO, which returns IOStatus instead of Status. However, the IOStatus is purged at the lower level of write path and transferred to Status.
The first job of this PR is to pass the IOStatus to the write path (flush, WAL write, and Compaction). The second job is to identify the Retryable IO Error as HardError, and set the bg_error_ as HardError. In this case, the DB Instance becomes read only. User is informed of the Status and need to take actions to deal with it (e.g., call db->Resume()).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6487
Test Plan: Added the testing case to error_handler_fs_test. Pass make asan_check
Reviewed By: anand1976
Differential Revision: D20685017
Pulled By: zhichao-cao
fbshipit-source-id: ff85f042896243abcd6ef37877834e26f36b6eb0
2020-03-27 23:03:05 +00:00
|
|
|
s = Flush();
|
2020-12-24 00:54:05 +00:00
|
|
|
ASSERT_OK(s);
|
Pass IOStatus to write path and set retryable IO Error as hard error in BG jobs (#6487)
Summary:
In the current code base, we use Status to get and store the returned status from the call. Specifically, for IO related functions, the current Status cannot reflect the IO Error details such as error scope, error retryable attribute, and others. With the implementation of https://github.com/facebook/rocksdb/issues/5761, we have the new Wrapper for IO, which returns IOStatus instead of Status. However, the IOStatus is purged at the lower level of write path and transferred to Status.
The first job of this PR is to pass the IOStatus to the write path (flush, WAL write, and Compaction). The second job is to identify the Retryable IO Error as HardError, and set the bg_error_ as HardError. In this case, the DB Instance becomes read only. User is informed of the Status and need to take actions to deal with it (e.g., call db->Resume()).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6487
Test Plan: Added the testing case to error_handler_fs_test. Pass make asan_check
Reviewed By: anand1976
Differential Revision: D20685017
Pulled By: zhichao-cao
fbshipit-source-id: ff85f042896243abcd6ef37877834e26f36b6eb0
2020-03-27 23:03:05 +00:00
|
|
|
|
|
|
|
|
listener->OverrideBGError(Status(error_msg, Status::Severity::kHardError));
|
|
|
|
|
listener->EnableAutoRecovery(false);
|
2020-04-03 01:06:26 +00:00
|
|
|
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->LoadDependency(
|
|
|
|
|
// Wait for flush of 2nd L0 file before starting compaction
|
|
|
|
|
{{"DBImpl::FlushMemTable:FlushMemTableFinished",
|
|
|
|
|
"BackgroundCallCompaction:0"},
|
|
|
|
|
// Wait for compaction to detect manifest write error
|
|
|
|
|
{"BackgroundCallCompaction:1", "CompactionManifestWriteError:0"},
|
|
|
|
|
// Make compaction thread wait for error to be cleared
|
|
|
|
|
{"CompactionManifestWriteError:1",
|
|
|
|
|
"DBImpl::BackgroundCallCompaction:FoundObsoleteFiles"}});
|
|
|
|
|
// trigger manifest write failure in compaction thread
|
Pass IOStatus to write path and set retryable IO Error as hard error in BG jobs (#6487)
Summary:
In the current code base, we use Status to get and store the returned status from the call. Specifically, for IO related functions, the current Status cannot reflect the IO Error details such as error scope, error retryable attribute, and others. With the implementation of https://github.com/facebook/rocksdb/issues/5761, we have the new Wrapper for IO, which returns IOStatus instead of Status. However, the IOStatus is purged at the lower level of write path and transferred to Status.
The first job of this PR is to pass the IOStatus to the write path (flush, WAL write, and Compaction). The second job is to identify the Retryable IO Error as HardError, and set the bg_error_ as HardError. In this case, the DB Instance becomes read only. User is informed of the Status and need to take actions to deal with it (e.g., call db->Resume()).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6487
Test Plan: Added the testing case to error_handler_fs_test. Pass make asan_check
Reviewed By: anand1976
Differential Revision: D20685017
Pulled By: zhichao-cao
fbshipit-source-id: ff85f042896243abcd6ef37877834e26f36b6eb0
2020-03-27 23:03:05 +00:00
|
|
|
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->SetCallBack(
|
2020-04-03 01:06:26 +00:00
|
|
|
"BackgroundCallCompaction:0", [&](void*) { fail_manifest.store(true); });
|
|
|
|
|
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->SetCallBack(
|
|
|
|
|
"VersionSet::LogAndApply:WriteManifest", [&](void*) {
|
|
|
|
|
if (fail_manifest.load()) {
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
fault_fs_->SetFilesystemActive(false, error_msg);
|
Revamp cache_bench to resemble a real workload (#6629)
Summary:
I suspect LRUCache could use some optimization, and to support
such an effort, a good benchmarking tool is needed. The existing
cache_bench was heavily skewed toward insertion and lookup misses, and
did not saturate memory with other work. This change should improve
those things to better resemble a real workload.
(All below using clang compiler, for some consistency, but not
necessarily same version and settings.)
The real workload is from production MySQL on RocksDB, filtering stacks
containing "LRU", "ShardedCache" or "CacheShard."
Lookup inclusive: 66%
Insert inclusive: 17%
Release inclusive: 15%
An alternate simulated workload is MySQL running a LinkBench read test:
Lookup inclusive: 54%
Insert inclusive: 24%
Release inclusive: 21%
cache_bench default settings, prior to this change:
Lookup inclusive: 35.8%
Insert inclusive: 63.6%
Release inclusive: 0%
cache_bench after this change (intended as somewhat "tighter" workload
than average production, more like LinkBench):
Lookup inclusive: 52%
Insert inclusive: 20%
Release inclusive: 26%
And top exclusive stacks (portion of stack samples as filtered above):
Production MySQL:
LRUHandleTable::FindPointer: 25.3%
rocksdb::operator==: 15.1% <-- Slice ==
LRUCacheShard::LRU_Remove: 13.8%
ShardedCache::Lookup: 8.9%
__pthread_mutex_lock: 7.1%
LRUCacheShard::LRU_Insert: 6.3%
MurmurHash64A: 4.8% <-- Since upgraded to XXH3p
...
Old cache_bench:
LRUHandleTable::FindPointer: 23.6%
__pthread_mutex_lock: 15.0%
__pthread_mutex_unlock_usercnt: 11.7%
__lll_lock_wait: 8.6%
__lll_unlock_wake: 6.8%
LRUCacheShard::LRU_Insert: 6.0%
ShardedCache::Lookup: 4.4%
LRUCacheShard::LRU_Remove: 2.8%
...
rocksdb::operator==: 0.2% <-- Slice ==
...
New cache_bench:
LRUHandleTable::FindPointer: 22.8%
__pthread_mutex_unlock_usercnt: 14.3%
rocksdb::operator==: 10.5% <-- Slice ==
LRUCacheShard::LRU_Insert: 9.0%
__pthread_mutex_lock: 5.9%
LRUCacheShard::LRU_Remove: 5.0%
...
ShardedCache::Lookup: 2.9%
...
So there's a bit more lock contention in the benchmark than in
production, but otherwise looks similar enough to me. At least it's a
big improvement over the existing code.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6629
Test Plan: No production code changes, ran cache_bench with ASAN
Reviewed By: ltamasi
Differential Revision: D20824318
Pulled By: pdillinger
fbshipit-source-id: 6f8dc5891ead0f87edbed3a615ecd5289d9abe12
2020-04-03 17:24:09 +00:00
|
|
|
}
|
|
|
|
|
});
|
Pass IOStatus to write path and set retryable IO Error as hard error in BG jobs (#6487)
Summary:
In the current code base, we use Status to get and store the returned status from the call. Specifically, for IO related functions, the current Status cannot reflect the IO Error details such as error scope, error retryable attribute, and others. With the implementation of https://github.com/facebook/rocksdb/issues/5761, we have the new Wrapper for IO, which returns IOStatus instead of Status. However, the IOStatus is purged at the lower level of write path and transferred to Status.
The first job of this PR is to pass the IOStatus to the write path (flush, WAL write, and Compaction). The second job is to identify the Retryable IO Error as HardError, and set the bg_error_ as HardError. In this case, the DB Instance becomes read only. User is informed of the Status and need to take actions to deal with it (e.g., call db->Resume()).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6487
Test Plan: Added the testing case to error_handler_fs_test. Pass make asan_check
Reviewed By: anand1976
Differential Revision: D20685017
Pulled By: zhichao-cao
fbshipit-source-id: ff85f042896243abcd6ef37877834e26f36b6eb0
2020-03-27 23:03:05 +00:00
|
|
|
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->EnableProcessing();
|
|
|
|
|
|
2020-10-02 23:39:17 +00:00
|
|
|
ASSERT_OK(Put(Key(1), "val"));
|
Pass IOStatus to write path and set retryable IO Error as hard error in BG jobs (#6487)
Summary:
In the current code base, we use Status to get and store the returned status from the call. Specifically, for IO related functions, the current Status cannot reflect the IO Error details such as error scope, error retryable attribute, and others. With the implementation of https://github.com/facebook/rocksdb/issues/5761, we have the new Wrapper for IO, which returns IOStatus instead of Status. However, the IOStatus is purged at the lower level of write path and transferred to Status.
The first job of this PR is to pass the IOStatus to the write path (flush, WAL write, and Compaction). The second job is to identify the Retryable IO Error as HardError, and set the bg_error_ as HardError. In this case, the DB Instance becomes read only. User is informed of the Status and need to take actions to deal with it (e.g., call db->Resume()).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6487
Test Plan: Added the testing case to error_handler_fs_test. Pass make asan_check
Reviewed By: anand1976
Differential Revision: D20685017
Pulled By: zhichao-cao
fbshipit-source-id: ff85f042896243abcd6ef37877834e26f36b6eb0
2020-03-27 23:03:05 +00:00
|
|
|
s = Flush();
|
2020-12-24 00:54:05 +00:00
|
|
|
ASSERT_OK(s);
|
Pass IOStatus to write path and set retryable IO Error as hard error in BG jobs (#6487)
Summary:
In the current code base, we use Status to get and store the returned status from the call. Specifically, for IO related functions, the current Status cannot reflect the IO Error details such as error scope, error retryable attribute, and others. With the implementation of https://github.com/facebook/rocksdb/issues/5761, we have the new Wrapper for IO, which returns IOStatus instead of Status. However, the IOStatus is purged at the lower level of write path and transferred to Status.
The first job of this PR is to pass the IOStatus to the write path (flush, WAL write, and Compaction). The second job is to identify the Retryable IO Error as HardError, and set the bg_error_ as HardError. In this case, the DB Instance becomes read only. User is informed of the Status and need to take actions to deal with it (e.g., call db->Resume()).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6487
Test Plan: Added the testing case to error_handler_fs_test. Pass make asan_check
Reviewed By: anand1976
Differential Revision: D20685017
Pulled By: zhichao-cao
fbshipit-source-id: ff85f042896243abcd6ef37877834e26f36b6eb0
2020-03-27 23:03:05 +00:00
|
|
|
|
2020-04-03 01:06:26 +00:00
|
|
|
TEST_SYNC_POINT("CompactionManifestWriteError:0");
|
2023-11-11 16:11:11 +00:00
|
|
|
ASSERT_FALSE(dbfull()->TEST_GetFilesToQuarantine().empty());
|
2020-04-03 01:06:26 +00:00
|
|
|
TEST_SYNC_POINT("CompactionManifestWriteError:1");
|
|
|
|
|
|
Pass IOStatus to write path and set retryable IO Error as hard error in BG jobs (#6487)
Summary:
In the current code base, we use Status to get and store the returned status from the call. Specifically, for IO related functions, the current Status cannot reflect the IO Error details such as error scope, error retryable attribute, and others. With the implementation of https://github.com/facebook/rocksdb/issues/5761, we have the new Wrapper for IO, which returns IOStatus instead of Status. However, the IOStatus is purged at the lower level of write path and transferred to Status.
The first job of this PR is to pass the IOStatus to the write path (flush, WAL write, and Compaction). The second job is to identify the Retryable IO Error as HardError, and set the bg_error_ as HardError. In this case, the DB Instance becomes read only. User is informed of the Status and need to take actions to deal with it (e.g., call db->Resume()).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6487
Test Plan: Added the testing case to error_handler_fs_test. Pass make asan_check
Reviewed By: anand1976
Differential Revision: D20685017
Pulled By: zhichao-cao
fbshipit-source-id: ff85f042896243abcd6ef37877834e26f36b6eb0
2020-03-27 23:03:05 +00:00
|
|
|
s = dbfull()->TEST_WaitForCompact();
|
|
|
|
|
ASSERT_EQ(s.severity(), ROCKSDB_NAMESPACE::Status::Severity::kHardError);
|
|
|
|
|
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
fault_fs_->SetFilesystemActive(true);
|
Pass IOStatus to write path and set retryable IO Error as hard error in BG jobs (#6487)
Summary:
In the current code base, we use Status to get and store the returned status from the call. Specifically, for IO related functions, the current Status cannot reflect the IO Error details such as error scope, error retryable attribute, and others. With the implementation of https://github.com/facebook/rocksdb/issues/5761, we have the new Wrapper for IO, which returns IOStatus instead of Status. However, the IOStatus is purged at the lower level of write path and transferred to Status.
The first job of this PR is to pass the IOStatus to the write path (flush, WAL write, and Compaction). The second job is to identify the Retryable IO Error as HardError, and set the bg_error_ as HardError. In this case, the DB Instance becomes read only. User is informed of the Status and need to take actions to deal with it (e.g., call db->Resume()).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6487
Test Plan: Added the testing case to error_handler_fs_test. Pass make asan_check
Reviewed By: anand1976
Differential Revision: D20685017
Pulled By: zhichao-cao
fbshipit-source-id: ff85f042896243abcd6ef37877834e26f36b6eb0
2020-03-27 23:03:05 +00:00
|
|
|
SyncPoint::GetInstance()->ClearAllCallBacks();
|
|
|
|
|
SyncPoint::GetInstance()->DisableProcessing();
|
|
|
|
|
s = dbfull()->Resume();
|
2020-12-24 00:54:05 +00:00
|
|
|
ASSERT_OK(s);
|
Pass IOStatus to write path and set retryable IO Error as hard error in BG jobs (#6487)
Summary:
In the current code base, we use Status to get and store the returned status from the call. Specifically, for IO related functions, the current Status cannot reflect the IO Error details such as error scope, error retryable attribute, and others. With the implementation of https://github.com/facebook/rocksdb/issues/5761, we have the new Wrapper for IO, which returns IOStatus instead of Status. However, the IOStatus is purged at the lower level of write path and transferred to Status.
The first job of this PR is to pass the IOStatus to the write path (flush, WAL write, and Compaction). The second job is to identify the Retryable IO Error as HardError, and set the bg_error_ as HardError. In this case, the DB Instance becomes read only. User is informed of the Status and need to take actions to deal with it (e.g., call db->Resume()).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6487
Test Plan: Added the testing case to error_handler_fs_test. Pass make asan_check
Reviewed By: anand1976
Differential Revision: D20685017
Pulled By: zhichao-cao
fbshipit-source-id: ff85f042896243abcd6ef37877834e26f36b6eb0
2020-03-27 23:03:05 +00:00
|
|
|
|
|
|
|
|
new_manifest = GetManifestNameFromLiveFiles();
|
|
|
|
|
ASSERT_NE(new_manifest, old_manifest);
|
2023-11-11 16:11:11 +00:00
|
|
|
ASSERT_TRUE(dbfull()->TEST_GetFilesToQuarantine().empty());
|
Pass IOStatus to write path and set retryable IO Error as hard error in BG jobs (#6487)
Summary:
In the current code base, we use Status to get and store the returned status from the call. Specifically, for IO related functions, the current Status cannot reflect the IO Error details such as error scope, error retryable attribute, and others. With the implementation of https://github.com/facebook/rocksdb/issues/5761, we have the new Wrapper for IO, which returns IOStatus instead of Status. However, the IOStatus is purged at the lower level of write path and transferred to Status.
The first job of this PR is to pass the IOStatus to the write path (flush, WAL write, and Compaction). The second job is to identify the Retryable IO Error as HardError, and set the bg_error_ as HardError. In this case, the DB Instance becomes read only. User is informed of the Status and need to take actions to deal with it (e.g., call db->Resume()).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6487
Test Plan: Added the testing case to error_handler_fs_test. Pass make asan_check
Reviewed By: anand1976
Differential Revision: D20685017
Pulled By: zhichao-cao
fbshipit-source-id: ff85f042896243abcd6ef37877834e26f36b6eb0
2020-03-27 23:03:05 +00:00
|
|
|
|
|
|
|
|
Reopen(options);
|
|
|
|
|
ASSERT_EQ("val", Get(Key(0)));
|
|
|
|
|
ASSERT_EQ("val", Get(Key(1)));
|
|
|
|
|
ASSERT_EQ("val", Get(Key(2)));
|
|
|
|
|
Close();
|
|
|
|
|
}
|
|
|
|
|
|
2020-03-04 20:30:34 +00:00
|
|
|
TEST_F(DBErrorHandlingFSTest, CompactionWriteError) {
|
2023-10-31 23:13:36 +00:00
|
|
|
std::shared_ptr<ErrorHandlerFSListener> listener =
|
|
|
|
|
std::make_shared<ErrorHandlerFSListener>();
|
2018-06-28 19:23:57 +00:00
|
|
|
Options options = GetDefaultOptions();
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
options.env = fault_env_.get();
|
2018-06-28 19:23:57 +00:00
|
|
|
options.create_if_missing = true;
|
|
|
|
|
options.level0_file_num_compaction_trigger = 2;
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
options.listeners.emplace_back(listener);
|
2018-06-28 19:23:57 +00:00
|
|
|
Status s;
|
|
|
|
|
DestroyAndReopen(options);
|
|
|
|
|
|
2020-10-02 23:39:17 +00:00
|
|
|
ASSERT_OK(Put(Key(0), "va;"));
|
|
|
|
|
ASSERT_OK(Put(Key(2), "va;"));
|
2018-06-28 19:23:57 +00:00
|
|
|
s = Flush();
|
2020-12-24 00:54:05 +00:00
|
|
|
ASSERT_OK(s);
|
2018-06-28 19:23:57 +00:00
|
|
|
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
listener->OverrideBGError(
|
2020-03-04 20:30:34 +00:00
|
|
|
Status(Status::NoSpace(), Status::Severity::kHardError));
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
listener->EnableAutoRecovery(false);
|
2020-02-20 20:07:53 +00:00
|
|
|
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->LoadDependency(
|
2019-12-12 22:05:48 +00:00
|
|
|
{{"DBImpl::FlushMemTable:FlushMemTableFinished",
|
|
|
|
|
"BackgroundCallCompaction:0"}});
|
2020-02-20 20:07:53 +00:00
|
|
|
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->SetCallBack(
|
|
|
|
|
"BackgroundCallCompaction:0", [&](void*) {
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
fault_fs_->SetFilesystemActive(false,
|
|
|
|
|
IOStatus::NoSpace("Out of space"));
|
2018-06-28 19:23:57 +00:00
|
|
|
});
|
2020-02-20 20:07:53 +00:00
|
|
|
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->EnableProcessing();
|
2018-06-28 19:23:57 +00:00
|
|
|
|
2020-10-02 23:39:17 +00:00
|
|
|
ASSERT_OK(Put(Key(1), "val"));
|
2018-06-28 19:23:57 +00:00
|
|
|
s = Flush();
|
2020-12-24 00:54:05 +00:00
|
|
|
ASSERT_OK(s);
|
2018-06-28 19:23:57 +00:00
|
|
|
|
|
|
|
|
s = dbfull()->TEST_WaitForCompact();
|
2020-02-20 20:07:53 +00:00
|
|
|
ASSERT_EQ(s.severity(), ROCKSDB_NAMESPACE::Status::Severity::kHardError);
|
2018-06-28 19:23:57 +00:00
|
|
|
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
fault_fs_->SetFilesystemActive(true);
|
2018-06-28 19:23:57 +00:00
|
|
|
s = dbfull()->Resume();
|
2020-12-24 00:54:05 +00:00
|
|
|
ASSERT_OK(s);
|
2018-06-28 19:23:57 +00:00
|
|
|
Destroy(options);
|
|
|
|
|
}
|
|
|
|
|
|
2021-02-15 17:43:11 +00:00
|
|
|
TEST_F(DBErrorHandlingFSTest, DISABLED_CompactionWriteRetryableError) {
|
2023-10-31 23:13:36 +00:00
|
|
|
std::shared_ptr<ErrorHandlerFSListener> listener =
|
|
|
|
|
std::make_shared<ErrorHandlerFSListener>();
|
Pass IOStatus to write path and set retryable IO Error as hard error in BG jobs (#6487)
Summary:
In the current code base, we use Status to get and store the returned status from the call. Specifically, for IO related functions, the current Status cannot reflect the IO Error details such as error scope, error retryable attribute, and others. With the implementation of https://github.com/facebook/rocksdb/issues/5761, we have the new Wrapper for IO, which returns IOStatus instead of Status. However, the IOStatus is purged at the lower level of write path and transferred to Status.
The first job of this PR is to pass the IOStatus to the write path (flush, WAL write, and Compaction). The second job is to identify the Retryable IO Error as HardError, and set the bg_error_ as HardError. In this case, the DB Instance becomes read only. User is informed of the Status and need to take actions to deal with it (e.g., call db->Resume()).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6487
Test Plan: Added the testing case to error_handler_fs_test. Pass make asan_check
Reviewed By: anand1976
Differential Revision: D20685017
Pulled By: zhichao-cao
fbshipit-source-id: ff85f042896243abcd6ef37877834e26f36b6eb0
2020-03-27 23:03:05 +00:00
|
|
|
Options options = GetDefaultOptions();
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
options.env = fault_env_.get();
|
Pass IOStatus to write path and set retryable IO Error as hard error in BG jobs (#6487)
Summary:
In the current code base, we use Status to get and store the returned status from the call. Specifically, for IO related functions, the current Status cannot reflect the IO Error details such as error scope, error retryable attribute, and others. With the implementation of https://github.com/facebook/rocksdb/issues/5761, we have the new Wrapper for IO, which returns IOStatus instead of Status. However, the IOStatus is purged at the lower level of write path and transferred to Status.
The first job of this PR is to pass the IOStatus to the write path (flush, WAL write, and Compaction). The second job is to identify the Retryable IO Error as HardError, and set the bg_error_ as HardError. In this case, the DB Instance becomes read only. User is informed of the Status and need to take actions to deal with it (e.g., call db->Resume()).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6487
Test Plan: Added the testing case to error_handler_fs_test. Pass make asan_check
Reviewed By: anand1976
Differential Revision: D20685017
Pulled By: zhichao-cao
fbshipit-source-id: ff85f042896243abcd6ef37877834e26f36b6eb0
2020-03-27 23:03:05 +00:00
|
|
|
options.create_if_missing = true;
|
|
|
|
|
options.level0_file_num_compaction_trigger = 2;
|
|
|
|
|
options.listeners.emplace_back(listener);
|
2020-07-15 18:02:44 +00:00
|
|
|
options.max_bgerror_resume_count = 0;
|
Pass IOStatus to write path and set retryable IO Error as hard error in BG jobs (#6487)
Summary:
In the current code base, we use Status to get and store the returned status from the call. Specifically, for IO related functions, the current Status cannot reflect the IO Error details such as error scope, error retryable attribute, and others. With the implementation of https://github.com/facebook/rocksdb/issues/5761, we have the new Wrapper for IO, which returns IOStatus instead of Status. However, the IOStatus is purged at the lower level of write path and transferred to Status.
The first job of this PR is to pass the IOStatus to the write path (flush, WAL write, and Compaction). The second job is to identify the Retryable IO Error as HardError, and set the bg_error_ as HardError. In this case, the DB Instance becomes read only. User is informed of the Status and need to take actions to deal with it (e.g., call db->Resume()).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6487
Test Plan: Added the testing case to error_handler_fs_test. Pass make asan_check
Reviewed By: anand1976
Differential Revision: D20685017
Pulled By: zhichao-cao
fbshipit-source-id: ff85f042896243abcd6ef37877834e26f36b6eb0
2020-03-27 23:03:05 +00:00
|
|
|
Status s;
|
|
|
|
|
DestroyAndReopen(options);
|
|
|
|
|
|
|
|
|
|
IOStatus error_msg = IOStatus::IOError("Retryable IO Error");
|
|
|
|
|
error_msg.SetRetryable(true);
|
|
|
|
|
|
2020-10-02 23:39:17 +00:00
|
|
|
ASSERT_OK(Put(Key(0), "va;"));
|
|
|
|
|
ASSERT_OK(Put(Key(2), "va;"));
|
Pass IOStatus to write path and set retryable IO Error as hard error in BG jobs (#6487)
Summary:
In the current code base, we use Status to get and store the returned status from the call. Specifically, for IO related functions, the current Status cannot reflect the IO Error details such as error scope, error retryable attribute, and others. With the implementation of https://github.com/facebook/rocksdb/issues/5761, we have the new Wrapper for IO, which returns IOStatus instead of Status. However, the IOStatus is purged at the lower level of write path and transferred to Status.
The first job of this PR is to pass the IOStatus to the write path (flush, WAL write, and Compaction). The second job is to identify the Retryable IO Error as HardError, and set the bg_error_ as HardError. In this case, the DB Instance becomes read only. User is informed of the Status and need to take actions to deal with it (e.g., call db->Resume()).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6487
Test Plan: Added the testing case to error_handler_fs_test. Pass make asan_check
Reviewed By: anand1976
Differential Revision: D20685017
Pulled By: zhichao-cao
fbshipit-source-id: ff85f042896243abcd6ef37877834e26f36b6eb0
2020-03-27 23:03:05 +00:00
|
|
|
s = Flush();
|
2020-12-24 00:54:05 +00:00
|
|
|
ASSERT_OK(s);
|
Pass IOStatus to write path and set retryable IO Error as hard error in BG jobs (#6487)
Summary:
In the current code base, we use Status to get and store the returned status from the call. Specifically, for IO related functions, the current Status cannot reflect the IO Error details such as error scope, error retryable attribute, and others. With the implementation of https://github.com/facebook/rocksdb/issues/5761, we have the new Wrapper for IO, which returns IOStatus instead of Status. However, the IOStatus is purged at the lower level of write path and transferred to Status.
The first job of this PR is to pass the IOStatus to the write path (flush, WAL write, and Compaction). The second job is to identify the Retryable IO Error as HardError, and set the bg_error_ as HardError. In this case, the DB Instance becomes read only. User is informed of the Status and need to take actions to deal with it (e.g., call db->Resume()).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6487
Test Plan: Added the testing case to error_handler_fs_test. Pass make asan_check
Reviewed By: anand1976
Differential Revision: D20685017
Pulled By: zhichao-cao
fbshipit-source-id: ff85f042896243abcd6ef37877834e26f36b6eb0
2020-03-27 23:03:05 +00:00
|
|
|
|
|
|
|
|
listener->OverrideBGError(Status(error_msg, Status::Severity::kHardError));
|
|
|
|
|
listener->EnableAutoRecovery(false);
|
|
|
|
|
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->LoadDependency(
|
|
|
|
|
{{"DBImpl::FlushMemTable:FlushMemTableFinished",
|
|
|
|
|
"BackgroundCallCompaction:0"}});
|
|
|
|
|
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->SetCallBack(
|
2020-07-15 18:02:44 +00:00
|
|
|
"CompactionJob::OpenCompactionOutputFile",
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
[&](void*) { fault_fs_->SetFilesystemActive(false, error_msg); });
|
2021-01-28 01:56:17 +00:00
|
|
|
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->SetCallBack(
|
|
|
|
|
"DBImpl::BackgroundCompaction:Finish",
|
|
|
|
|
[&](void*) { CancelAllBackgroundWork(dbfull()); });
|
Pass IOStatus to write path and set retryable IO Error as hard error in BG jobs (#6487)
Summary:
In the current code base, we use Status to get and store the returned status from the call. Specifically, for IO related functions, the current Status cannot reflect the IO Error details such as error scope, error retryable attribute, and others. With the implementation of https://github.com/facebook/rocksdb/issues/5761, we have the new Wrapper for IO, which returns IOStatus instead of Status. However, the IOStatus is purged at the lower level of write path and transferred to Status.
The first job of this PR is to pass the IOStatus to the write path (flush, WAL write, and Compaction). The second job is to identify the Retryable IO Error as HardError, and set the bg_error_ as HardError. In this case, the DB Instance becomes read only. User is informed of the Status and need to take actions to deal with it (e.g., call db->Resume()).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6487
Test Plan: Added the testing case to error_handler_fs_test. Pass make asan_check
Reviewed By: anand1976
Differential Revision: D20685017
Pulled By: zhichao-cao
fbshipit-source-id: ff85f042896243abcd6ef37877834e26f36b6eb0
2020-03-27 23:03:05 +00:00
|
|
|
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->EnableProcessing();
|
|
|
|
|
|
2020-10-02 23:39:17 +00:00
|
|
|
ASSERT_OK(Put(Key(1), "val"));
|
Pass IOStatus to write path and set retryable IO Error as hard error in BG jobs (#6487)
Summary:
In the current code base, we use Status to get and store the returned status from the call. Specifically, for IO related functions, the current Status cannot reflect the IO Error details such as error scope, error retryable attribute, and others. With the implementation of https://github.com/facebook/rocksdb/issues/5761, we have the new Wrapper for IO, which returns IOStatus instead of Status. However, the IOStatus is purged at the lower level of write path and transferred to Status.
The first job of this PR is to pass the IOStatus to the write path (flush, WAL write, and Compaction). The second job is to identify the Retryable IO Error as HardError, and set the bg_error_ as HardError. In this case, the DB Instance becomes read only. User is informed of the Status and need to take actions to deal with it (e.g., call db->Resume()).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6487
Test Plan: Added the testing case to error_handler_fs_test. Pass make asan_check
Reviewed By: anand1976
Differential Revision: D20685017
Pulled By: zhichao-cao
fbshipit-source-id: ff85f042896243abcd6ef37877834e26f36b6eb0
2020-03-27 23:03:05 +00:00
|
|
|
s = Flush();
|
2020-12-24 00:54:05 +00:00
|
|
|
ASSERT_OK(s);
|
Pass IOStatus to write path and set retryable IO Error as hard error in BG jobs (#6487)
Summary:
In the current code base, we use Status to get and store the returned status from the call. Specifically, for IO related functions, the current Status cannot reflect the IO Error details such as error scope, error retryable attribute, and others. With the implementation of https://github.com/facebook/rocksdb/issues/5761, we have the new Wrapper for IO, which returns IOStatus instead of Status. However, the IOStatus is purged at the lower level of write path and transferred to Status.
The first job of this PR is to pass the IOStatus to the write path (flush, WAL write, and Compaction). The second job is to identify the Retryable IO Error as HardError, and set the bg_error_ as HardError. In this case, the DB Instance becomes read only. User is informed of the Status and need to take actions to deal with it (e.g., call db->Resume()).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6487
Test Plan: Added the testing case to error_handler_fs_test. Pass make asan_check
Reviewed By: anand1976
Differential Revision: D20685017
Pulled By: zhichao-cao
fbshipit-source-id: ff85f042896243abcd6ef37877834e26f36b6eb0
2020-03-27 23:03:05 +00:00
|
|
|
|
2021-01-28 01:56:17 +00:00
|
|
|
s = dbfull()->TEST_GetBGError();
|
|
|
|
|
ASSERT_OK(s);
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
fault_fs_->SetFilesystemActive(true);
|
Pass IOStatus to write path and set retryable IO Error as hard error in BG jobs (#6487)
Summary:
In the current code base, we use Status to get and store the returned status from the call. Specifically, for IO related functions, the current Status cannot reflect the IO Error details such as error scope, error retryable attribute, and others. With the implementation of https://github.com/facebook/rocksdb/issues/5761, we have the new Wrapper for IO, which returns IOStatus instead of Status. However, the IOStatus is purged at the lower level of write path and transferred to Status.
The first job of this PR is to pass the IOStatus to the write path (flush, WAL write, and Compaction). The second job is to identify the Retryable IO Error as HardError, and set the bg_error_ as HardError. In this case, the DB Instance becomes read only. User is informed of the Status and need to take actions to deal with it (e.g., call db->Resume()).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6487
Test Plan: Added the testing case to error_handler_fs_test. Pass make asan_check
Reviewed By: anand1976
Differential Revision: D20685017
Pulled By: zhichao-cao
fbshipit-source-id: ff85f042896243abcd6ef37877834e26f36b6eb0
2020-03-27 23:03:05 +00:00
|
|
|
SyncPoint::GetInstance()->ClearAllCallBacks();
|
2021-01-08 00:30:06 +00:00
|
|
|
SyncPoint::GetInstance()->DisableProcessing();
|
|
|
|
|
s = dbfull()->Resume();
|
|
|
|
|
ASSERT_OK(s);
|
|
|
|
|
Destroy(options);
|
|
|
|
|
}
|
|
|
|
|
|
2021-02-15 17:43:11 +00:00
|
|
|
TEST_F(DBErrorHandlingFSTest, DISABLED_CompactionWriteFileScopeError) {
|
2023-10-31 23:13:36 +00:00
|
|
|
std::shared_ptr<ErrorHandlerFSListener> listener =
|
|
|
|
|
std::make_shared<ErrorHandlerFSListener>();
|
2021-01-08 00:30:06 +00:00
|
|
|
Options options = GetDefaultOptions();
|
|
|
|
|
options.env = fault_env_.get();
|
|
|
|
|
options.create_if_missing = true;
|
|
|
|
|
options.level0_file_num_compaction_trigger = 2;
|
|
|
|
|
options.listeners.emplace_back(listener);
|
|
|
|
|
options.max_bgerror_resume_count = 0;
|
|
|
|
|
Status s;
|
|
|
|
|
DestroyAndReopen(options);
|
|
|
|
|
|
|
|
|
|
IOStatus error_msg = IOStatus::IOError("File Scope Data Loss Error");
|
|
|
|
|
error_msg.SetDataLoss(true);
|
|
|
|
|
error_msg.SetScope(
|
|
|
|
|
ROCKSDB_NAMESPACE::IOStatus::IOErrorScope::kIOErrorScopeFile);
|
|
|
|
|
error_msg.SetRetryable(false);
|
|
|
|
|
|
|
|
|
|
ASSERT_OK(Put(Key(0), "va;"));
|
|
|
|
|
ASSERT_OK(Put(Key(2), "va;"));
|
|
|
|
|
s = Flush();
|
|
|
|
|
ASSERT_OK(s);
|
|
|
|
|
|
|
|
|
|
listener->OverrideBGError(Status(error_msg, Status::Severity::kHardError));
|
|
|
|
|
listener->EnableAutoRecovery(false);
|
|
|
|
|
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->LoadDependency(
|
|
|
|
|
{{"DBImpl::FlushMemTable:FlushMemTableFinished",
|
|
|
|
|
"BackgroundCallCompaction:0"}});
|
|
|
|
|
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->SetCallBack(
|
|
|
|
|
"CompactionJob::OpenCompactionOutputFile",
|
|
|
|
|
[&](void*) { fault_fs_->SetFilesystemActive(false, error_msg); });
|
2021-01-28 01:56:17 +00:00
|
|
|
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->SetCallBack(
|
|
|
|
|
"DBImpl::BackgroundCompaction:Finish",
|
|
|
|
|
[&](void*) { CancelAllBackgroundWork(dbfull()); });
|
2021-01-08 00:30:06 +00:00
|
|
|
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->EnableProcessing();
|
|
|
|
|
|
|
|
|
|
ASSERT_OK(Put(Key(1), "val"));
|
|
|
|
|
s = Flush();
|
|
|
|
|
ASSERT_OK(s);
|
|
|
|
|
|
2021-01-28 01:56:17 +00:00
|
|
|
s = dbfull()->TEST_GetBGError();
|
|
|
|
|
ASSERT_OK(s);
|
2021-01-08 00:30:06 +00:00
|
|
|
|
|
|
|
|
fault_fs_->SetFilesystemActive(true);
|
|
|
|
|
SyncPoint::GetInstance()->ClearAllCallBacks();
|
Pass IOStatus to write path and set retryable IO Error as hard error in BG jobs (#6487)
Summary:
In the current code base, we use Status to get and store the returned status from the call. Specifically, for IO related functions, the current Status cannot reflect the IO Error details such as error scope, error retryable attribute, and others. With the implementation of https://github.com/facebook/rocksdb/issues/5761, we have the new Wrapper for IO, which returns IOStatus instead of Status. However, the IOStatus is purged at the lower level of write path and transferred to Status.
The first job of this PR is to pass the IOStatus to the write path (flush, WAL write, and Compaction). The second job is to identify the Retryable IO Error as HardError, and set the bg_error_ as HardError. In this case, the DB Instance becomes read only. User is informed of the Status and need to take actions to deal with it (e.g., call db->Resume()).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6487
Test Plan: Added the testing case to error_handler_fs_test. Pass make asan_check
Reviewed By: anand1976
Differential Revision: D20685017
Pulled By: zhichao-cao
fbshipit-source-id: ff85f042896243abcd6ef37877834e26f36b6eb0
2020-03-27 23:03:05 +00:00
|
|
|
SyncPoint::GetInstance()->DisableProcessing();
|
|
|
|
|
s = dbfull()->Resume();
|
2020-12-24 00:54:05 +00:00
|
|
|
ASSERT_OK(s);
|
Pass IOStatus to write path and set retryable IO Error as hard error in BG jobs (#6487)
Summary:
In the current code base, we use Status to get and store the returned status from the call. Specifically, for IO related functions, the current Status cannot reflect the IO Error details such as error scope, error retryable attribute, and others. With the implementation of https://github.com/facebook/rocksdb/issues/5761, we have the new Wrapper for IO, which returns IOStatus instead of Status. However, the IOStatus is purged at the lower level of write path and transferred to Status.
The first job of this PR is to pass the IOStatus to the write path (flush, WAL write, and Compaction). The second job is to identify the Retryable IO Error as HardError, and set the bg_error_ as HardError. In this case, the DB Instance becomes read only. User is informed of the Status and need to take actions to deal with it (e.g., call db->Resume()).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6487
Test Plan: Added the testing case to error_handler_fs_test. Pass make asan_check
Reviewed By: anand1976
Differential Revision: D20685017
Pulled By: zhichao-cao
fbshipit-source-id: ff85f042896243abcd6ef37877834e26f36b6eb0
2020-03-27 23:03:05 +00:00
|
|
|
Destroy(options);
|
|
|
|
|
}
|
|
|
|
|
|
2020-03-04 20:30:34 +00:00
|
|
|
TEST_F(DBErrorHandlingFSTest, CorruptionError) {
|
2018-06-28 19:23:57 +00:00
|
|
|
Options options = GetDefaultOptions();
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
options.env = fault_env_.get();
|
2018-06-28 19:23:57 +00:00
|
|
|
options.create_if_missing = true;
|
|
|
|
|
options.level0_file_num_compaction_trigger = 2;
|
|
|
|
|
Status s;
|
|
|
|
|
DestroyAndReopen(options);
|
|
|
|
|
|
2020-10-02 23:39:17 +00:00
|
|
|
ASSERT_OK(Put(Key(0), "va;"));
|
|
|
|
|
ASSERT_OK(Put(Key(2), "va;"));
|
2018-06-28 19:23:57 +00:00
|
|
|
s = Flush();
|
2020-12-24 00:54:05 +00:00
|
|
|
ASSERT_OK(s);
|
2018-06-28 19:23:57 +00:00
|
|
|
|
2020-02-20 20:07:53 +00:00
|
|
|
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->LoadDependency(
|
2019-12-12 22:05:48 +00:00
|
|
|
{{"DBImpl::FlushMemTable:FlushMemTableFinished",
|
|
|
|
|
"BackgroundCallCompaction:0"}});
|
2020-02-20 20:07:53 +00:00
|
|
|
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->SetCallBack(
|
|
|
|
|
"BackgroundCallCompaction:0", [&](void*) {
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
fault_fs_->SetFilesystemActive(false,
|
|
|
|
|
IOStatus::Corruption("Corruption"));
|
2018-06-28 19:23:57 +00:00
|
|
|
});
|
2020-02-20 20:07:53 +00:00
|
|
|
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->EnableProcessing();
|
2018-06-28 19:23:57 +00:00
|
|
|
|
2020-10-02 23:39:17 +00:00
|
|
|
ASSERT_OK(Put(Key(1), "val"));
|
2018-06-28 19:23:57 +00:00
|
|
|
s = Flush();
|
2020-12-24 00:54:05 +00:00
|
|
|
ASSERT_OK(s);
|
2018-06-28 19:23:57 +00:00
|
|
|
|
|
|
|
|
s = dbfull()->TEST_WaitForCompact();
|
2020-02-20 20:07:53 +00:00
|
|
|
ASSERT_EQ(s.severity(),
|
|
|
|
|
ROCKSDB_NAMESPACE::Status::Severity::kUnrecoverableError);
|
2018-06-28 19:23:57 +00:00
|
|
|
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
fault_fs_->SetFilesystemActive(true);
|
2018-06-28 19:23:57 +00:00
|
|
|
s = dbfull()->Resume();
|
2020-12-24 00:54:05 +00:00
|
|
|
ASSERT_NOK(s);
|
2018-06-28 19:23:57 +00:00
|
|
|
Destroy(options);
|
|
|
|
|
}
|
|
|
|
|
|
2020-03-04 20:30:34 +00:00
|
|
|
TEST_F(DBErrorHandlingFSTest, AutoRecoverFlushError) {
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
if (mem_env_ != nullptr) {
|
|
|
|
|
ROCKSDB_GTEST_SKIP("Test requires non-mock environment");
|
|
|
|
|
return;
|
|
|
|
|
}
|
2023-10-31 23:13:36 +00:00
|
|
|
std::shared_ptr<ErrorHandlerFSListener> listener =
|
|
|
|
|
std::make_shared<ErrorHandlerFSListener>();
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
Options options = GetDefaultOptions();
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
options.env = fault_env_.get();
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
options.create_if_missing = true;
|
|
|
|
|
options.listeners.emplace_back(listener);
|
2021-03-18 05:36:42 +00:00
|
|
|
options.statistics = CreateDBStatistics();
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
Status s;
|
|
|
|
|
|
|
|
|
|
listener->EnableAutoRecovery();
|
|
|
|
|
DestroyAndReopen(options);
|
|
|
|
|
|
2020-10-02 23:39:17 +00:00
|
|
|
ASSERT_OK(Put(Key(0), "val"));
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
SyncPoint::GetInstance()->SetCallBack("FlushJob::Start", [&](void*) {
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
fault_fs_->SetFilesystemActive(false, IOStatus::NoSpace("Out of space"));
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
});
|
|
|
|
|
SyncPoint::GetInstance()->EnableProcessing();
|
|
|
|
|
s = Flush();
|
2020-02-20 20:07:53 +00:00
|
|
|
ASSERT_EQ(s.severity(), ROCKSDB_NAMESPACE::Status::Severity::kHardError);
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
SyncPoint::GetInstance()->DisableProcessing();
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
fault_fs_->SetFilesystemActive(true);
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
ASSERT_EQ(listener->WaitForRecovery(5000000), true);
|
|
|
|
|
|
|
|
|
|
s = Put(Key(1), "val");
|
2020-12-24 00:54:05 +00:00
|
|
|
ASSERT_OK(s);
|
2021-03-18 05:36:42 +00:00
|
|
|
ASSERT_EQ(1, options.statistics->getAndResetTickerCount(
|
|
|
|
|
ERROR_HANDLER_BG_ERROR_COUNT));
|
|
|
|
|
ASSERT_EQ(1, options.statistics->getAndResetTickerCount(
|
|
|
|
|
ERROR_HANDLER_BG_IO_ERROR_COUNT));
|
|
|
|
|
ASSERT_EQ(0, options.statistics->getAndResetTickerCount(
|
|
|
|
|
ERROR_HANDLER_BG_RETRYABLE_IO_ERROR_COUNT));
|
|
|
|
|
ASSERT_EQ(0, options.statistics->getAndResetTickerCount(
|
|
|
|
|
ERROR_HANDLER_AUTORESUME_COUNT));
|
|
|
|
|
ASSERT_EQ(0, options.statistics->getAndResetTickerCount(
|
|
|
|
|
ERROR_HANDLER_AUTORESUME_RETRY_TOTAL_COUNT));
|
|
|
|
|
ASSERT_EQ(0, options.statistics->getAndResetTickerCount(
|
|
|
|
|
ERROR_HANDLER_AUTORESUME_SUCCESS_COUNT));
|
2024-04-12 16:57:56 +00:00
|
|
|
ASSERT_OK(dbfull()->SyncWAL());
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
|
|
|
|
|
Reopen(options);
|
|
|
|
|
ASSERT_EQ("val", Get(Key(0)));
|
|
|
|
|
ASSERT_EQ("val", Get(Key(1)));
|
|
|
|
|
Destroy(options);
|
|
|
|
|
}
|
|
|
|
|
|
2020-03-04 20:30:34 +00:00
|
|
|
TEST_F(DBErrorHandlingFSTest, FailRecoverFlushError) {
|
2023-10-31 23:13:36 +00:00
|
|
|
std::shared_ptr<ErrorHandlerFSListener> listener =
|
|
|
|
|
std::make_shared<ErrorHandlerFSListener>();
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
Options options = GetDefaultOptions();
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
options.env = fault_env_.get();
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
options.create_if_missing = true;
|
|
|
|
|
options.listeners.emplace_back(listener);
|
|
|
|
|
Status s;
|
|
|
|
|
|
|
|
|
|
listener->EnableAutoRecovery();
|
|
|
|
|
DestroyAndReopen(options);
|
|
|
|
|
|
2020-10-02 23:39:17 +00:00
|
|
|
ASSERT_OK(Put(Key(0), "val"));
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
SyncPoint::GetInstance()->SetCallBack("FlushJob::Start", [&](void*) {
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
fault_fs_->SetFilesystemActive(false, IOStatus::NoSpace("Out of space"));
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
});
|
|
|
|
|
SyncPoint::GetInstance()->EnableProcessing();
|
|
|
|
|
s = Flush();
|
2020-02-20 20:07:53 +00:00
|
|
|
ASSERT_EQ(s.severity(), ROCKSDB_NAMESPACE::Status::Severity::kHardError);
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
// We should be able to shutdown the database while auto recovery is going
|
|
|
|
|
// on in the background
|
|
|
|
|
Close();
|
2020-12-24 00:54:05 +00:00
|
|
|
DestroyDB(dbname_, options).PermitUncheckedError();
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
}
|
|
|
|
|
|
2020-03-04 20:30:34 +00:00
|
|
|
TEST_F(DBErrorHandlingFSTest, WALWriteError) {
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
if (mem_env_ != nullptr) {
|
|
|
|
|
ROCKSDB_GTEST_SKIP("Test requires non-mock environment");
|
|
|
|
|
return;
|
|
|
|
|
}
|
2023-10-31 23:13:36 +00:00
|
|
|
std::shared_ptr<ErrorHandlerFSListener> listener =
|
|
|
|
|
std::make_shared<ErrorHandlerFSListener>();
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
Options options = GetDefaultOptions();
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
options.env = fault_env_.get();
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
options.create_if_missing = true;
|
|
|
|
|
options.writable_file_max_buffer_size = 32768;
|
|
|
|
|
options.listeners.emplace_back(listener);
|
|
|
|
|
Status s;
|
2018-09-17 20:08:13 +00:00
|
|
|
Random rnd(301);
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
|
|
|
|
|
listener->EnableAutoRecovery();
|
|
|
|
|
DestroyAndReopen(options);
|
|
|
|
|
|
|
|
|
|
{
|
|
|
|
|
WriteBatch batch;
|
|
|
|
|
|
2020-03-04 20:30:34 +00:00
|
|
|
for (auto i = 0; i < 100; ++i) {
|
2020-10-02 23:39:17 +00:00
|
|
|
ASSERT_OK(batch.Put(Key(i), rnd.RandomString(1024)));
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
WriteOptions wopts;
|
|
|
|
|
wopts.sync = true;
|
2020-12-24 00:54:05 +00:00
|
|
|
ASSERT_OK(dbfull()->Write(wopts, &batch));
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
};
|
|
|
|
|
|
|
|
|
|
{
|
|
|
|
|
WriteBatch batch;
|
|
|
|
|
int write_error = 0;
|
|
|
|
|
|
2020-03-04 20:30:34 +00:00
|
|
|
for (auto i = 100; i < 199; ++i) {
|
2020-10-02 23:39:17 +00:00
|
|
|
ASSERT_OK(batch.Put(Key(i), rnd.RandomString(1024)));
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
}
|
|
|
|
|
|
2020-03-04 20:30:34 +00:00
|
|
|
SyncPoint::GetInstance()->SetCallBack(
|
|
|
|
|
"WritableFileWriter::Append:BeforePrepareWrite", [&](void*) {
|
|
|
|
|
write_error++;
|
|
|
|
|
if (write_error > 2) {
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
fault_fs_->SetFilesystemActive(false,
|
|
|
|
|
IOStatus::NoSpace("Out of space"));
|
2020-03-04 20:30:34 +00:00
|
|
|
}
|
|
|
|
|
});
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
SyncPoint::GetInstance()->EnableProcessing();
|
|
|
|
|
WriteOptions wopts;
|
|
|
|
|
wopts.sync = true;
|
|
|
|
|
s = dbfull()->Write(wopts, &batch);
|
|
|
|
|
ASSERT_EQ(s, s.NoSpace());
|
|
|
|
|
}
|
|
|
|
|
SyncPoint::GetInstance()->DisableProcessing();
|
2022-09-06 19:59:02 +00:00
|
|
|
// `ClearAllCallBacks()` is needed in addition to `DisableProcessing()` to
|
|
|
|
|
// drain all callbacks. Otherwise, a pending callback in the background
|
|
|
|
|
// could re-disable `fault_fs_` after we enable it below.
|
|
|
|
|
SyncPoint::GetInstance()->ClearAllCallBacks();
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
fault_fs_->SetFilesystemActive(true);
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
ASSERT_EQ(listener->WaitForRecovery(5000000), true);
|
2020-03-04 20:30:34 +00:00
|
|
|
for (auto i = 0; i < 199; ++i) {
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
if (i < 100) {
|
|
|
|
|
ASSERT_NE(Get(Key(i)), "NOT_FOUND");
|
|
|
|
|
} else {
|
|
|
|
|
ASSERT_EQ(Get(Key(i)), "NOT_FOUND");
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
Reopen(options);
|
2020-03-04 20:30:34 +00:00
|
|
|
for (auto i = 0; i < 199; ++i) {
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
if (i < 100) {
|
|
|
|
|
ASSERT_NE(Get(Key(i)), "NOT_FOUND");
|
|
|
|
|
} else {
|
|
|
|
|
ASSERT_EQ(Get(Key(i)), "NOT_FOUND");
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
Close();
|
|
|
|
|
}
|
|
|
|
|
|
Pass IOStatus to write path and set retryable IO Error as hard error in BG jobs (#6487)
Summary:
In the current code base, we use Status to get and store the returned status from the call. Specifically, for IO related functions, the current Status cannot reflect the IO Error details such as error scope, error retryable attribute, and others. With the implementation of https://github.com/facebook/rocksdb/issues/5761, we have the new Wrapper for IO, which returns IOStatus instead of Status. However, the IOStatus is purged at the lower level of write path and transferred to Status.
The first job of this PR is to pass the IOStatus to the write path (flush, WAL write, and Compaction). The second job is to identify the Retryable IO Error as HardError, and set the bg_error_ as HardError. In this case, the DB Instance becomes read only. User is informed of the Status and need to take actions to deal with it (e.g., call db->Resume()).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6487
Test Plan: Added the testing case to error_handler_fs_test. Pass make asan_check
Reviewed By: anand1976
Differential Revision: D20685017
Pulled By: zhichao-cao
fbshipit-source-id: ff85f042896243abcd6ef37877834e26f36b6eb0
2020-03-27 23:03:05 +00:00
|
|
|
TEST_F(DBErrorHandlingFSTest, WALWriteRetryableError) {
|
2023-10-31 23:13:36 +00:00
|
|
|
std::shared_ptr<ErrorHandlerFSListener> listener =
|
|
|
|
|
std::make_shared<ErrorHandlerFSListener>();
|
Pass IOStatus to write path and set retryable IO Error as hard error in BG jobs (#6487)
Summary:
In the current code base, we use Status to get and store the returned status from the call. Specifically, for IO related functions, the current Status cannot reflect the IO Error details such as error scope, error retryable attribute, and others. With the implementation of https://github.com/facebook/rocksdb/issues/5761, we have the new Wrapper for IO, which returns IOStatus instead of Status. However, the IOStatus is purged at the lower level of write path and transferred to Status.
The first job of this PR is to pass the IOStatus to the write path (flush, WAL write, and Compaction). The second job is to identify the Retryable IO Error as HardError, and set the bg_error_ as HardError. In this case, the DB Instance becomes read only. User is informed of the Status and need to take actions to deal with it (e.g., call db->Resume()).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6487
Test Plan: Added the testing case to error_handler_fs_test. Pass make asan_check
Reviewed By: anand1976
Differential Revision: D20685017
Pulled By: zhichao-cao
fbshipit-source-id: ff85f042896243abcd6ef37877834e26f36b6eb0
2020-03-27 23:03:05 +00:00
|
|
|
Options options = GetDefaultOptions();
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
options.env = fault_env_.get();
|
Pass IOStatus to write path and set retryable IO Error as hard error in BG jobs (#6487)
Summary:
In the current code base, we use Status to get and store the returned status from the call. Specifically, for IO related functions, the current Status cannot reflect the IO Error details such as error scope, error retryable attribute, and others. With the implementation of https://github.com/facebook/rocksdb/issues/5761, we have the new Wrapper for IO, which returns IOStatus instead of Status. However, the IOStatus is purged at the lower level of write path and transferred to Status.
The first job of this PR is to pass the IOStatus to the write path (flush, WAL write, and Compaction). The second job is to identify the Retryable IO Error as HardError, and set the bg_error_ as HardError. In this case, the DB Instance becomes read only. User is informed of the Status and need to take actions to deal with it (e.g., call db->Resume()).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6487
Test Plan: Added the testing case to error_handler_fs_test. Pass make asan_check
Reviewed By: anand1976
Differential Revision: D20685017
Pulled By: zhichao-cao
fbshipit-source-id: ff85f042896243abcd6ef37877834e26f36b6eb0
2020-03-27 23:03:05 +00:00
|
|
|
options.create_if_missing = true;
|
|
|
|
|
options.writable_file_max_buffer_size = 32768;
|
|
|
|
|
options.listeners.emplace_back(listener);
|
|
|
|
|
options.paranoid_checks = true;
|
2020-07-15 18:02:44 +00:00
|
|
|
options.max_bgerror_resume_count = 0;
|
Pass IOStatus to write path and set retryable IO Error as hard error in BG jobs (#6487)
Summary:
In the current code base, we use Status to get and store the returned status from the call. Specifically, for IO related functions, the current Status cannot reflect the IO Error details such as error scope, error retryable attribute, and others. With the implementation of https://github.com/facebook/rocksdb/issues/5761, we have the new Wrapper for IO, which returns IOStatus instead of Status. However, the IOStatus is purged at the lower level of write path and transferred to Status.
The first job of this PR is to pass the IOStatus to the write path (flush, WAL write, and Compaction). The second job is to identify the Retryable IO Error as HardError, and set the bg_error_ as HardError. In this case, the DB Instance becomes read only. User is informed of the Status and need to take actions to deal with it (e.g., call db->Resume()).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6487
Test Plan: Added the testing case to error_handler_fs_test. Pass make asan_check
Reviewed By: anand1976
Differential Revision: D20685017
Pulled By: zhichao-cao
fbshipit-source-id: ff85f042896243abcd6ef37877834e26f36b6eb0
2020-03-27 23:03:05 +00:00
|
|
|
Random rnd(301);
|
|
|
|
|
|
|
|
|
|
DestroyAndReopen(options);
|
|
|
|
|
|
|
|
|
|
IOStatus error_msg = IOStatus::IOError("Retryable IO Error");
|
|
|
|
|
error_msg.SetRetryable(true);
|
|
|
|
|
|
|
|
|
|
// For the first batch, write is successful, require sync
|
|
|
|
|
{
|
|
|
|
|
WriteBatch batch;
|
|
|
|
|
|
|
|
|
|
for (auto i = 0; i < 100; ++i) {
|
2020-10-02 23:39:17 +00:00
|
|
|
ASSERT_OK(batch.Put(Key(i), rnd.RandomString(1024)));
|
Pass IOStatus to write path and set retryable IO Error as hard error in BG jobs (#6487)
Summary:
In the current code base, we use Status to get and store the returned status from the call. Specifically, for IO related functions, the current Status cannot reflect the IO Error details such as error scope, error retryable attribute, and others. With the implementation of https://github.com/facebook/rocksdb/issues/5761, we have the new Wrapper for IO, which returns IOStatus instead of Status. However, the IOStatus is purged at the lower level of write path and transferred to Status.
The first job of this PR is to pass the IOStatus to the write path (flush, WAL write, and Compaction). The second job is to identify the Retryable IO Error as HardError, and set the bg_error_ as HardError. In this case, the DB Instance becomes read only. User is informed of the Status and need to take actions to deal with it (e.g., call db->Resume()).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6487
Test Plan: Added the testing case to error_handler_fs_test. Pass make asan_check
Reviewed By: anand1976
Differential Revision: D20685017
Pulled By: zhichao-cao
fbshipit-source-id: ff85f042896243abcd6ef37877834e26f36b6eb0
2020-03-27 23:03:05 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
WriteOptions wopts;
|
|
|
|
|
wopts.sync = true;
|
2020-12-24 00:54:05 +00:00
|
|
|
ASSERT_OK(dbfull()->Write(wopts, &batch));
|
Pass IOStatus to write path and set retryable IO Error as hard error in BG jobs (#6487)
Summary:
In the current code base, we use Status to get and store the returned status from the call. Specifically, for IO related functions, the current Status cannot reflect the IO Error details such as error scope, error retryable attribute, and others. With the implementation of https://github.com/facebook/rocksdb/issues/5761, we have the new Wrapper for IO, which returns IOStatus instead of Status. However, the IOStatus is purged at the lower level of write path and transferred to Status.
The first job of this PR is to pass the IOStatus to the write path (flush, WAL write, and Compaction). The second job is to identify the Retryable IO Error as HardError, and set the bg_error_ as HardError. In this case, the DB Instance becomes read only. User is informed of the Status and need to take actions to deal with it (e.g., call db->Resume()).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6487
Test Plan: Added the testing case to error_handler_fs_test. Pass make asan_check
Reviewed By: anand1976
Differential Revision: D20685017
Pulled By: zhichao-cao
fbshipit-source-id: ff85f042896243abcd6ef37877834e26f36b6eb0
2020-03-27 23:03:05 +00:00
|
|
|
};
|
|
|
|
|
|
|
|
|
|
// For the second batch, the first 2 file Append are successful, then the
|
|
|
|
|
// following Append fails due to file system retryable IOError.
|
|
|
|
|
{
|
|
|
|
|
WriteBatch batch;
|
|
|
|
|
int write_error = 0;
|
|
|
|
|
|
|
|
|
|
for (auto i = 100; i < 200; ++i) {
|
2020-10-02 23:39:17 +00:00
|
|
|
ASSERT_OK(batch.Put(Key(i), rnd.RandomString(1024)));
|
Pass IOStatus to write path and set retryable IO Error as hard error in BG jobs (#6487)
Summary:
In the current code base, we use Status to get and store the returned status from the call. Specifically, for IO related functions, the current Status cannot reflect the IO Error details such as error scope, error retryable attribute, and others. With the implementation of https://github.com/facebook/rocksdb/issues/5761, we have the new Wrapper for IO, which returns IOStatus instead of Status. However, the IOStatus is purged at the lower level of write path and transferred to Status.
The first job of this PR is to pass the IOStatus to the write path (flush, WAL write, and Compaction). The second job is to identify the Retryable IO Error as HardError, and set the bg_error_ as HardError. In this case, the DB Instance becomes read only. User is informed of the Status and need to take actions to deal with it (e.g., call db->Resume()).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6487
Test Plan: Added the testing case to error_handler_fs_test. Pass make asan_check
Reviewed By: anand1976
Differential Revision: D20685017
Pulled By: zhichao-cao
fbshipit-source-id: ff85f042896243abcd6ef37877834e26f36b6eb0
2020-03-27 23:03:05 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
SyncPoint::GetInstance()->SetCallBack(
|
|
|
|
|
"WritableFileWriter::Append:BeforePrepareWrite", [&](void*) {
|
|
|
|
|
write_error++;
|
|
|
|
|
if (write_error > 2) {
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
fault_fs_->SetFilesystemActive(false, error_msg);
|
Pass IOStatus to write path and set retryable IO Error as hard error in BG jobs (#6487)
Summary:
In the current code base, we use Status to get and store the returned status from the call. Specifically, for IO related functions, the current Status cannot reflect the IO Error details such as error scope, error retryable attribute, and others. With the implementation of https://github.com/facebook/rocksdb/issues/5761, we have the new Wrapper for IO, which returns IOStatus instead of Status. However, the IOStatus is purged at the lower level of write path and transferred to Status.
The first job of this PR is to pass the IOStatus to the write path (flush, WAL write, and Compaction). The second job is to identify the Retryable IO Error as HardError, and set the bg_error_ as HardError. In this case, the DB Instance becomes read only. User is informed of the Status and need to take actions to deal with it (e.g., call db->Resume()).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6487
Test Plan: Added the testing case to error_handler_fs_test. Pass make asan_check
Reviewed By: anand1976
Differential Revision: D20685017
Pulled By: zhichao-cao
fbshipit-source-id: ff85f042896243abcd6ef37877834e26f36b6eb0
2020-03-27 23:03:05 +00:00
|
|
|
}
|
|
|
|
|
});
|
|
|
|
|
SyncPoint::GetInstance()->EnableProcessing();
|
|
|
|
|
WriteOptions wopts;
|
|
|
|
|
wopts.sync = true;
|
2020-12-24 00:54:05 +00:00
|
|
|
Status s = dbfull()->Write(wopts, &batch);
|
|
|
|
|
ASSERT_TRUE(s.IsIOError());
|
Pass IOStatus to write path and set retryable IO Error as hard error in BG jobs (#6487)
Summary:
In the current code base, we use Status to get and store the returned status from the call. Specifically, for IO related functions, the current Status cannot reflect the IO Error details such as error scope, error retryable attribute, and others. With the implementation of https://github.com/facebook/rocksdb/issues/5761, we have the new Wrapper for IO, which returns IOStatus instead of Status. However, the IOStatus is purged at the lower level of write path and transferred to Status.
The first job of this PR is to pass the IOStatus to the write path (flush, WAL write, and Compaction). The second job is to identify the Retryable IO Error as HardError, and set the bg_error_ as HardError. In this case, the DB Instance becomes read only. User is informed of the Status and need to take actions to deal with it (e.g., call db->Resume()).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6487
Test Plan: Added the testing case to error_handler_fs_test. Pass make asan_check
Reviewed By: anand1976
Differential Revision: D20685017
Pulled By: zhichao-cao
fbshipit-source-id: ff85f042896243abcd6ef37877834e26f36b6eb0
2020-03-27 23:03:05 +00:00
|
|
|
}
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
fault_fs_->SetFilesystemActive(true);
|
Pass IOStatus to write path and set retryable IO Error as hard error in BG jobs (#6487)
Summary:
In the current code base, we use Status to get and store the returned status from the call. Specifically, for IO related functions, the current Status cannot reflect the IO Error details such as error scope, error retryable attribute, and others. With the implementation of https://github.com/facebook/rocksdb/issues/5761, we have the new Wrapper for IO, which returns IOStatus instead of Status. However, the IOStatus is purged at the lower level of write path and transferred to Status.
The first job of this PR is to pass the IOStatus to the write path (flush, WAL write, and Compaction). The second job is to identify the Retryable IO Error as HardError, and set the bg_error_ as HardError. In this case, the DB Instance becomes read only. User is informed of the Status and need to take actions to deal with it (e.g., call db->Resume()).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6487
Test Plan: Added the testing case to error_handler_fs_test. Pass make asan_check
Reviewed By: anand1976
Differential Revision: D20685017
Pulled By: zhichao-cao
fbshipit-source-id: ff85f042896243abcd6ef37877834e26f36b6eb0
2020-03-27 23:03:05 +00:00
|
|
|
SyncPoint::GetInstance()->ClearAllCallBacks();
|
|
|
|
|
SyncPoint::GetInstance()->DisableProcessing();
|
|
|
|
|
|
|
|
|
|
// Data in corrupted WAL are not stored
|
|
|
|
|
for (auto i = 0; i < 199; ++i) {
|
|
|
|
|
if (i < 100) {
|
|
|
|
|
ASSERT_NE(Get(Key(i)), "NOT_FOUND");
|
|
|
|
|
} else {
|
|
|
|
|
ASSERT_EQ(Get(Key(i)), "NOT_FOUND");
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
// Resume and write a new batch, should be in the WAL
|
2020-12-24 00:54:05 +00:00
|
|
|
ASSERT_OK(dbfull()->Resume());
|
Pass IOStatus to write path and set retryable IO Error as hard error in BG jobs (#6487)
Summary:
In the current code base, we use Status to get and store the returned status from the call. Specifically, for IO related functions, the current Status cannot reflect the IO Error details such as error scope, error retryable attribute, and others. With the implementation of https://github.com/facebook/rocksdb/issues/5761, we have the new Wrapper for IO, which returns IOStatus instead of Status. However, the IOStatus is purged at the lower level of write path and transferred to Status.
The first job of this PR is to pass the IOStatus to the write path (flush, WAL write, and Compaction). The second job is to identify the Retryable IO Error as HardError, and set the bg_error_ as HardError. In this case, the DB Instance becomes read only. User is informed of the Status and need to take actions to deal with it (e.g., call db->Resume()).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6487
Test Plan: Added the testing case to error_handler_fs_test. Pass make asan_check
Reviewed By: anand1976
Differential Revision: D20685017
Pulled By: zhichao-cao
fbshipit-source-id: ff85f042896243abcd6ef37877834e26f36b6eb0
2020-03-27 23:03:05 +00:00
|
|
|
{
|
|
|
|
|
WriteBatch batch;
|
|
|
|
|
|
|
|
|
|
for (auto i = 200; i < 300; ++i) {
|
2020-10-02 23:39:17 +00:00
|
|
|
ASSERT_OK(batch.Put(Key(i), rnd.RandomString(1024)));
|
Pass IOStatus to write path and set retryable IO Error as hard error in BG jobs (#6487)
Summary:
In the current code base, we use Status to get and store the returned status from the call. Specifically, for IO related functions, the current Status cannot reflect the IO Error details such as error scope, error retryable attribute, and others. With the implementation of https://github.com/facebook/rocksdb/issues/5761, we have the new Wrapper for IO, which returns IOStatus instead of Status. However, the IOStatus is purged at the lower level of write path and transferred to Status.
The first job of this PR is to pass the IOStatus to the write path (flush, WAL write, and Compaction). The second job is to identify the Retryable IO Error as HardError, and set the bg_error_ as HardError. In this case, the DB Instance becomes read only. User is informed of the Status and need to take actions to deal with it (e.g., call db->Resume()).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6487
Test Plan: Added the testing case to error_handler_fs_test. Pass make asan_check
Reviewed By: anand1976
Differential Revision: D20685017
Pulled By: zhichao-cao
fbshipit-source-id: ff85f042896243abcd6ef37877834e26f36b6eb0
2020-03-27 23:03:05 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
WriteOptions wopts;
|
|
|
|
|
wopts.sync = true;
|
2020-12-24 00:54:05 +00:00
|
|
|
ASSERT_OK(dbfull()->Write(wopts, &batch));
|
Pass IOStatus to write path and set retryable IO Error as hard error in BG jobs (#6487)
Summary:
In the current code base, we use Status to get and store the returned status from the call. Specifically, for IO related functions, the current Status cannot reflect the IO Error details such as error scope, error retryable attribute, and others. With the implementation of https://github.com/facebook/rocksdb/issues/5761, we have the new Wrapper for IO, which returns IOStatus instead of Status. However, the IOStatus is purged at the lower level of write path and transferred to Status.
The first job of this PR is to pass the IOStatus to the write path (flush, WAL write, and Compaction). The second job is to identify the Retryable IO Error as HardError, and set the bg_error_ as HardError. In this case, the DB Instance becomes read only. User is informed of the Status and need to take actions to deal with it (e.g., call db->Resume()).
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6487
Test Plan: Added the testing case to error_handler_fs_test. Pass make asan_check
Reviewed By: anand1976
Differential Revision: D20685017
Pulled By: zhichao-cao
fbshipit-source-id: ff85f042896243abcd6ef37877834e26f36b6eb0
2020-03-27 23:03:05 +00:00
|
|
|
};
|
|
|
|
|
|
|
|
|
|
Reopen(options);
|
|
|
|
|
for (auto i = 0; i < 300; ++i) {
|
|
|
|
|
if (i < 100 || i >= 200) {
|
|
|
|
|
ASSERT_NE(Get(Key(i)), "NOT_FOUND");
|
|
|
|
|
} else {
|
|
|
|
|
ASSERT_EQ(Get(Key(i)), "NOT_FOUND");
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
Close();
|
|
|
|
|
}
|
|
|
|
|
|
2020-03-04 20:30:34 +00:00
|
|
|
TEST_F(DBErrorHandlingFSTest, MultiCFWALWriteError) {
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
if (mem_env_ != nullptr) {
|
|
|
|
|
ROCKSDB_GTEST_SKIP("Test requires non-mock environment");
|
|
|
|
|
return;
|
|
|
|
|
}
|
2023-10-31 23:13:36 +00:00
|
|
|
std::shared_ptr<ErrorHandlerFSListener> listener =
|
|
|
|
|
std::make_shared<ErrorHandlerFSListener>();
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
Options options = GetDefaultOptions();
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
options.env = fault_env_.get();
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
options.create_if_missing = true;
|
|
|
|
|
options.writable_file_max_buffer_size = 32768;
|
|
|
|
|
options.listeners.emplace_back(listener);
|
2018-09-17 20:08:13 +00:00
|
|
|
Random rnd(301);
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
|
|
|
|
|
listener->EnableAutoRecovery();
|
|
|
|
|
CreateAndReopenWithCF({"one", "two", "three"}, options);
|
|
|
|
|
|
|
|
|
|
{
|
|
|
|
|
WriteBatch batch;
|
|
|
|
|
|
|
|
|
|
for (auto i = 1; i < 4; ++i) {
|
|
|
|
|
for (auto j = 0; j < 100; ++j) {
|
2020-10-02 23:39:17 +00:00
|
|
|
ASSERT_OK(batch.Put(handles_[i], Key(j), rnd.RandomString(1024)));
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
WriteOptions wopts;
|
|
|
|
|
wopts.sync = true;
|
2020-12-24 00:54:05 +00:00
|
|
|
ASSERT_OK(dbfull()->Write(wopts, &batch));
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
};
|
|
|
|
|
|
|
|
|
|
{
|
|
|
|
|
WriteBatch batch;
|
|
|
|
|
int write_error = 0;
|
|
|
|
|
|
|
|
|
|
// Write to one CF
|
|
|
|
|
for (auto i = 100; i < 199; ++i) {
|
2020-10-02 23:39:17 +00:00
|
|
|
ASSERT_OK(batch.Put(handles_[2], Key(i), rnd.RandomString(1024)));
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
SyncPoint::GetInstance()->SetCallBack(
|
|
|
|
|
"WritableFileWriter::Append:BeforePrepareWrite", [&](void*) {
|
|
|
|
|
write_error++;
|
|
|
|
|
if (write_error > 2) {
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
fault_fs_->SetFilesystemActive(false,
|
|
|
|
|
IOStatus::NoSpace("Out of space"));
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
}
|
|
|
|
|
});
|
|
|
|
|
SyncPoint::GetInstance()->EnableProcessing();
|
|
|
|
|
WriteOptions wopts;
|
|
|
|
|
wopts.sync = true;
|
2020-12-24 00:54:05 +00:00
|
|
|
Status s = dbfull()->Write(wopts, &batch);
|
|
|
|
|
ASSERT_TRUE(s.IsNoSpace());
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
}
|
|
|
|
|
SyncPoint::GetInstance()->DisableProcessing();
|
2022-09-06 19:59:02 +00:00
|
|
|
// `ClearAllCallBacks()` is needed in addition to `DisableProcessing()` to
|
|
|
|
|
// drain all callbacks. Otherwise, a pending callback in the background
|
|
|
|
|
// could re-disable `fault_fs_` after we enable it below.
|
|
|
|
|
SyncPoint::GetInstance()->ClearAllCallBacks();
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
fault_fs_->SetFilesystemActive(true);
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
ASSERT_EQ(listener->WaitForRecovery(5000000), true);
|
|
|
|
|
|
|
|
|
|
for (auto i = 1; i < 4; ++i) {
|
|
|
|
|
// Every CF should have been flushed
|
|
|
|
|
ASSERT_EQ(NumTableFilesAtLevel(0, i), 1);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
for (auto i = 1; i < 4; ++i) {
|
|
|
|
|
for (auto j = 0; j < 199; ++j) {
|
|
|
|
|
if (j < 100) {
|
|
|
|
|
ASSERT_NE(Get(i, Key(j)), "NOT_FOUND");
|
|
|
|
|
} else {
|
|
|
|
|
ASSERT_EQ(Get(i, Key(j)), "NOT_FOUND");
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
ReopenWithColumnFamilies({"default", "one", "two", "three"}, options);
|
|
|
|
|
for (auto i = 1; i < 4; ++i) {
|
|
|
|
|
for (auto j = 0; j < 199; ++j) {
|
|
|
|
|
if (j < 100) {
|
|
|
|
|
ASSERT_NE(Get(i, Key(j)), "NOT_FOUND");
|
|
|
|
|
} else {
|
|
|
|
|
ASSERT_EQ(Get(i, Key(j)), "NOT_FOUND");
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
Close();
|
|
|
|
|
}
|
|
|
|
|
|
2020-03-04 20:30:34 +00:00
|
|
|
TEST_F(DBErrorHandlingFSTest, MultiDBCompactionError) {
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
if (mem_env_ != nullptr) {
|
|
|
|
|
ROCKSDB_GTEST_SKIP("Test requires non-mock environment");
|
|
|
|
|
return;
|
|
|
|
|
}
|
|
|
|
|
FaultInjectionTestEnv* def_env = new FaultInjectionTestEnv(env_);
|
2020-03-24 04:50:42 +00:00
|
|
|
std::vector<std::unique_ptr<Env>> fault_envs;
|
2020-03-04 20:30:34 +00:00
|
|
|
std::vector<FaultInjectionTestFS*> fault_fs;
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
std::vector<Options> options;
|
2020-03-04 20:30:34 +00:00
|
|
|
std::vector<std::shared_ptr<ErrorHandlerFSListener>> listener;
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
std::vector<DB*> db;
|
|
|
|
|
std::shared_ptr<SstFileManager> sfm(NewSstFileManager(def_env));
|
|
|
|
|
int kNumDbInstances = 3;
|
2018-09-17 20:08:13 +00:00
|
|
|
Random rnd(301);
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
|
|
|
|
|
for (auto i = 0; i < kNumDbInstances; ++i) {
|
2020-03-04 20:30:34 +00:00
|
|
|
listener.emplace_back(new ErrorHandlerFSListener());
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
options.emplace_back(GetDefaultOptions());
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
fault_fs.emplace_back(new FaultInjectionTestFS(env_->GetFileSystem()));
|
2020-03-24 04:50:42 +00:00
|
|
|
std::shared_ptr<FileSystem> fs(fault_fs.back());
|
|
|
|
|
fault_envs.emplace_back(new CompositeEnvWrapper(def_env, fs));
|
|
|
|
|
options[i].env = fault_envs.back().get();
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
options[i].create_if_missing = true;
|
|
|
|
|
options[i].level0_file_num_compaction_trigger = 2;
|
|
|
|
|
options[i].writable_file_max_buffer_size = 32768;
|
|
|
|
|
options[i].listeners.emplace_back(listener[i]);
|
|
|
|
|
options[i].sst_file_manager = sfm;
|
|
|
|
|
DB* dbptr;
|
|
|
|
|
char buf[16];
|
|
|
|
|
|
|
|
|
|
listener[i]->EnableAutoRecovery();
|
|
|
|
|
// Setup for returning error for the 3rd SST, which would be level 1
|
2020-03-04 20:30:34 +00:00
|
|
|
listener[i]->InjectFileCreationError(fault_fs[i], 3,
|
|
|
|
|
IOStatus::NoSpace("Out of space"));
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
snprintf(buf, sizeof(buf), "_%d", i);
|
2020-12-24 00:54:05 +00:00
|
|
|
ASSERT_OK(DestroyDB(dbname_ + std::string(buf), options[i]));
|
|
|
|
|
ASSERT_OK(DB::Open(options[i], dbname_ + std::string(buf), &dbptr));
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
db.emplace_back(dbptr);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
for (auto i = 0; i < kNumDbInstances; ++i) {
|
|
|
|
|
WriteBatch batch;
|
|
|
|
|
|
|
|
|
|
for (auto j = 0; j <= 100; ++j) {
|
2020-10-02 23:39:17 +00:00
|
|
|
ASSERT_OK(batch.Put(Key(j), rnd.RandomString(1024)));
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
WriteOptions wopts;
|
|
|
|
|
wopts.sync = true;
|
2020-12-24 00:54:05 +00:00
|
|
|
ASSERT_OK(db[i]->Write(wopts, &batch));
|
|
|
|
|
ASSERT_OK(db[i]->Flush(FlushOptions()));
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
def_env->SetFilesystemActive(false, Status::NoSpace("Out of space"));
|
|
|
|
|
for (auto i = 0; i < kNumDbInstances; ++i) {
|
|
|
|
|
WriteBatch batch;
|
|
|
|
|
|
|
|
|
|
// Write to one CF
|
|
|
|
|
for (auto j = 100; j < 199; ++j) {
|
2020-10-02 23:39:17 +00:00
|
|
|
ASSERT_OK(batch.Put(Key(j), rnd.RandomString(1024)));
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
WriteOptions wopts;
|
|
|
|
|
wopts.sync = true;
|
2020-12-24 00:54:05 +00:00
|
|
|
ASSERT_OK(db[i]->Write(wopts, &batch));
|
|
|
|
|
ASSERT_OK(db[i]->Flush(FlushOptions()));
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
for (auto i = 0; i < kNumDbInstances; ++i) {
|
Remove wait_unscheduled from waitForCompact internal API (#11443)
Summary:
Context:
In pull request https://github.com/facebook/rocksdb/issues/11436, we are introducing a new public API `waitForCompact(const WaitForCompactOptions& wait_for_compact_options)`. This API invokes the internal implementation `waitForCompact(bool wait_unscheduled=false)`. The unscheduled parameter indicates the compactions that are not yet scheduled but are required to process items in the queue.
In certain cases, we are unable to wait for compactions, such as during a shutdown or when background jobs are paused. It is important to return the appropriate status in these scenarios. For all other cases, we should wait for all compaction and flush jobs, including the unscheduled ones. The primary purpose of this new API is to wait until the system has resolved its compaction debt. Currently, the usage of `wait_unscheduled` is limited to test code.
This pull request eliminates the usage of wait_unscheduled. The internal `waitForCompact()` API now waits for unscheduled compactions unless the db is undergoing a shutdown. In the event of a shutdown, the API returns `Status::ShutdownInProgress()`.
Additionally, a new parameter, `abort_on_pause`, has been introduced with a default value of `false`. This parameter addresses the possibility of waiting indefinitely for unscheduled jobs if `PauseBackgroundWork()` was called before `waitForCompact()` is invoked. By setting `abort_on_pause` to `true`, the API will immediately return `Status::Aborted`.
Furthermore, all tests that previously called `waitForCompact(true)` have been fixed.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11443
Test Plan:
Existing tests that involve a shutdown in progress:
- DBCompactionTest::CompactRangeShutdownWhileDelayed
- DBTestWithParam::PreShutdownMultipleCompaction
- DBTestWithParam::PreShutdownCompactionMiddle
Reviewed By: pdillinger
Differential Revision: D45923426
Pulled By: jaykorean
fbshipit-source-id: 7dc93fe6a6841a7d9d2d72866fa647090dba8eae
2023-05-18 01:13:50 +00:00
|
|
|
Status s = static_cast<DBImpl*>(db[i])->TEST_WaitForCompact();
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
ASSERT_EQ(s.severity(), Status::Severity::kSoftError);
|
2020-03-04 20:30:34 +00:00
|
|
|
fault_fs[i]->SetFilesystemActive(true);
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
def_env->SetFilesystemActive(true);
|
|
|
|
|
for (auto i = 0; i < kNumDbInstances; ++i) {
|
|
|
|
|
std::string prop;
|
|
|
|
|
ASSERT_EQ(listener[i]->WaitForRecovery(5000000), true);
|
Remove wait_unscheduled from waitForCompact internal API (#11443)
Summary:
Context:
In pull request https://github.com/facebook/rocksdb/issues/11436, we are introducing a new public API `waitForCompact(const WaitForCompactOptions& wait_for_compact_options)`. This API invokes the internal implementation `waitForCompact(bool wait_unscheduled=false)`. The unscheduled parameter indicates the compactions that are not yet scheduled but are required to process items in the queue.
In certain cases, we are unable to wait for compactions, such as during a shutdown or when background jobs are paused. It is important to return the appropriate status in these scenarios. For all other cases, we should wait for all compaction and flush jobs, including the unscheduled ones. The primary purpose of this new API is to wait until the system has resolved its compaction debt. Currently, the usage of `wait_unscheduled` is limited to test code.
This pull request eliminates the usage of wait_unscheduled. The internal `waitForCompact()` API now waits for unscheduled compactions unless the db is undergoing a shutdown. In the event of a shutdown, the API returns `Status::ShutdownInProgress()`.
Additionally, a new parameter, `abort_on_pause`, has been introduced with a default value of `false`. This parameter addresses the possibility of waiting indefinitely for unscheduled jobs if `PauseBackgroundWork()` was called before `waitForCompact()` is invoked. By setting `abort_on_pause` to `true`, the API will immediately return `Status::Aborted`.
Furthermore, all tests that previously called `waitForCompact(true)` have been fixed.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11443
Test Plan:
Existing tests that involve a shutdown in progress:
- DBCompactionTest::CompactRangeShutdownWhileDelayed
- DBTestWithParam::PreShutdownMultipleCompaction
- DBTestWithParam::PreShutdownCompactionMiddle
Reviewed By: pdillinger
Differential Revision: D45923426
Pulled By: jaykorean
fbshipit-source-id: 7dc93fe6a6841a7d9d2d72866fa647090dba8eae
2023-05-18 01:13:50 +00:00
|
|
|
ASSERT_OK(static_cast<DBImpl*>(db[i])->TEST_WaitForCompact());
|
2022-05-06 20:03:58 +00:00
|
|
|
EXPECT_TRUE(db[i]->GetProperty(
|
|
|
|
|
"rocksdb.num-files-at-level" + std::to_string(0), &prop));
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
EXPECT_EQ(atoi(prop.c_str()), 0);
|
2022-05-06 20:03:58 +00:00
|
|
|
EXPECT_TRUE(db[i]->GetProperty(
|
|
|
|
|
"rocksdb.num-files-at-level" + std::to_string(1), &prop));
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
EXPECT_EQ(atoi(prop.c_str()), 1);
|
|
|
|
|
}
|
|
|
|
|
|
2021-01-26 18:58:31 +00:00
|
|
|
SstFileManagerImpl* sfmImpl =
|
|
|
|
|
static_cast_with_check<SstFileManagerImpl>(sfm.get());
|
|
|
|
|
sfmImpl->Close();
|
|
|
|
|
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
for (auto i = 0; i < kNumDbInstances; ++i) {
|
|
|
|
|
char buf[16];
|
|
|
|
|
snprintf(buf, sizeof(buf), "_%d", i);
|
|
|
|
|
delete db[i];
|
2020-03-04 20:30:34 +00:00
|
|
|
fault_fs[i]->SetFilesystemActive(true);
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
if (getenv("KEEP_DB")) {
|
|
|
|
|
printf("DB is still at %s%s\n", dbname_.c_str(), buf);
|
|
|
|
|
} else {
|
2020-12-24 00:54:05 +00:00
|
|
|
ASSERT_OK(DestroyDB(dbname_ + std::string(buf), options[i]));
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
options.clear();
|
|
|
|
|
sfm.reset();
|
|
|
|
|
delete def_env;
|
|
|
|
|
}
|
|
|
|
|
|
2020-03-04 20:30:34 +00:00
|
|
|
TEST_F(DBErrorHandlingFSTest, MultiDBVariousErrors) {
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
if (mem_env_ != nullptr) {
|
|
|
|
|
ROCKSDB_GTEST_SKIP("Test requires non-mock environment");
|
|
|
|
|
return;
|
|
|
|
|
}
|
|
|
|
|
FaultInjectionTestEnv* def_env = new FaultInjectionTestEnv(env_);
|
2020-03-24 04:50:42 +00:00
|
|
|
std::vector<std::unique_ptr<Env>> fault_envs;
|
2020-03-04 20:30:34 +00:00
|
|
|
std::vector<FaultInjectionTestFS*> fault_fs;
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
std::vector<Options> options;
|
2020-03-04 20:30:34 +00:00
|
|
|
std::vector<std::shared_ptr<ErrorHandlerFSListener>> listener;
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
std::vector<DB*> db;
|
|
|
|
|
std::shared_ptr<SstFileManager> sfm(NewSstFileManager(def_env));
|
|
|
|
|
int kNumDbInstances = 3;
|
2018-09-17 20:08:13 +00:00
|
|
|
Random rnd(301);
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
|
|
|
|
|
for (auto i = 0; i < kNumDbInstances; ++i) {
|
2020-03-04 20:30:34 +00:00
|
|
|
listener.emplace_back(new ErrorHandlerFSListener());
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
options.emplace_back(GetDefaultOptions());
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
fault_fs.emplace_back(new FaultInjectionTestFS(env_->GetFileSystem()));
|
2020-03-24 04:50:42 +00:00
|
|
|
std::shared_ptr<FileSystem> fs(fault_fs.back());
|
|
|
|
|
fault_envs.emplace_back(new CompositeEnvWrapper(def_env, fs));
|
|
|
|
|
options[i].env = fault_envs.back().get();
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
options[i].create_if_missing = true;
|
|
|
|
|
options[i].level0_file_num_compaction_trigger = 2;
|
|
|
|
|
options[i].writable_file_max_buffer_size = 32768;
|
|
|
|
|
options[i].listeners.emplace_back(listener[i]);
|
|
|
|
|
options[i].sst_file_manager = sfm;
|
|
|
|
|
DB* dbptr;
|
|
|
|
|
char buf[16];
|
|
|
|
|
|
|
|
|
|
listener[i]->EnableAutoRecovery();
|
|
|
|
|
switch (i) {
|
|
|
|
|
case 0:
|
|
|
|
|
// Setup for returning error for the 3rd SST, which would be level 1
|
2020-03-04 20:30:34 +00:00
|
|
|
listener[i]->InjectFileCreationError(fault_fs[i], 3,
|
|
|
|
|
IOStatus::NoSpace("Out of space"));
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
break;
|
|
|
|
|
case 1:
|
|
|
|
|
// Setup for returning error after the 1st SST, which would result
|
|
|
|
|
// in a hard error
|
2020-03-04 20:30:34 +00:00
|
|
|
listener[i]->InjectFileCreationError(fault_fs[i], 2,
|
|
|
|
|
IOStatus::NoSpace("Out of space"));
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
break;
|
|
|
|
|
default:
|
|
|
|
|
break;
|
|
|
|
|
}
|
|
|
|
|
snprintf(buf, sizeof(buf), "_%d", i);
|
2020-12-24 00:54:05 +00:00
|
|
|
ASSERT_OK(DestroyDB(dbname_ + std::string(buf), options[i]));
|
|
|
|
|
ASSERT_OK(DB::Open(options[i], dbname_ + std::string(buf), &dbptr));
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
db.emplace_back(dbptr);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
for (auto i = 0; i < kNumDbInstances; ++i) {
|
|
|
|
|
WriteBatch batch;
|
|
|
|
|
|
|
|
|
|
for (auto j = 0; j <= 100; ++j) {
|
2020-10-02 23:39:17 +00:00
|
|
|
ASSERT_OK(batch.Put(Key(j), rnd.RandomString(1024)));
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
WriteOptions wopts;
|
|
|
|
|
wopts.sync = true;
|
2020-12-24 00:54:05 +00:00
|
|
|
ASSERT_OK(db[i]->Write(wopts, &batch));
|
|
|
|
|
ASSERT_OK(db[i]->Flush(FlushOptions()));
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
def_env->SetFilesystemActive(false, Status::NoSpace("Out of space"));
|
|
|
|
|
for (auto i = 0; i < kNumDbInstances; ++i) {
|
|
|
|
|
WriteBatch batch;
|
|
|
|
|
|
|
|
|
|
// Write to one CF
|
|
|
|
|
for (auto j = 100; j < 199; ++j) {
|
2020-10-02 23:39:17 +00:00
|
|
|
ASSERT_OK(batch.Put(Key(j), rnd.RandomString(1024)));
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
WriteOptions wopts;
|
|
|
|
|
wopts.sync = true;
|
2020-12-24 00:54:05 +00:00
|
|
|
ASSERT_OK(db[i]->Write(wopts, &batch));
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
if (i != 1) {
|
2020-12-24 00:54:05 +00:00
|
|
|
ASSERT_OK(db[i]->Flush(FlushOptions()));
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
} else {
|
2020-12-24 00:54:05 +00:00
|
|
|
ASSERT_TRUE(db[i]->Flush(FlushOptions()).IsNoSpace());
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
for (auto i = 0; i < kNumDbInstances; ++i) {
|
Remove wait_unscheduled from waitForCompact internal API (#11443)
Summary:
Context:
In pull request https://github.com/facebook/rocksdb/issues/11436, we are introducing a new public API `waitForCompact(const WaitForCompactOptions& wait_for_compact_options)`. This API invokes the internal implementation `waitForCompact(bool wait_unscheduled=false)`. The unscheduled parameter indicates the compactions that are not yet scheduled but are required to process items in the queue.
In certain cases, we are unable to wait for compactions, such as during a shutdown or when background jobs are paused. It is important to return the appropriate status in these scenarios. For all other cases, we should wait for all compaction and flush jobs, including the unscheduled ones. The primary purpose of this new API is to wait until the system has resolved its compaction debt. Currently, the usage of `wait_unscheduled` is limited to test code.
This pull request eliminates the usage of wait_unscheduled. The internal `waitForCompact()` API now waits for unscheduled compactions unless the db is undergoing a shutdown. In the event of a shutdown, the API returns `Status::ShutdownInProgress()`.
Additionally, a new parameter, `abort_on_pause`, has been introduced with a default value of `false`. This parameter addresses the possibility of waiting indefinitely for unscheduled jobs if `PauseBackgroundWork()` was called before `waitForCompact()` is invoked. By setting `abort_on_pause` to `true`, the API will immediately return `Status::Aborted`.
Furthermore, all tests that previously called `waitForCompact(true)` have been fixed.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11443
Test Plan:
Existing tests that involve a shutdown in progress:
- DBCompactionTest::CompactRangeShutdownWhileDelayed
- DBTestWithParam::PreShutdownMultipleCompaction
- DBTestWithParam::PreShutdownCompactionMiddle
Reviewed By: pdillinger
Differential Revision: D45923426
Pulled By: jaykorean
fbshipit-source-id: 7dc93fe6a6841a7d9d2d72866fa647090dba8eae
2023-05-18 01:13:50 +00:00
|
|
|
Status s = static_cast<DBImpl*>(db[i])->TEST_WaitForCompact();
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
switch (i) {
|
|
|
|
|
case 0:
|
|
|
|
|
ASSERT_EQ(s.severity(), Status::Severity::kSoftError);
|
|
|
|
|
break;
|
|
|
|
|
case 1:
|
|
|
|
|
ASSERT_EQ(s.severity(), Status::Severity::kHardError);
|
|
|
|
|
break;
|
|
|
|
|
case 2:
|
2020-12-24 00:54:05 +00:00
|
|
|
ASSERT_OK(s);
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
break;
|
|
|
|
|
}
|
2020-03-04 20:30:34 +00:00
|
|
|
fault_fs[i]->SetFilesystemActive(true);
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
def_env->SetFilesystemActive(true);
|
|
|
|
|
for (auto i = 0; i < kNumDbInstances; ++i) {
|
|
|
|
|
std::string prop;
|
|
|
|
|
if (i < 2) {
|
|
|
|
|
ASSERT_EQ(listener[i]->WaitForRecovery(5000000), true);
|
|
|
|
|
}
|
|
|
|
|
if (i == 1) {
|
Remove wait_unscheduled from waitForCompact internal API (#11443)
Summary:
Context:
In pull request https://github.com/facebook/rocksdb/issues/11436, we are introducing a new public API `waitForCompact(const WaitForCompactOptions& wait_for_compact_options)`. This API invokes the internal implementation `waitForCompact(bool wait_unscheduled=false)`. The unscheduled parameter indicates the compactions that are not yet scheduled but are required to process items in the queue.
In certain cases, we are unable to wait for compactions, such as during a shutdown or when background jobs are paused. It is important to return the appropriate status in these scenarios. For all other cases, we should wait for all compaction and flush jobs, including the unscheduled ones. The primary purpose of this new API is to wait until the system has resolved its compaction debt. Currently, the usage of `wait_unscheduled` is limited to test code.
This pull request eliminates the usage of wait_unscheduled. The internal `waitForCompact()` API now waits for unscheduled compactions unless the db is undergoing a shutdown. In the event of a shutdown, the API returns `Status::ShutdownInProgress()`.
Additionally, a new parameter, `abort_on_pause`, has been introduced with a default value of `false`. This parameter addresses the possibility of waiting indefinitely for unscheduled jobs if `PauseBackgroundWork()` was called before `waitForCompact()` is invoked. By setting `abort_on_pause` to `true`, the API will immediately return `Status::Aborted`.
Furthermore, all tests that previously called `waitForCompact(true)` have been fixed.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/11443
Test Plan:
Existing tests that involve a shutdown in progress:
- DBCompactionTest::CompactRangeShutdownWhileDelayed
- DBTestWithParam::PreShutdownMultipleCompaction
- DBTestWithParam::PreShutdownCompactionMiddle
Reviewed By: pdillinger
Differential Revision: D45923426
Pulled By: jaykorean
fbshipit-source-id: 7dc93fe6a6841a7d9d2d72866fa647090dba8eae
2023-05-18 01:13:50 +00:00
|
|
|
ASSERT_OK(static_cast<DBImpl*>(db[i])->TEST_WaitForCompact());
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
}
|
2022-05-06 20:03:58 +00:00
|
|
|
EXPECT_TRUE(db[i]->GetProperty(
|
|
|
|
|
"rocksdb.num-files-at-level" + std::to_string(0), &prop));
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
EXPECT_EQ(atoi(prop.c_str()), 0);
|
2022-05-06 20:03:58 +00:00
|
|
|
EXPECT_TRUE(db[i]->GetProperty(
|
|
|
|
|
"rocksdb.num-files-at-level" + std::to_string(1), &prop));
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
EXPECT_EQ(atoi(prop.c_str()), 1);
|
|
|
|
|
}
|
|
|
|
|
|
2021-01-26 18:58:31 +00:00
|
|
|
SstFileManagerImpl* sfmImpl =
|
|
|
|
|
static_cast_with_check<SstFileManagerImpl>(sfm.get());
|
|
|
|
|
sfmImpl->Close();
|
|
|
|
|
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
for (auto i = 0; i < kNumDbInstances; ++i) {
|
|
|
|
|
char buf[16];
|
|
|
|
|
snprintf(buf, sizeof(buf), "_%d", i);
|
2020-03-04 20:30:34 +00:00
|
|
|
fault_fs[i]->SetFilesystemActive(true);
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
delete db[i];
|
|
|
|
|
if (getenv("KEEP_DB")) {
|
|
|
|
|
printf("DB is still at %s%s\n", dbname_.c_str(), buf);
|
|
|
|
|
} else {
|
2020-12-24 00:54:05 +00:00
|
|
|
EXPECT_OK(DestroyDB(dbname_ + std::string(buf), options[i]));
|
Auto recovery from out of space errors (#4164)
Summary:
This commit implements automatic recovery from a Status::NoSpace() error
during background operations such as write callback, flush and
compaction. The broad design is as follows -
1. Compaction errors are treated as soft errors and don't put the
database in read-only mode. A compaction is delayed until enough free
disk space is available to accomodate the compaction outputs, which is
estimated based on the input size. This means that users can continue to
write, and we rely on the WriteController to delay or stop writes if the
compaction debt becomes too high due to persistent low disk space
condition
2. Errors during write callback and flush are treated as hard errors,
i.e the database is put in read-only mode and goes back to read-write
only fater certain recovery actions are taken.
3. Both types of recovery rely on the SstFileManagerImpl to poll for
sufficient disk space. We assume that there is a 1-1 mapping between an
SFM and the underlying OS storage container. For cases where multiple
DBs are hosted on a single storage container, the user is expected to
allocate a single SFM instance and use the same one for all the DBs. If
no SFM is specified by the user, DBImpl::Open() will allocate one, but
this will be one per DB and each DB will recover independently. The
recovery implemented by SFM is as follows -
a) On the first occurance of an out of space error during compaction,
subsequent
compactions will be delayed until the disk free space check indicates
enough available space. The required space is computed as the sum of
input sizes.
b) The free space check requirement will be removed once the amount of
free space is greater than the size reserved by in progress
compactions when the first error occured
c) If the out of space error is a hard error, a background thread in
SFM will poll for sufficient headroom before triggering the recovery
of the database and putting it in write-only mode. The headroom is
calculated as the sum of the write_buffer_size of all the DB instances
associated with the SFM
4. EventListener callbacks will be called at the start and completion of
automatic recovery. Users can disable the auto recov ery in the start
callback, and later initiate it manually by calling DB::Resume()
Todo:
1. More extensive testing
2. Add disk full condition to db_stress (follow-on PR)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164
Differential Revision: D9846378
Pulled By: anand1976
fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
2018-09-15 20:36:19 +00:00
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
options.clear();
|
|
|
|
|
delete def_env;
|
|
|
|
|
}
|
2020-07-15 18:02:44 +00:00
|
|
|
|
2020-09-18 03:22:35 +00:00
|
|
|
// When Put the KV-pair, the write option is set to disable WAL.
|
|
|
|
|
// If retryable error happens in this condition, map the bg error
|
|
|
|
|
// to soft error and trigger auto resume. During auto resume, SwitchMemtable
|
|
|
|
|
// is disabled to avoid small SST tables. Write can still be applied before
|
|
|
|
|
// the bg error is cleaned unless the memtable is full.
|
2023-10-31 23:13:36 +00:00
|
|
|
TEST_F(DBErrorHandlingFSTest, FlushWritNoWALRetryableErrorAutoRecover1) {
|
2020-09-18 03:22:35 +00:00
|
|
|
// Activate the FS before the first resume
|
2023-10-31 23:13:36 +00:00
|
|
|
std::shared_ptr<ErrorHandlerFSListener> listener =
|
|
|
|
|
std::make_shared<ErrorHandlerFSListener>();
|
2020-09-18 03:22:35 +00:00
|
|
|
Options options = GetDefaultOptions();
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
options.env = fault_env_.get();
|
2020-09-18 03:22:35 +00:00
|
|
|
options.create_if_missing = true;
|
|
|
|
|
options.listeners.emplace_back(listener);
|
|
|
|
|
options.max_bgerror_resume_count = 2;
|
|
|
|
|
options.bgerror_resume_retry_interval = 100000; // 0.1 second
|
2021-03-18 05:36:42 +00:00
|
|
|
options.statistics = CreateDBStatistics();
|
2020-09-18 03:22:35 +00:00
|
|
|
Status s;
|
|
|
|
|
|
|
|
|
|
listener->EnableAutoRecovery(false);
|
|
|
|
|
DestroyAndReopen(options);
|
|
|
|
|
|
|
|
|
|
IOStatus error_msg = IOStatus::IOError("Retryable IO Error");
|
|
|
|
|
error_msg.SetRetryable(true);
|
|
|
|
|
|
|
|
|
|
WriteOptions wo = WriteOptions();
|
|
|
|
|
wo.disableWAL = true;
|
2020-10-02 23:39:17 +00:00
|
|
|
ASSERT_OK(Put(Key(1), "val1", wo));
|
2020-09-18 03:22:35 +00:00
|
|
|
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->LoadDependency(
|
|
|
|
|
{{"RecoverFromRetryableBGIOError:LoopOut",
|
2023-10-31 23:13:36 +00:00
|
|
|
"FlushWritNoWALRetryableeErrorAutoRecover1:1"}});
|
2020-09-18 03:22:35 +00:00
|
|
|
SyncPoint::GetInstance()->SetCallBack(
|
|
|
|
|
"BuildTable:BeforeFinishBuildTable",
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
[&](void*) { fault_fs_->SetFilesystemActive(false, error_msg); });
|
2020-09-18 03:22:35 +00:00
|
|
|
|
|
|
|
|
SyncPoint::GetInstance()->EnableProcessing();
|
|
|
|
|
s = Flush();
|
|
|
|
|
ASSERT_EQ("val1", Get(Key(1)));
|
|
|
|
|
ASSERT_EQ(s.severity(), ROCKSDB_NAMESPACE::Status::Severity::kSoftError);
|
2023-10-31 23:13:36 +00:00
|
|
|
TEST_SYNC_POINT("FlushWritNoWALRetryableeErrorAutoRecover1:1");
|
2020-09-18 03:22:35 +00:00
|
|
|
ASSERT_EQ("val1", Get(Key(1)));
|
|
|
|
|
ASSERT_EQ("val1", Get(Key(1)));
|
|
|
|
|
SyncPoint::GetInstance()->DisableProcessing();
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
fault_fs_->SetFilesystemActive(true);
|
2021-03-18 05:36:42 +00:00
|
|
|
ASSERT_EQ(3, options.statistics->getAndResetTickerCount(
|
|
|
|
|
ERROR_HANDLER_BG_ERROR_COUNT));
|
|
|
|
|
ASSERT_EQ(3, options.statistics->getAndResetTickerCount(
|
|
|
|
|
ERROR_HANDLER_BG_IO_ERROR_COUNT));
|
|
|
|
|
ASSERT_EQ(3, options.statistics->getAndResetTickerCount(
|
|
|
|
|
ERROR_HANDLER_BG_RETRYABLE_IO_ERROR_COUNT));
|
|
|
|
|
ASSERT_EQ(1, options.statistics->getAndResetTickerCount(
|
|
|
|
|
ERROR_HANDLER_AUTORESUME_COUNT));
|
2021-03-31 04:43:47 +00:00
|
|
|
ASSERT_LE(0, options.statistics->getAndResetTickerCount(
|
2021-03-18 05:36:42 +00:00
|
|
|
ERROR_HANDLER_AUTORESUME_RETRY_TOTAL_COUNT));
|
2021-03-31 04:43:47 +00:00
|
|
|
ASSERT_LE(0, options.statistics->getAndResetTickerCount(
|
2021-03-18 05:36:42 +00:00
|
|
|
ERROR_HANDLER_AUTORESUME_SUCCESS_COUNT));
|
|
|
|
|
HistogramData autoresume_retry;
|
|
|
|
|
options.statistics->histogramData(ERROR_HANDLER_AUTORESUME_RETRY_COUNT,
|
|
|
|
|
&autoresume_retry);
|
2021-03-31 04:43:47 +00:00
|
|
|
ASSERT_GE(autoresume_retry.max, 0);
|
2020-10-02 23:39:17 +00:00
|
|
|
ASSERT_OK(Put(Key(2), "val2", wo));
|
2020-09-18 03:22:35 +00:00
|
|
|
s = Flush();
|
|
|
|
|
// Since auto resume fails, the bg error is not cleand, flush will
|
|
|
|
|
// return the bg_error set before.
|
|
|
|
|
ASSERT_EQ(s.severity(), ROCKSDB_NAMESPACE::Status::Severity::kSoftError);
|
|
|
|
|
ASSERT_EQ("val2", Get(Key(2)));
|
|
|
|
|
|
|
|
|
|
// call auto resume
|
2020-12-24 00:54:05 +00:00
|
|
|
ASSERT_OK(dbfull()->Resume());
|
2020-10-02 23:39:17 +00:00
|
|
|
ASSERT_OK(Put(Key(3), "val3", wo));
|
2020-09-18 03:22:35 +00:00
|
|
|
// After resume is successful, the flush should be ok.
|
2020-12-24 00:54:05 +00:00
|
|
|
ASSERT_OK(Flush());
|
2020-09-18 03:22:35 +00:00
|
|
|
ASSERT_EQ("val3", Get(Key(3)));
|
|
|
|
|
Destroy(options);
|
|
|
|
|
}
|
|
|
|
|
|
2023-10-31 23:13:36 +00:00
|
|
|
TEST_F(DBErrorHandlingFSTest, MultipleRecoveryThreads) {
|
|
|
|
|
// This test creates a scenario where second write's recovery can get started
|
|
|
|
|
// while mutex is released for a short period during
|
|
|
|
|
// NotifyOnErrorRecoveryEnd() from the first write's recovery. This is to make
|
|
|
|
|
// sure RecoverFromRetryableBGIOError() from the second write's recovery
|
|
|
|
|
// thread does not start with recovery_in_prog_ = false;
|
|
|
|
|
|
|
|
|
|
std::shared_ptr<ErrorHandlerFSListener> listener =
|
|
|
|
|
std::make_shared<ErrorHandlerFSListener>();
|
|
|
|
|
Options options = GetDefaultOptions();
|
|
|
|
|
options.env = fault_env_.get();
|
|
|
|
|
options.create_if_missing = true;
|
|
|
|
|
options.listeners.emplace_back(listener);
|
|
|
|
|
options.max_bgerror_resume_count = 100;
|
|
|
|
|
options.bgerror_resume_retry_interval = 1000000; // 1 second
|
|
|
|
|
options.statistics = CreateDBStatistics();
|
|
|
|
|
|
|
|
|
|
listener->EnableAutoRecovery(false);
|
|
|
|
|
DestroyAndReopen(options);
|
|
|
|
|
|
|
|
|
|
IOStatus error_msg = IOStatus::IOError("Retryable IO Error");
|
|
|
|
|
error_msg.SetRetryable(true);
|
|
|
|
|
|
|
|
|
|
WriteOptions wo = WriteOptions();
|
|
|
|
|
wo.disableWAL = true;
|
|
|
|
|
fault_fs_->SetFilesystemActive(false, error_msg);
|
|
|
|
|
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->LoadDependency(
|
|
|
|
|
{{"NotifyOnErrorRecoveryEnd:MutexUnlocked:1",
|
|
|
|
|
"MultipleRecoveryThreads:1"},
|
|
|
|
|
{"MultipleRecoveryThreads:2",
|
|
|
|
|
"NotifyOnErrorRecoveryEnd:MutexUnlocked:2"},
|
|
|
|
|
{"StartRecoverFromRetryableBGIOError:BeforeWaitingForOtherThread",
|
|
|
|
|
"MultipleRecoveryThreads:3"},
|
|
|
|
|
{"RecoverFromRetryableBGIOError:RecoverSuccess",
|
|
|
|
|
"MultipleRecoveryThreads:4"},
|
|
|
|
|
{"MultipleRecoveryThreads:4",
|
|
|
|
|
"StartRecoverFromRetryableBGIOError:AfterWaitingForOtherThread"}});
|
|
|
|
|
SyncPoint::GetInstance()->EnableProcessing();
|
|
|
|
|
|
|
|
|
|
// First write with read fault injected and recovery will start
|
|
|
|
|
{
|
|
|
|
|
ASSERT_OK(Put(Key(1), "val1", wo));
|
|
|
|
|
Status s = Flush();
|
|
|
|
|
ASSERT_NOK(s);
|
|
|
|
|
}
|
|
|
|
|
// Remove read fault injection so that first recovery can go through
|
|
|
|
|
fault_fs_->SetFilesystemActive(true);
|
|
|
|
|
|
|
|
|
|
// At this point, first recovery is now at NotifyOnErrorRecoveryEnd. Mutex is
|
|
|
|
|
// released.
|
|
|
|
|
TEST_SYNC_POINT("MultipleRecoveryThreads:1");
|
|
|
|
|
|
|
|
|
|
ROCKSDB_NAMESPACE::port::Thread second_write([&] {
|
|
|
|
|
// Second write with read fault injected
|
|
|
|
|
fault_fs_->SetFilesystemActive(false, error_msg);
|
|
|
|
|
ASSERT_OK(Put(Key(2), "val2", wo));
|
|
|
|
|
Status s = Flush();
|
|
|
|
|
ASSERT_NOK(s);
|
|
|
|
|
});
|
|
|
|
|
// Second bg thread before waiting for the first thread's recovery thread
|
|
|
|
|
TEST_SYNC_POINT("MultipleRecoveryThreads:3");
|
|
|
|
|
// First thread's recovery thread continues
|
|
|
|
|
TEST_SYNC_POINT("MultipleRecoveryThreads:2");
|
|
|
|
|
// Wait for the first thread's recovery to finish
|
|
|
|
|
// (this sets recovery_in_prog_ = false)
|
|
|
|
|
// And second thread continues and starts recovery thread
|
|
|
|
|
TEST_SYNC_POINT("MultipleRecoveryThreads:4");
|
|
|
|
|
second_write.join();
|
|
|
|
|
// Remove error injection so that second thread recovery can go through
|
|
|
|
|
fault_fs_->SetFilesystemActive(true);
|
|
|
|
|
|
|
|
|
|
// Set up sync point so that we can wait for the recovery thread to finish
|
|
|
|
|
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->LoadDependency(
|
|
|
|
|
{{"RecoverFromRetryableBGIOError:RecoverSuccess",
|
|
|
|
|
"MultipleRecoveryThreads:6"}});
|
|
|
|
|
|
|
|
|
|
// Wait for the second thread's recovery to be done
|
|
|
|
|
TEST_SYNC_POINT("MultipleRecoveryThreads:6");
|
|
|
|
|
|
|
|
|
|
SyncPoint::GetInstance()->DisableProcessing();
|
|
|
|
|
SyncPoint::GetInstance()->ClearAllCallBacks();
|
|
|
|
|
Destroy(options);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
TEST_F(DBErrorHandlingFSTest, FlushWritNoWALRetryableErrorAutoRecover2) {
|
2020-09-18 03:22:35 +00:00
|
|
|
// Activate the FS before the first resume
|
2023-10-31 23:13:36 +00:00
|
|
|
std::shared_ptr<ErrorHandlerFSListener> listener =
|
|
|
|
|
std::make_shared<ErrorHandlerFSListener>();
|
2020-09-18 03:22:35 +00:00
|
|
|
Options options = GetDefaultOptions();
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
options.env = fault_env_.get();
|
2020-09-18 03:22:35 +00:00
|
|
|
options.create_if_missing = true;
|
|
|
|
|
options.listeners.emplace_back(listener);
|
|
|
|
|
options.max_bgerror_resume_count = 2;
|
|
|
|
|
options.bgerror_resume_retry_interval = 100000; // 0.1 second
|
2021-03-18 05:36:42 +00:00
|
|
|
options.statistics = CreateDBStatistics();
|
2020-09-18 03:22:35 +00:00
|
|
|
Status s;
|
|
|
|
|
|
|
|
|
|
listener->EnableAutoRecovery(false);
|
|
|
|
|
DestroyAndReopen(options);
|
|
|
|
|
|
|
|
|
|
IOStatus error_msg = IOStatus::IOError("Retryable IO Error");
|
|
|
|
|
error_msg.SetRetryable(true);
|
|
|
|
|
|
|
|
|
|
WriteOptions wo = WriteOptions();
|
|
|
|
|
wo.disableWAL = true;
|
2020-10-02 23:39:17 +00:00
|
|
|
ASSERT_OK(Put(Key(1), "val1", wo));
|
2020-09-18 03:22:35 +00:00
|
|
|
SyncPoint::GetInstance()->SetCallBack(
|
|
|
|
|
"BuildTable:BeforeFinishBuildTable",
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
[&](void*) { fault_fs_->SetFilesystemActive(false, error_msg); });
|
2020-09-18 03:22:35 +00:00
|
|
|
|
|
|
|
|
SyncPoint::GetInstance()->EnableProcessing();
|
|
|
|
|
s = Flush();
|
|
|
|
|
ASSERT_EQ("val1", Get(Key(1)));
|
|
|
|
|
ASSERT_EQ(s.severity(), ROCKSDB_NAMESPACE::Status::Severity::kSoftError);
|
|
|
|
|
SyncPoint::GetInstance()->DisableProcessing();
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
fault_fs_->SetFilesystemActive(true);
|
2020-09-18 03:22:35 +00:00
|
|
|
ASSERT_EQ(listener->WaitForRecovery(5000000), true);
|
|
|
|
|
ASSERT_EQ("val1", Get(Key(1)));
|
2021-03-18 05:36:42 +00:00
|
|
|
ASSERT_EQ(1, options.statistics->getAndResetTickerCount(
|
|
|
|
|
ERROR_HANDLER_BG_ERROR_COUNT));
|
|
|
|
|
ASSERT_EQ(1, options.statistics->getAndResetTickerCount(
|
|
|
|
|
ERROR_HANDLER_BG_IO_ERROR_COUNT));
|
|
|
|
|
ASSERT_EQ(1, options.statistics->getAndResetTickerCount(
|
|
|
|
|
ERROR_HANDLER_BG_RETRYABLE_IO_ERROR_COUNT));
|
|
|
|
|
ASSERT_EQ(1, options.statistics->getAndResetTickerCount(
|
|
|
|
|
ERROR_HANDLER_AUTORESUME_COUNT));
|
2021-03-26 04:41:44 +00:00
|
|
|
ASSERT_LE(0, options.statistics->getAndResetTickerCount(
|
2021-03-18 05:36:42 +00:00
|
|
|
ERROR_HANDLER_AUTORESUME_RETRY_TOTAL_COUNT));
|
2021-03-26 04:41:44 +00:00
|
|
|
ASSERT_LE(0, options.statistics->getAndResetTickerCount(
|
2021-03-18 05:36:42 +00:00
|
|
|
ERROR_HANDLER_AUTORESUME_SUCCESS_COUNT));
|
|
|
|
|
HistogramData autoresume_retry;
|
|
|
|
|
options.statistics->histogramData(ERROR_HANDLER_AUTORESUME_RETRY_COUNT,
|
|
|
|
|
&autoresume_retry);
|
2021-03-26 04:41:44 +00:00
|
|
|
ASSERT_GE(autoresume_retry.max, 0);
|
2020-10-02 23:39:17 +00:00
|
|
|
ASSERT_OK(Put(Key(2), "val2", wo));
|
2020-09-18 03:22:35 +00:00
|
|
|
s = Flush();
|
|
|
|
|
// Since auto resume is successful, the bg error is cleaned, flush will
|
|
|
|
|
// be successful.
|
|
|
|
|
ASSERT_OK(s);
|
|
|
|
|
ASSERT_EQ("val2", Get(Key(2)));
|
|
|
|
|
Destroy(options);
|
|
|
|
|
}
|
|
|
|
|
|
2021-11-08 23:47:53 +00:00
|
|
|
// Auto resume fromt the flush retryable IO error. Activate the FS before the
|
|
|
|
|
// first resume. Resume is successful
|
2023-10-31 23:13:36 +00:00
|
|
|
TEST_F(DBErrorHandlingFSTest, FlushWritRetryableErrorAutoRecover1) {
|
2020-07-15 18:02:44 +00:00
|
|
|
// Activate the FS before the first resume
|
2023-10-31 23:13:36 +00:00
|
|
|
std::shared_ptr<ErrorHandlerFSListener> listener =
|
|
|
|
|
std::make_shared<ErrorHandlerFSListener>();
|
2020-07-15 18:02:44 +00:00
|
|
|
Options options = GetDefaultOptions();
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
options.env = fault_env_.get();
|
2020-07-15 18:02:44 +00:00
|
|
|
options.create_if_missing = true;
|
|
|
|
|
options.listeners.emplace_back(listener);
|
|
|
|
|
options.max_bgerror_resume_count = 2;
|
|
|
|
|
options.bgerror_resume_retry_interval = 100000; // 0.1 second
|
|
|
|
|
Status s;
|
|
|
|
|
|
|
|
|
|
listener->EnableAutoRecovery(false);
|
|
|
|
|
DestroyAndReopen(options);
|
|
|
|
|
|
|
|
|
|
IOStatus error_msg = IOStatus::IOError("Retryable IO Error");
|
|
|
|
|
error_msg.SetRetryable(true);
|
|
|
|
|
|
2020-10-02 23:39:17 +00:00
|
|
|
ASSERT_OK(Put(Key(1), "val1"));
|
2020-07-15 18:02:44 +00:00
|
|
|
SyncPoint::GetInstance()->SetCallBack(
|
|
|
|
|
"BuildTable:BeforeFinishBuildTable",
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
[&](void*) { fault_fs_->SetFilesystemActive(false, error_msg); });
|
2020-07-18 06:26:07 +00:00
|
|
|
|
2020-07-15 18:02:44 +00:00
|
|
|
SyncPoint::GetInstance()->EnableProcessing();
|
|
|
|
|
s = Flush();
|
2021-03-26 04:41:44 +00:00
|
|
|
ASSERT_EQ(s.severity(), ROCKSDB_NAMESPACE::Status::Severity::kSoftError);
|
2020-07-15 18:02:44 +00:00
|
|
|
SyncPoint::GetInstance()->DisableProcessing();
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
fault_fs_->SetFilesystemActive(true);
|
2020-07-18 06:26:07 +00:00
|
|
|
ASSERT_EQ(listener->WaitForRecovery(5000000), true);
|
2020-07-15 18:02:44 +00:00
|
|
|
|
|
|
|
|
ASSERT_EQ("val1", Get(Key(1)));
|
|
|
|
|
Reopen(options);
|
|
|
|
|
ASSERT_EQ("val1", Get(Key(1)));
|
2020-10-02 23:39:17 +00:00
|
|
|
ASSERT_OK(Put(Key(2), "val2"));
|
2020-12-24 00:54:05 +00:00
|
|
|
ASSERT_OK(Flush());
|
2020-07-15 18:02:44 +00:00
|
|
|
ASSERT_EQ("val2", Get(Key(2)));
|
|
|
|
|
|
|
|
|
|
Destroy(options);
|
|
|
|
|
}
|
|
|
|
|
|
2021-11-08 23:47:53 +00:00
|
|
|
// Auto resume fromt the flush retryable IO error and set the retry limit count.
|
|
|
|
|
// Never activate the FS and auto resume should fail at the end
|
2023-10-31 23:13:36 +00:00
|
|
|
TEST_F(DBErrorHandlingFSTest, FlushWritRetryableErrorAutoRecover2) {
|
2020-07-15 18:02:44 +00:00
|
|
|
// Fail all the resume and let user to resume
|
2023-10-31 23:13:36 +00:00
|
|
|
std::shared_ptr<ErrorHandlerFSListener> listener =
|
|
|
|
|
std::make_shared<ErrorHandlerFSListener>();
|
2020-07-15 18:02:44 +00:00
|
|
|
Options options = GetDefaultOptions();
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
options.env = fault_env_.get();
|
2020-07-15 18:02:44 +00:00
|
|
|
options.create_if_missing = true;
|
|
|
|
|
options.listeners.emplace_back(listener);
|
|
|
|
|
options.max_bgerror_resume_count = 2;
|
|
|
|
|
options.bgerror_resume_retry_interval = 100000; // 0.1 second
|
|
|
|
|
Status s;
|
|
|
|
|
|
|
|
|
|
listener->EnableAutoRecovery(false);
|
|
|
|
|
DestroyAndReopen(options);
|
|
|
|
|
|
|
|
|
|
IOStatus error_msg = IOStatus::IOError("Retryable IO Error");
|
|
|
|
|
error_msg.SetRetryable(true);
|
|
|
|
|
|
2020-10-02 23:39:17 +00:00
|
|
|
ASSERT_OK(Put(Key(1), "val1"));
|
2020-07-15 18:02:44 +00:00
|
|
|
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->LoadDependency(
|
2023-10-31 23:13:36 +00:00
|
|
|
{{"FlushWritRetryableeErrorAutoRecover2:0",
|
2020-07-15 18:02:44 +00:00
|
|
|
"RecoverFromRetryableBGIOError:BeforeStart"},
|
|
|
|
|
{"RecoverFromRetryableBGIOError:LoopOut",
|
2023-10-31 23:13:36 +00:00
|
|
|
"FlushWritRetryableeErrorAutoRecover2:1"}});
|
2020-07-15 18:02:44 +00:00
|
|
|
SyncPoint::GetInstance()->SetCallBack(
|
|
|
|
|
"BuildTable:BeforeFinishBuildTable",
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
[&](void*) { fault_fs_->SetFilesystemActive(false, error_msg); });
|
2020-07-15 18:02:44 +00:00
|
|
|
SyncPoint::GetInstance()->EnableProcessing();
|
|
|
|
|
s = Flush();
|
2021-03-26 04:41:44 +00:00
|
|
|
ASSERT_EQ(s.severity(), ROCKSDB_NAMESPACE::Status::Severity::kSoftError);
|
2023-10-31 23:13:36 +00:00
|
|
|
TEST_SYNC_POINT("FlushWritRetryableeErrorAutoRecover2:0");
|
|
|
|
|
TEST_SYNC_POINT("FlushWritRetryableeErrorAutoRecover2:1");
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
fault_fs_->SetFilesystemActive(true);
|
2020-07-15 18:02:44 +00:00
|
|
|
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->ClearAllCallBacks();
|
|
|
|
|
SyncPoint::GetInstance()->DisableProcessing();
|
|
|
|
|
|
|
|
|
|
ASSERT_EQ("val1", Get(Key(1)));
|
|
|
|
|
// Auto resume fails due to FS does not recover during resume. User call
|
|
|
|
|
// resume manually here.
|
|
|
|
|
s = dbfull()->Resume();
|
|
|
|
|
ASSERT_EQ("val1", Get(Key(1)));
|
2020-12-24 00:54:05 +00:00
|
|
|
ASSERT_OK(s);
|
2020-10-02 23:39:17 +00:00
|
|
|
ASSERT_OK(Put(Key(2), "val2"));
|
2020-12-24 00:54:05 +00:00
|
|
|
ASSERT_OK(Flush());
|
2020-07-15 18:02:44 +00:00
|
|
|
ASSERT_EQ("val2", Get(Key(2)));
|
|
|
|
|
|
|
|
|
|
Destroy(options);
|
|
|
|
|
}
|
|
|
|
|
|
2021-11-08 23:47:53 +00:00
|
|
|
// Auto resume fromt the flush retryable IO error and set the retry limit count.
|
|
|
|
|
// Fail the first resume and let the second resume be successful.
|
2020-07-15 18:02:44 +00:00
|
|
|
TEST_F(DBErrorHandlingFSTest, ManifestWriteRetryableErrorAutoRecover) {
|
|
|
|
|
// Fail the first resume and let the second resume be successful
|
2023-10-31 23:13:36 +00:00
|
|
|
std::shared_ptr<ErrorHandlerFSListener> listener =
|
|
|
|
|
std::make_shared<ErrorHandlerFSListener>();
|
2020-07-15 18:02:44 +00:00
|
|
|
Options options = GetDefaultOptions();
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
options.env = fault_env_.get();
|
2020-07-15 18:02:44 +00:00
|
|
|
options.create_if_missing = true;
|
|
|
|
|
options.listeners.emplace_back(listener);
|
|
|
|
|
options.max_bgerror_resume_count = 2;
|
|
|
|
|
options.bgerror_resume_retry_interval = 100000; // 0.1 second
|
|
|
|
|
Status s;
|
|
|
|
|
std::string old_manifest;
|
|
|
|
|
std::string new_manifest;
|
|
|
|
|
|
|
|
|
|
listener->EnableAutoRecovery(false);
|
|
|
|
|
DestroyAndReopen(options);
|
|
|
|
|
old_manifest = GetManifestNameFromLiveFiles();
|
|
|
|
|
|
|
|
|
|
IOStatus error_msg = IOStatus::IOError("Retryable IO Error");
|
|
|
|
|
error_msg.SetRetryable(true);
|
|
|
|
|
|
2020-10-02 23:39:17 +00:00
|
|
|
ASSERT_OK(Put(Key(0), "val"));
|
|
|
|
|
ASSERT_OK(Flush());
|
|
|
|
|
ASSERT_OK(Put(Key(1), "val"));
|
2020-07-15 18:02:44 +00:00
|
|
|
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->LoadDependency(
|
|
|
|
|
{{"RecoverFromRetryableBGIOError:BeforeStart",
|
|
|
|
|
"ManifestWriteRetryableErrorAutoRecover:0"},
|
|
|
|
|
{"ManifestWriteRetryableErrorAutoRecover:1",
|
|
|
|
|
"RecoverFromRetryableBGIOError:BeforeWait1"},
|
|
|
|
|
{"RecoverFromRetryableBGIOError:RecoverSuccess",
|
|
|
|
|
"ManifestWriteRetryableErrorAutoRecover:2"}});
|
|
|
|
|
SyncPoint::GetInstance()->SetCallBack(
|
|
|
|
|
"VersionSet::LogAndApply:WriteManifest",
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
[&](void*) { fault_fs_->SetFilesystemActive(false, error_msg); });
|
2020-07-15 18:02:44 +00:00
|
|
|
SyncPoint::GetInstance()->EnableProcessing();
|
|
|
|
|
s = Flush();
|
2021-03-26 04:41:44 +00:00
|
|
|
ASSERT_EQ(s.severity(), ROCKSDB_NAMESPACE::Status::Severity::kSoftError);
|
2020-07-15 18:02:44 +00:00
|
|
|
TEST_SYNC_POINT("ManifestWriteRetryableErrorAutoRecover:0");
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
fault_fs_->SetFilesystemActive(true);
|
2020-07-15 18:02:44 +00:00
|
|
|
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->ClearAllCallBacks();
|
|
|
|
|
TEST_SYNC_POINT("ManifestWriteRetryableErrorAutoRecover:1");
|
|
|
|
|
TEST_SYNC_POINT("ManifestWriteRetryableErrorAutoRecover:2");
|
|
|
|
|
SyncPoint::GetInstance()->DisableProcessing();
|
|
|
|
|
|
|
|
|
|
new_manifest = GetManifestNameFromLiveFiles();
|
|
|
|
|
ASSERT_NE(new_manifest, old_manifest);
|
|
|
|
|
|
|
|
|
|
Reopen(options);
|
|
|
|
|
ASSERT_EQ("val", Get(Key(0)));
|
|
|
|
|
ASSERT_EQ("val", Get(Key(1)));
|
|
|
|
|
Close();
|
|
|
|
|
}
|
|
|
|
|
|
2020-12-03 02:22:05 +00:00
|
|
|
TEST_F(DBErrorHandlingFSTest, ManifestWriteNoWALRetryableErrorAutoRecover) {
|
|
|
|
|
// Fail the first resume and let the second resume be successful
|
2023-10-31 23:13:36 +00:00
|
|
|
std::shared_ptr<ErrorHandlerFSListener> listener =
|
|
|
|
|
std::make_shared<ErrorHandlerFSListener>();
|
2020-12-03 02:22:05 +00:00
|
|
|
Options options = GetDefaultOptions();
|
|
|
|
|
options.env = fault_env_.get();
|
|
|
|
|
options.create_if_missing = true;
|
|
|
|
|
options.listeners.emplace_back(listener);
|
|
|
|
|
options.max_bgerror_resume_count = 2;
|
|
|
|
|
options.bgerror_resume_retry_interval = 100000; // 0.1 second
|
|
|
|
|
Status s;
|
|
|
|
|
std::string old_manifest;
|
|
|
|
|
std::string new_manifest;
|
|
|
|
|
|
|
|
|
|
listener->EnableAutoRecovery(false);
|
|
|
|
|
DestroyAndReopen(options);
|
|
|
|
|
old_manifest = GetManifestNameFromLiveFiles();
|
|
|
|
|
|
|
|
|
|
IOStatus error_msg = IOStatus::IOError("Retryable IO Error");
|
|
|
|
|
error_msg.SetRetryable(true);
|
|
|
|
|
|
|
|
|
|
WriteOptions wo = WriteOptions();
|
|
|
|
|
wo.disableWAL = true;
|
|
|
|
|
ASSERT_OK(Put(Key(0), "val", wo));
|
|
|
|
|
ASSERT_OK(Flush());
|
|
|
|
|
ASSERT_OK(Put(Key(1), "val", wo));
|
|
|
|
|
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->LoadDependency(
|
|
|
|
|
{{"RecoverFromRetryableBGIOError:BeforeStart",
|
|
|
|
|
"ManifestWriteNoWALRetryableErrorAutoRecover:0"},
|
|
|
|
|
{"ManifestWriteNoWALRetryableErrorAutoRecover:1",
|
|
|
|
|
"RecoverFromRetryableBGIOError:BeforeWait1"},
|
|
|
|
|
{"RecoverFromRetryableBGIOError:RecoverSuccess",
|
|
|
|
|
"ManifestWriteNoWALRetryableErrorAutoRecover:2"}});
|
|
|
|
|
SyncPoint::GetInstance()->SetCallBack(
|
|
|
|
|
"VersionSet::LogAndApply:WriteManifest",
|
|
|
|
|
[&](void*) { fault_fs_->SetFilesystemActive(false, error_msg); });
|
|
|
|
|
SyncPoint::GetInstance()->EnableProcessing();
|
|
|
|
|
s = Flush();
|
|
|
|
|
ASSERT_EQ(s.severity(), ROCKSDB_NAMESPACE::Status::Severity::kSoftError);
|
|
|
|
|
TEST_SYNC_POINT("ManifestWriteNoWALRetryableErrorAutoRecover:0");
|
|
|
|
|
fault_fs_->SetFilesystemActive(true);
|
|
|
|
|
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->ClearAllCallBacks();
|
|
|
|
|
TEST_SYNC_POINT("ManifestWriteNoWALRetryableErrorAutoRecover:1");
|
|
|
|
|
TEST_SYNC_POINT("ManifestWriteNoWALRetryableErrorAutoRecover:2");
|
|
|
|
|
SyncPoint::GetInstance()->DisableProcessing();
|
|
|
|
|
|
|
|
|
|
new_manifest = GetManifestNameFromLiveFiles();
|
|
|
|
|
ASSERT_NE(new_manifest, old_manifest);
|
|
|
|
|
|
|
|
|
|
Reopen(options);
|
|
|
|
|
ASSERT_EQ("val", Get(Key(0)));
|
|
|
|
|
ASSERT_EQ("val", Get(Key(1)));
|
|
|
|
|
Close();
|
|
|
|
|
}
|
|
|
|
|
|
2020-07-15 18:02:44 +00:00
|
|
|
TEST_F(DBErrorHandlingFSTest,
|
|
|
|
|
CompactionManifestWriteRetryableErrorAutoRecover) {
|
2023-10-31 23:13:36 +00:00
|
|
|
std::shared_ptr<ErrorHandlerFSListener> listener =
|
|
|
|
|
std::make_shared<ErrorHandlerFSListener>();
|
2020-07-15 18:02:44 +00:00
|
|
|
Options options = GetDefaultOptions();
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
options.env = fault_env_.get();
|
2020-07-15 18:02:44 +00:00
|
|
|
options.create_if_missing = true;
|
|
|
|
|
options.level0_file_num_compaction_trigger = 2;
|
|
|
|
|
options.listeners.emplace_back(listener);
|
|
|
|
|
options.max_bgerror_resume_count = 2;
|
|
|
|
|
options.bgerror_resume_retry_interval = 100000; // 0.1 second
|
|
|
|
|
Status s;
|
|
|
|
|
std::string old_manifest;
|
|
|
|
|
std::string new_manifest;
|
|
|
|
|
std::atomic<bool> fail_manifest(false);
|
|
|
|
|
DestroyAndReopen(options);
|
|
|
|
|
old_manifest = GetManifestNameFromLiveFiles();
|
|
|
|
|
|
|
|
|
|
IOStatus error_msg = IOStatus::IOError("Retryable IO Error");
|
|
|
|
|
error_msg.SetRetryable(true);
|
|
|
|
|
|
2020-10-02 23:39:17 +00:00
|
|
|
ASSERT_OK(Put(Key(0), "val"));
|
|
|
|
|
ASSERT_OK(Put(Key(2), "val"));
|
2020-12-24 00:54:05 +00:00
|
|
|
ASSERT_OK(Flush());
|
2020-07-15 18:02:44 +00:00
|
|
|
|
|
|
|
|
listener->OverrideBGError(Status(error_msg, Status::Severity::kHardError));
|
|
|
|
|
listener->EnableAutoRecovery(false);
|
|
|
|
|
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->LoadDependency(
|
|
|
|
|
// Wait for flush of 2nd L0 file before starting compaction
|
|
|
|
|
{{"DBImpl::FlushMemTable:FlushMemTableFinished",
|
|
|
|
|
"BackgroundCallCompaction:0"},
|
|
|
|
|
// Wait for compaction to detect manifest write error
|
|
|
|
|
{"BackgroundCallCompaction:1", "CompactionManifestWriteErrorAR:0"},
|
|
|
|
|
// Make compaction thread wait for error to be cleared
|
|
|
|
|
{"CompactionManifestWriteErrorAR:1",
|
|
|
|
|
"DBImpl::BackgroundCallCompaction:FoundObsoleteFiles"},
|
|
|
|
|
{"CompactionManifestWriteErrorAR:2",
|
|
|
|
|
"RecoverFromRetryableBGIOError:BeforeStart"},
|
|
|
|
|
// Fail the first resume, before the wait in resume
|
|
|
|
|
{"RecoverFromRetryableBGIOError:BeforeResume0",
|
|
|
|
|
"CompactionManifestWriteErrorAR:3"},
|
|
|
|
|
// Activate the FS before the second resume
|
|
|
|
|
{"CompactionManifestWriteErrorAR:4",
|
|
|
|
|
"RecoverFromRetryableBGIOError:BeforeResume1"},
|
|
|
|
|
// Wait the auto resume be sucessful
|
|
|
|
|
{"RecoverFromRetryableBGIOError:RecoverSuccess",
|
|
|
|
|
"CompactionManifestWriteErrorAR:5"}});
|
|
|
|
|
// trigger manifest write failure in compaction thread
|
|
|
|
|
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->SetCallBack(
|
|
|
|
|
"BackgroundCallCompaction:0", [&](void*) { fail_manifest.store(true); });
|
|
|
|
|
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->SetCallBack(
|
|
|
|
|
"VersionSet::LogAndApply:WriteManifest", [&](void*) {
|
|
|
|
|
if (fail_manifest.load()) {
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
fault_fs_->SetFilesystemActive(false, error_msg);
|
2020-07-15 18:02:44 +00:00
|
|
|
}
|
|
|
|
|
});
|
|
|
|
|
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->EnableProcessing();
|
|
|
|
|
|
2020-10-02 23:39:17 +00:00
|
|
|
ASSERT_OK(Put(Key(1), "val"));
|
2020-07-15 18:02:44 +00:00
|
|
|
s = Flush();
|
2020-12-24 00:54:05 +00:00
|
|
|
ASSERT_OK(s);
|
2020-07-15 18:02:44 +00:00
|
|
|
|
|
|
|
|
TEST_SYNC_POINT("CompactionManifestWriteErrorAR:0");
|
|
|
|
|
TEST_SYNC_POINT("CompactionManifestWriteErrorAR:1");
|
|
|
|
|
|
|
|
|
|
s = dbfull()->TEST_WaitForCompact();
|
|
|
|
|
ASSERT_EQ(s.severity(), ROCKSDB_NAMESPACE::Status::Severity::kHardError);
|
|
|
|
|
TEST_SYNC_POINT("CompactionManifestWriteErrorAR:2");
|
|
|
|
|
TEST_SYNC_POINT("CompactionManifestWriteErrorAR:3");
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
fault_fs_->SetFilesystemActive(true);
|
2020-07-15 18:02:44 +00:00
|
|
|
SyncPoint::GetInstance()->ClearAllCallBacks();
|
|
|
|
|
TEST_SYNC_POINT("CompactionManifestWriteErrorAR:4");
|
|
|
|
|
TEST_SYNC_POINT("CompactionManifestWriteErrorAR:5");
|
|
|
|
|
SyncPoint::GetInstance()->DisableProcessing();
|
|
|
|
|
|
|
|
|
|
new_manifest = GetManifestNameFromLiveFiles();
|
|
|
|
|
ASSERT_NE(new_manifest, old_manifest);
|
|
|
|
|
|
|
|
|
|
Reopen(options);
|
|
|
|
|
ASSERT_EQ("val", Get(Key(0)));
|
|
|
|
|
ASSERT_EQ("val", Get(Key(1)));
|
|
|
|
|
ASSERT_EQ("val", Get(Key(2)));
|
|
|
|
|
Close();
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
TEST_F(DBErrorHandlingFSTest, CompactionWriteRetryableErrorAutoRecover) {
|
|
|
|
|
// In this test, in the first round of compaction, the FS is set to error.
|
|
|
|
|
// So the first compaction fails due to retryable IO error and it is mapped
|
|
|
|
|
// to soft error. Then, compaction is rescheduled, in the second round of
|
|
|
|
|
// compaction, the FS is set to active and compaction is successful, so
|
|
|
|
|
// the test will hit the CompactionJob::FinishCompactionOutputFile1 sync
|
|
|
|
|
// point.
|
2023-10-31 23:13:36 +00:00
|
|
|
std::shared_ptr<ErrorHandlerFSListener> listener =
|
|
|
|
|
std::make_shared<ErrorHandlerFSListener>();
|
2020-07-15 18:02:44 +00:00
|
|
|
Options options = GetDefaultOptions();
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
options.env = fault_env_.get();
|
2020-07-15 18:02:44 +00:00
|
|
|
options.create_if_missing = true;
|
|
|
|
|
options.level0_file_num_compaction_trigger = 2;
|
|
|
|
|
options.listeners.emplace_back(listener);
|
|
|
|
|
Status s;
|
|
|
|
|
std::atomic<bool> fail_first(false);
|
|
|
|
|
std::atomic<bool> fail_second(true);
|
|
|
|
|
DestroyAndReopen(options);
|
|
|
|
|
|
|
|
|
|
IOStatus error_msg = IOStatus::IOError("Retryable IO Error");
|
|
|
|
|
error_msg.SetRetryable(true);
|
|
|
|
|
|
2020-10-02 23:39:17 +00:00
|
|
|
ASSERT_OK(Put(Key(0), "va;"));
|
|
|
|
|
ASSERT_OK(Put(Key(2), "va;"));
|
2020-07-15 18:02:44 +00:00
|
|
|
s = Flush();
|
2020-12-24 00:54:05 +00:00
|
|
|
ASSERT_OK(s);
|
2020-07-15 18:02:44 +00:00
|
|
|
|
|
|
|
|
listener->OverrideBGError(Status(error_msg, Status::Severity::kHardError));
|
|
|
|
|
listener->EnableAutoRecovery(false);
|
|
|
|
|
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->LoadDependency(
|
|
|
|
|
{{"DBImpl::FlushMemTable:FlushMemTableFinished",
|
|
|
|
|
"BackgroundCallCompaction:0"},
|
|
|
|
|
{"CompactionJob::FinishCompactionOutputFile1",
|
|
|
|
|
"CompactionWriteRetryableErrorAutoRecover0"}});
|
|
|
|
|
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->SetCallBack(
|
|
|
|
|
"DBImpl::BackgroundCompaction:Start",
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
[&](void*) { fault_fs_->SetFilesystemActive(true); });
|
2020-07-15 18:02:44 +00:00
|
|
|
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->SetCallBack(
|
|
|
|
|
"BackgroundCallCompaction:0", [&](void*) { fail_first.store(true); });
|
|
|
|
|
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->SetCallBack(
|
|
|
|
|
"CompactionJob::OpenCompactionOutputFile", [&](void*) {
|
|
|
|
|
if (fail_first.load() && fail_second.load()) {
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
fault_fs_->SetFilesystemActive(false, error_msg);
|
2020-07-15 18:02:44 +00:00
|
|
|
fail_second.store(false);
|
|
|
|
|
}
|
|
|
|
|
});
|
|
|
|
|
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->EnableProcessing();
|
|
|
|
|
|
2020-10-02 23:39:17 +00:00
|
|
|
ASSERT_OK(Put(Key(1), "val"));
|
2020-07-15 18:02:44 +00:00
|
|
|
s = Flush();
|
2020-12-24 00:54:05 +00:00
|
|
|
ASSERT_OK(s);
|
2020-07-15 18:02:44 +00:00
|
|
|
|
|
|
|
|
s = dbfull()->TEST_WaitForCompact();
|
2021-01-28 01:56:17 +00:00
|
|
|
ASSERT_OK(s);
|
2020-07-15 18:02:44 +00:00
|
|
|
TEST_SYNC_POINT("CompactionWriteRetryableErrorAutoRecover0");
|
|
|
|
|
SyncPoint::GetInstance()->ClearAllCallBacks();
|
|
|
|
|
SyncPoint::GetInstance()->DisableProcessing();
|
|
|
|
|
Destroy(options);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
TEST_F(DBErrorHandlingFSTest, WALWriteRetryableErrorAutoRecover1) {
|
2023-10-31 23:13:36 +00:00
|
|
|
std::shared_ptr<ErrorHandlerFSListener> listener =
|
|
|
|
|
std::make_shared<ErrorHandlerFSListener>();
|
2020-07-15 18:02:44 +00:00
|
|
|
Options options = GetDefaultOptions();
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
options.env = fault_env_.get();
|
2020-07-15 18:02:44 +00:00
|
|
|
options.create_if_missing = true;
|
|
|
|
|
options.writable_file_max_buffer_size = 32768;
|
|
|
|
|
options.listeners.emplace_back(listener);
|
|
|
|
|
options.paranoid_checks = true;
|
|
|
|
|
options.max_bgerror_resume_count = 2;
|
|
|
|
|
options.bgerror_resume_retry_interval = 100000; // 0.1 second
|
|
|
|
|
Status s;
|
|
|
|
|
Random rnd(301);
|
|
|
|
|
|
|
|
|
|
DestroyAndReopen(options);
|
|
|
|
|
|
|
|
|
|
IOStatus error_msg = IOStatus::IOError("Retryable IO Error");
|
|
|
|
|
error_msg.SetRetryable(true);
|
|
|
|
|
|
|
|
|
|
// For the first batch, write is successful, require sync
|
|
|
|
|
{
|
|
|
|
|
WriteBatch batch;
|
|
|
|
|
|
|
|
|
|
for (auto i = 0; i < 100; ++i) {
|
2020-10-02 23:39:17 +00:00
|
|
|
ASSERT_OK(batch.Put(Key(i), rnd.RandomString(1024)));
|
2020-07-15 18:02:44 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
WriteOptions wopts;
|
|
|
|
|
wopts.sync = true;
|
2020-12-24 00:54:05 +00:00
|
|
|
ASSERT_OK(dbfull()->Write(wopts, &batch));
|
2020-07-15 18:02:44 +00:00
|
|
|
};
|
|
|
|
|
|
|
|
|
|
// For the second batch, the first 2 file Append are successful, then the
|
|
|
|
|
// following Append fails due to file system retryable IOError.
|
|
|
|
|
{
|
|
|
|
|
WriteBatch batch;
|
|
|
|
|
int write_error = 0;
|
|
|
|
|
|
|
|
|
|
for (auto i = 100; i < 200; ++i) {
|
2020-10-02 23:39:17 +00:00
|
|
|
ASSERT_OK(batch.Put(Key(i), rnd.RandomString(1024)));
|
2020-07-15 18:02:44 +00:00
|
|
|
}
|
|
|
|
|
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->LoadDependency(
|
2021-01-27 01:00:17 +00:00
|
|
|
{{"WALWriteErrorDone", "RecoverFromRetryableBGIOError:BeforeStart"},
|
|
|
|
|
{"RecoverFromRetryableBGIOError:BeforeResume0", "WALWriteError1:0"},
|
2020-07-15 18:02:44 +00:00
|
|
|
{"WALWriteError1:1", "RecoverFromRetryableBGIOError:BeforeResume1"},
|
|
|
|
|
{"RecoverFromRetryableBGIOError:RecoverSuccess", "WALWriteError1:2"}});
|
|
|
|
|
|
|
|
|
|
SyncPoint::GetInstance()->SetCallBack(
|
|
|
|
|
"WritableFileWriter::Append:BeforePrepareWrite", [&](void*) {
|
|
|
|
|
write_error++;
|
|
|
|
|
if (write_error > 2) {
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
fault_fs_->SetFilesystemActive(false, error_msg);
|
2020-07-15 18:02:44 +00:00
|
|
|
}
|
|
|
|
|
});
|
|
|
|
|
SyncPoint::GetInstance()->EnableProcessing();
|
|
|
|
|
WriteOptions wopts;
|
|
|
|
|
wopts.sync = true;
|
|
|
|
|
s = dbfull()->Write(wopts, &batch);
|
|
|
|
|
ASSERT_EQ(true, s.IsIOError());
|
2021-01-27 01:00:17 +00:00
|
|
|
TEST_SYNC_POINT("WALWriteErrorDone");
|
2020-07-15 18:02:44 +00:00
|
|
|
|
|
|
|
|
TEST_SYNC_POINT("WALWriteError1:0");
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
fault_fs_->SetFilesystemActive(true);
|
2020-07-15 18:02:44 +00:00
|
|
|
SyncPoint::GetInstance()->ClearAllCallBacks();
|
|
|
|
|
TEST_SYNC_POINT("WALWriteError1:1");
|
|
|
|
|
TEST_SYNC_POINT("WALWriteError1:2");
|
|
|
|
|
}
|
|
|
|
|
SyncPoint::GetInstance()->DisableProcessing();
|
|
|
|
|
|
|
|
|
|
// Data in corrupted WAL are not stored
|
|
|
|
|
for (auto i = 0; i < 199; ++i) {
|
|
|
|
|
if (i < 100) {
|
|
|
|
|
ASSERT_NE(Get(Key(i)), "NOT_FOUND");
|
|
|
|
|
} else {
|
|
|
|
|
ASSERT_EQ(Get(Key(i)), "NOT_FOUND");
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
// Resume and write a new batch, should be in the WAL
|
|
|
|
|
{
|
|
|
|
|
WriteBatch batch;
|
|
|
|
|
|
|
|
|
|
for (auto i = 200; i < 300; ++i) {
|
2020-10-02 23:39:17 +00:00
|
|
|
ASSERT_OK(batch.Put(Key(i), rnd.RandomString(1024)));
|
2020-07-15 18:02:44 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
WriteOptions wopts;
|
|
|
|
|
wopts.sync = true;
|
2020-12-24 00:54:05 +00:00
|
|
|
ASSERT_OK(dbfull()->Write(wopts, &batch));
|
2020-07-15 18:02:44 +00:00
|
|
|
};
|
|
|
|
|
|
|
|
|
|
Reopen(options);
|
|
|
|
|
for (auto i = 0; i < 300; ++i) {
|
|
|
|
|
if (i < 100 || i >= 200) {
|
|
|
|
|
ASSERT_NE(Get(Key(i)), "NOT_FOUND");
|
|
|
|
|
} else {
|
|
|
|
|
ASSERT_EQ(Get(Key(i)), "NOT_FOUND");
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
Close();
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
TEST_F(DBErrorHandlingFSTest, WALWriteRetryableErrorAutoRecover2) {
|
|
|
|
|
// Fail the first recover and try second time.
|
2023-10-31 23:13:36 +00:00
|
|
|
std::shared_ptr<ErrorHandlerFSListener> listener =
|
|
|
|
|
std::make_shared<ErrorHandlerFSListener>();
|
2020-07-15 18:02:44 +00:00
|
|
|
Options options = GetDefaultOptions();
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
options.env = fault_env_.get();
|
2020-07-15 18:02:44 +00:00
|
|
|
options.create_if_missing = true;
|
|
|
|
|
options.writable_file_max_buffer_size = 32768;
|
|
|
|
|
options.listeners.emplace_back(listener);
|
|
|
|
|
options.paranoid_checks = true;
|
|
|
|
|
options.max_bgerror_resume_count = 2;
|
|
|
|
|
options.bgerror_resume_retry_interval = 100000; // 0.1 second
|
|
|
|
|
Status s;
|
|
|
|
|
Random rnd(301);
|
|
|
|
|
|
|
|
|
|
DestroyAndReopen(options);
|
|
|
|
|
|
|
|
|
|
IOStatus error_msg = IOStatus::IOError("Retryable IO Error");
|
|
|
|
|
error_msg.SetRetryable(true);
|
|
|
|
|
|
|
|
|
|
// For the first batch, write is successful, require sync
|
|
|
|
|
{
|
|
|
|
|
WriteBatch batch;
|
|
|
|
|
|
|
|
|
|
for (auto i = 0; i < 100; ++i) {
|
2020-10-02 23:39:17 +00:00
|
|
|
ASSERT_OK(batch.Put(Key(i), rnd.RandomString(1024)));
|
2020-07-15 18:02:44 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
WriteOptions wopts;
|
|
|
|
|
wopts.sync = true;
|
2020-12-24 00:54:05 +00:00
|
|
|
ASSERT_OK(dbfull()->Write(wopts, &batch));
|
2020-07-15 18:02:44 +00:00
|
|
|
};
|
|
|
|
|
|
|
|
|
|
// For the second batch, the first 2 file Append are successful, then the
|
|
|
|
|
// following Append fails due to file system retryable IOError.
|
|
|
|
|
{
|
|
|
|
|
WriteBatch batch;
|
|
|
|
|
int write_error = 0;
|
|
|
|
|
|
|
|
|
|
for (auto i = 100; i < 200; ++i) {
|
2020-10-02 23:39:17 +00:00
|
|
|
ASSERT_OK(batch.Put(Key(i), rnd.RandomString(1024)));
|
2020-07-15 18:02:44 +00:00
|
|
|
}
|
|
|
|
|
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->LoadDependency(
|
|
|
|
|
{{"RecoverFromRetryableBGIOError:BeforeWait0", "WALWriteError2:0"},
|
|
|
|
|
{"WALWriteError2:1", "RecoverFromRetryableBGIOError:BeforeWait1"},
|
|
|
|
|
{"RecoverFromRetryableBGIOError:RecoverSuccess", "WALWriteError2:2"}});
|
|
|
|
|
|
|
|
|
|
SyncPoint::GetInstance()->SetCallBack(
|
|
|
|
|
"WritableFileWriter::Append:BeforePrepareWrite", [&](void*) {
|
|
|
|
|
write_error++;
|
|
|
|
|
if (write_error > 2) {
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
fault_fs_->SetFilesystemActive(false, error_msg);
|
2020-07-15 18:02:44 +00:00
|
|
|
}
|
|
|
|
|
});
|
|
|
|
|
SyncPoint::GetInstance()->EnableProcessing();
|
|
|
|
|
WriteOptions wopts;
|
|
|
|
|
wopts.sync = true;
|
|
|
|
|
s = dbfull()->Write(wopts, &batch);
|
|
|
|
|
ASSERT_EQ(true, s.IsIOError());
|
|
|
|
|
|
|
|
|
|
TEST_SYNC_POINT("WALWriteError2:0");
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
fault_fs_->SetFilesystemActive(true);
|
2020-07-15 18:02:44 +00:00
|
|
|
SyncPoint::GetInstance()->ClearAllCallBacks();
|
|
|
|
|
TEST_SYNC_POINT("WALWriteError2:1");
|
|
|
|
|
TEST_SYNC_POINT("WALWriteError2:2");
|
|
|
|
|
}
|
|
|
|
|
SyncPoint::GetInstance()->DisableProcessing();
|
|
|
|
|
|
|
|
|
|
// Data in corrupted WAL are not stored
|
|
|
|
|
for (auto i = 0; i < 199; ++i) {
|
|
|
|
|
if (i < 100) {
|
|
|
|
|
ASSERT_NE(Get(Key(i)), "NOT_FOUND");
|
|
|
|
|
} else {
|
|
|
|
|
ASSERT_EQ(Get(Key(i)), "NOT_FOUND");
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
// Resume and write a new batch, should be in the WAL
|
|
|
|
|
{
|
|
|
|
|
WriteBatch batch;
|
|
|
|
|
|
|
|
|
|
for (auto i = 200; i < 300; ++i) {
|
2020-10-02 23:39:17 +00:00
|
|
|
ASSERT_OK(batch.Put(Key(i), rnd.RandomString(1024)));
|
2020-07-15 18:02:44 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
WriteOptions wopts;
|
|
|
|
|
wopts.sync = true;
|
2020-12-24 00:54:05 +00:00
|
|
|
ASSERT_OK(dbfull()->Write(wopts, &batch));
|
2020-07-15 18:02:44 +00:00
|
|
|
};
|
|
|
|
|
|
|
|
|
|
Reopen(options);
|
|
|
|
|
for (auto i = 0; i < 300; ++i) {
|
|
|
|
|
if (i < 100 || i >= 200) {
|
|
|
|
|
ASSERT_NE(Get(Key(i)), "NOT_FOUND");
|
|
|
|
|
} else {
|
|
|
|
|
ASSERT_EQ(Get(Key(i)), "NOT_FOUND");
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
Close();
|
|
|
|
|
}
|
|
|
|
|
|
2021-12-08 22:22:41 +00:00
|
|
|
// Fail auto resume from a flush retryable error and verify that
|
|
|
|
|
// OnErrorRecoveryEnd listener callback is called
|
2023-10-31 23:13:36 +00:00
|
|
|
TEST_F(DBErrorHandlingFSTest, FlushWritRetryableErrorAbortRecovery) {
|
2021-12-08 22:22:41 +00:00
|
|
|
// Activate the FS before the first resume
|
2023-10-31 23:13:36 +00:00
|
|
|
std::shared_ptr<ErrorHandlerFSListener> listener =
|
|
|
|
|
std::make_shared<ErrorHandlerFSListener>();
|
2021-12-08 22:22:41 +00:00
|
|
|
Options options = GetDefaultOptions();
|
|
|
|
|
options.env = fault_env_.get();
|
|
|
|
|
options.create_if_missing = true;
|
|
|
|
|
options.listeners.emplace_back(listener);
|
|
|
|
|
options.max_bgerror_resume_count = 2;
|
|
|
|
|
options.bgerror_resume_retry_interval = 100000; // 0.1 second
|
|
|
|
|
Status s;
|
|
|
|
|
|
|
|
|
|
listener->EnableAutoRecovery(false);
|
|
|
|
|
DestroyAndReopen(options);
|
|
|
|
|
|
|
|
|
|
IOStatus error_msg = IOStatus::IOError("Retryable IO Error");
|
|
|
|
|
error_msg.SetRetryable(true);
|
|
|
|
|
|
|
|
|
|
ASSERT_OK(Put(Key(1), "val1"));
|
|
|
|
|
SyncPoint::GetInstance()->SetCallBack(
|
|
|
|
|
"BuildTable:BeforeFinishBuildTable",
|
|
|
|
|
[&](void*) { fault_fs_->SetFilesystemActive(false, error_msg); });
|
|
|
|
|
|
|
|
|
|
SyncPoint::GetInstance()->EnableProcessing();
|
|
|
|
|
s = Flush();
|
|
|
|
|
ASSERT_EQ(s.severity(), ROCKSDB_NAMESPACE::Status::Severity::kSoftError);
|
|
|
|
|
ASSERT_EQ(listener->WaitForRecovery(5000000), true);
|
|
|
|
|
ASSERT_EQ(listener->new_bg_error(), Status::Aborted());
|
|
|
|
|
SyncPoint::GetInstance()->DisableProcessing();
|
|
|
|
|
fault_fs_->SetFilesystemActive(true);
|
|
|
|
|
|
|
|
|
|
Destroy(options);
|
|
|
|
|
}
|
|
|
|
|
|
2023-10-25 18:59:09 +00:00
|
|
|
TEST_F(DBErrorHandlingFSTest, FlushErrorRecoveryRaceWithDBDestruction) {
|
|
|
|
|
Options options = GetDefaultOptions();
|
|
|
|
|
options.env = fault_env_.get();
|
|
|
|
|
options.create_if_missing = true;
|
|
|
|
|
std::shared_ptr<ErrorHandlerFSListener> listener =
|
|
|
|
|
std::make_shared<ErrorHandlerFSListener>();
|
|
|
|
|
options.listeners.emplace_back(listener);
|
|
|
|
|
DestroyAndReopen(options);
|
|
|
|
|
ASSERT_OK(Put("k1", "val"));
|
|
|
|
|
|
|
|
|
|
// Inject retryable flush error
|
|
|
|
|
bool error_set = false;
|
|
|
|
|
SyncPoint::GetInstance()->SetCallBack(
|
|
|
|
|
"BuildTable:BeforeOutputValidation", [&](void*) {
|
|
|
|
|
if (error_set) {
|
|
|
|
|
return;
|
|
|
|
|
}
|
|
|
|
|
IOStatus st = IOStatus::IOError("Injected");
|
|
|
|
|
st.SetRetryable(true);
|
|
|
|
|
fault_fs_->SetFilesystemActive(false, st);
|
|
|
|
|
error_set = true;
|
|
|
|
|
});
|
|
|
|
|
|
|
|
|
|
port::Thread db_close_thread;
|
|
|
|
|
SyncPoint::GetInstance()->SetCallBack(
|
|
|
|
|
"BuildTable:BeforeDeleteFile", [&](void*) {
|
|
|
|
|
// Clear retryable flush error injection
|
|
|
|
|
fault_fs_->SetFilesystemActive(true);
|
|
|
|
|
|
|
|
|
|
// Coerce race between ending auto recovery in db destruction and flush
|
|
|
|
|
// error recovery
|
|
|
|
|
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->LoadDependency(
|
|
|
|
|
{{"PostEndAutoRecovery", "FlushJob::WriteLevel0Table"}});
|
|
|
|
|
db_close_thread = port::Thread([&] { Close(); });
|
|
|
|
|
});
|
|
|
|
|
SyncPoint::GetInstance()->EnableProcessing();
|
|
|
|
|
|
|
|
|
|
Status s = Flush();
|
|
|
|
|
ASSERT_NOK(s);
|
|
|
|
|
|
|
|
|
|
int placeholder = 1;
|
|
|
|
|
listener->WaitForRecovery(placeholder);
|
|
|
|
|
ASSERT_TRUE(listener->new_bg_error().IsShutdownInProgress());
|
|
|
|
|
|
|
|
|
|
// Prior to the fix, the db close will crash due to the recovery thread for
|
|
|
|
|
// flush error is not joined by the time of destruction.
|
|
|
|
|
db_close_thread.join();
|
|
|
|
|
|
|
|
|
|
SyncPoint::GetInstance()->DisableProcessing();
|
|
|
|
|
SyncPoint::GetInstance()->ClearAllCallBacks();
|
|
|
|
|
Destroy(options);
|
|
|
|
|
}
|
|
|
|
|
|
2022-03-15 21:45:34 +00:00
|
|
|
TEST_F(DBErrorHandlingFSTest, FlushReadError) {
|
|
|
|
|
std::shared_ptr<ErrorHandlerFSListener> listener =
|
|
|
|
|
std::make_shared<ErrorHandlerFSListener>();
|
|
|
|
|
Options options = GetDefaultOptions();
|
|
|
|
|
options.env = fault_env_.get();
|
|
|
|
|
options.create_if_missing = true;
|
|
|
|
|
options.listeners.emplace_back(listener);
|
|
|
|
|
options.statistics = CreateDBStatistics();
|
|
|
|
|
Status s;
|
|
|
|
|
|
|
|
|
|
listener->EnableAutoRecovery(false);
|
|
|
|
|
DestroyAndReopen(options);
|
|
|
|
|
|
|
|
|
|
ASSERT_OK(Put(Key(0), "val"));
|
|
|
|
|
SyncPoint::GetInstance()->SetCallBack(
|
|
|
|
|
"BuildTable:BeforeOutputValidation", [&](void*) {
|
|
|
|
|
IOStatus st = IOStatus::IOError();
|
|
|
|
|
st.SetRetryable(true);
|
|
|
|
|
st.SetScope(IOStatus::IOErrorScope::kIOErrorScopeFile);
|
|
|
|
|
fault_fs_->SetFilesystemActive(false, st);
|
|
|
|
|
});
|
|
|
|
|
SyncPoint::GetInstance()->SetCallBack(
|
|
|
|
|
"BuildTable:BeforeDeleteFile",
|
|
|
|
|
[&](void*) { fault_fs_->SetFilesystemActive(true, IOStatus::OK()); });
|
|
|
|
|
SyncPoint::GetInstance()->EnableProcessing();
|
|
|
|
|
s = Flush();
|
|
|
|
|
ASSERT_EQ(s.severity(), ROCKSDB_NAMESPACE::Status::Severity::kSoftError);
|
|
|
|
|
SyncPoint::GetInstance()->DisableProcessing();
|
|
|
|
|
fault_fs_->SetFilesystemActive(true);
|
|
|
|
|
ASSERT_EQ(listener->WaitForRecovery(5000000), true);
|
|
|
|
|
ASSERT_EQ(1, options.statistics->getAndResetTickerCount(
|
|
|
|
|
ERROR_HANDLER_BG_ERROR_COUNT));
|
|
|
|
|
ASSERT_EQ(1, options.statistics->getAndResetTickerCount(
|
|
|
|
|
ERROR_HANDLER_BG_IO_ERROR_COUNT));
|
|
|
|
|
ASSERT_EQ(1, options.statistics->getAndResetTickerCount(
|
|
|
|
|
ERROR_HANDLER_BG_RETRYABLE_IO_ERROR_COUNT));
|
|
|
|
|
ASSERT_LE(1, options.statistics->getAndResetTickerCount(
|
|
|
|
|
ERROR_HANDLER_AUTORESUME_COUNT));
|
|
|
|
|
ASSERT_LE(0, options.statistics->getAndResetTickerCount(
|
|
|
|
|
ERROR_HANDLER_AUTORESUME_RETRY_TOTAL_COUNT));
|
|
|
|
|
s = dbfull()->TEST_GetBGError();
|
|
|
|
|
ASSERT_OK(s);
|
|
|
|
|
|
|
|
|
|
Reopen(GetDefaultOptions());
|
|
|
|
|
ASSERT_EQ("val", Get(Key(0)));
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
TEST_F(DBErrorHandlingFSTest, AtomicFlushReadError) {
|
|
|
|
|
std::shared_ptr<ErrorHandlerFSListener> listener =
|
|
|
|
|
std::make_shared<ErrorHandlerFSListener>();
|
|
|
|
|
Options options = GetDefaultOptions();
|
|
|
|
|
options.env = fault_env_.get();
|
|
|
|
|
options.create_if_missing = true;
|
|
|
|
|
options.listeners.emplace_back(listener);
|
|
|
|
|
options.statistics = CreateDBStatistics();
|
|
|
|
|
Status s;
|
|
|
|
|
|
|
|
|
|
listener->EnableAutoRecovery(false);
|
|
|
|
|
options.atomic_flush = true;
|
|
|
|
|
CreateAndReopenWithCF({"pikachu"}, options);
|
|
|
|
|
|
|
|
|
|
ASSERT_OK(Put(0, Key(0), "val"));
|
|
|
|
|
ASSERT_OK(Put(1, Key(0), "val"));
|
|
|
|
|
SyncPoint::GetInstance()->SetCallBack(
|
|
|
|
|
"BuildTable:BeforeOutputValidation", [&](void*) {
|
|
|
|
|
IOStatus st = IOStatus::IOError();
|
|
|
|
|
st.SetRetryable(true);
|
|
|
|
|
st.SetScope(IOStatus::IOErrorScope::kIOErrorScopeFile);
|
|
|
|
|
fault_fs_->SetFilesystemActive(false, st);
|
|
|
|
|
});
|
|
|
|
|
SyncPoint::GetInstance()->SetCallBack(
|
|
|
|
|
"BuildTable:BeforeDeleteFile",
|
|
|
|
|
[&](void*) { fault_fs_->SetFilesystemActive(true, IOStatus::OK()); });
|
|
|
|
|
SyncPoint::GetInstance()->EnableProcessing();
|
|
|
|
|
s = Flush({0, 1});
|
|
|
|
|
ASSERT_EQ(s.severity(), ROCKSDB_NAMESPACE::Status::Severity::kSoftError);
|
|
|
|
|
SyncPoint::GetInstance()->DisableProcessing();
|
|
|
|
|
fault_fs_->SetFilesystemActive(true);
|
|
|
|
|
ASSERT_EQ(listener->WaitForRecovery(5000000), true);
|
|
|
|
|
ASSERT_EQ(1, options.statistics->getAndResetTickerCount(
|
|
|
|
|
ERROR_HANDLER_BG_ERROR_COUNT));
|
|
|
|
|
ASSERT_EQ(1, options.statistics->getAndResetTickerCount(
|
|
|
|
|
ERROR_HANDLER_BG_IO_ERROR_COUNT));
|
|
|
|
|
ASSERT_EQ(1, options.statistics->getAndResetTickerCount(
|
|
|
|
|
ERROR_HANDLER_BG_RETRYABLE_IO_ERROR_COUNT));
|
|
|
|
|
ASSERT_LE(1, options.statistics->getAndResetTickerCount(
|
|
|
|
|
ERROR_HANDLER_AUTORESUME_COUNT));
|
|
|
|
|
ASSERT_LE(0, options.statistics->getAndResetTickerCount(
|
|
|
|
|
ERROR_HANDLER_AUTORESUME_RETRY_TOTAL_COUNT));
|
|
|
|
|
s = dbfull()->TEST_GetBGError();
|
|
|
|
|
ASSERT_OK(s);
|
|
|
|
|
|
2023-08-09 22:46:44 +00:00
|
|
|
ASSERT_OK(TryReopenWithColumnFamilies({kDefaultColumnFamilyName, "pikachu"},
|
|
|
|
|
GetDefaultOptions()));
|
2022-03-15 21:45:34 +00:00
|
|
|
ASSERT_EQ("val", Get(Key(0)));
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
TEST_F(DBErrorHandlingFSTest, AtomicFlushNoSpaceError) {
|
|
|
|
|
std::shared_ptr<ErrorHandlerFSListener> listener =
|
|
|
|
|
std::make_shared<ErrorHandlerFSListener>();
|
|
|
|
|
Options options = GetDefaultOptions();
|
|
|
|
|
options.env = fault_env_.get();
|
|
|
|
|
options.create_if_missing = true;
|
|
|
|
|
options.listeners.emplace_back(listener);
|
|
|
|
|
options.statistics = CreateDBStatistics();
|
|
|
|
|
Status s;
|
|
|
|
|
|
|
|
|
|
listener->EnableAutoRecovery(true);
|
|
|
|
|
options.atomic_flush = true;
|
|
|
|
|
CreateAndReopenWithCF({"pikachu"}, options);
|
|
|
|
|
|
|
|
|
|
ASSERT_OK(Put(0, Key(0), "val"));
|
|
|
|
|
ASSERT_OK(Put(1, Key(0), "val"));
|
|
|
|
|
SyncPoint::GetInstance()->SetCallBack("BuildTable:create_file", [&](void*) {
|
|
|
|
|
IOStatus st = IOStatus::NoSpace();
|
|
|
|
|
fault_fs_->SetFilesystemActive(false, st);
|
|
|
|
|
});
|
|
|
|
|
SyncPoint::GetInstance()->SetCallBack(
|
|
|
|
|
"BuildTable:BeforeDeleteFile",
|
|
|
|
|
[&](void*) { fault_fs_->SetFilesystemActive(true, IOStatus::OK()); });
|
|
|
|
|
SyncPoint::GetInstance()->EnableProcessing();
|
|
|
|
|
s = Flush({0, 1});
|
|
|
|
|
ASSERT_EQ(s.severity(), ROCKSDB_NAMESPACE::Status::Severity::kHardError);
|
|
|
|
|
SyncPoint::GetInstance()->DisableProcessing();
|
|
|
|
|
fault_fs_->SetFilesystemActive(true);
|
|
|
|
|
ASSERT_EQ(listener->WaitForRecovery(5000000), true);
|
|
|
|
|
ASSERT_LE(1, options.statistics->getAndResetTickerCount(
|
|
|
|
|
ERROR_HANDLER_BG_ERROR_COUNT));
|
|
|
|
|
ASSERT_LE(1, options.statistics->getAndResetTickerCount(
|
|
|
|
|
ERROR_HANDLER_BG_IO_ERROR_COUNT));
|
|
|
|
|
s = dbfull()->TEST_GetBGError();
|
|
|
|
|
ASSERT_OK(s);
|
|
|
|
|
|
2023-08-09 22:46:44 +00:00
|
|
|
ASSERT_OK(TryReopenWithColumnFamilies({kDefaultColumnFamilyName, "pikachu"},
|
|
|
|
|
GetDefaultOptions()));
|
2022-03-15 21:45:34 +00:00
|
|
|
ASSERT_EQ("val", Get(Key(0)));
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
TEST_F(DBErrorHandlingFSTest, CompactionReadRetryableErrorAutoRecover) {
|
|
|
|
|
// In this test, in the first round of compaction, the FS is set to error.
|
|
|
|
|
// So the first compaction fails due to retryable IO error and it is mapped
|
|
|
|
|
// to soft error. Then, compaction is rescheduled, in the second round of
|
|
|
|
|
// compaction, the FS is set to active and compaction is successful, so
|
|
|
|
|
// the test will hit the CompactionJob::FinishCompactionOutputFile1 sync
|
|
|
|
|
// point.
|
|
|
|
|
std::shared_ptr<ErrorHandlerFSListener> listener =
|
|
|
|
|
std::make_shared<ErrorHandlerFSListener>();
|
|
|
|
|
Options options = GetDefaultOptions();
|
|
|
|
|
options.env = fault_env_.get();
|
|
|
|
|
options.create_if_missing = true;
|
|
|
|
|
options.level0_file_num_compaction_trigger = 2;
|
|
|
|
|
options.listeners.emplace_back(listener);
|
|
|
|
|
BlockBasedTableOptions table_options;
|
|
|
|
|
table_options.no_block_cache = true;
|
|
|
|
|
options.table_factory.reset(NewBlockBasedTableFactory(table_options));
|
|
|
|
|
Status s;
|
|
|
|
|
std::atomic<bool> fail_first(false);
|
|
|
|
|
std::atomic<bool> fail_second(true);
|
|
|
|
|
Random rnd(301);
|
|
|
|
|
DestroyAndReopen(options);
|
|
|
|
|
|
|
|
|
|
IOStatus error_msg = IOStatus::IOError("Retryable IO Error");
|
|
|
|
|
error_msg.SetRetryable(true);
|
|
|
|
|
|
|
|
|
|
for (int i = 0; i < 100; ++i) {
|
|
|
|
|
ASSERT_OK(Put(Key(i), rnd.RandomString(1024)));
|
|
|
|
|
}
|
|
|
|
|
s = Flush();
|
|
|
|
|
ASSERT_OK(s);
|
|
|
|
|
|
|
|
|
|
listener->OverrideBGError(Status(error_msg, Status::Severity::kHardError));
|
|
|
|
|
listener->EnableAutoRecovery(false);
|
|
|
|
|
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->LoadDependency(
|
|
|
|
|
{{"DBImpl::FlushMemTable:FlushMemTableFinished",
|
|
|
|
|
"BackgroundCallCompaction:0"},
|
|
|
|
|
{"CompactionJob::FinishCompactionOutputFile1",
|
|
|
|
|
"CompactionWriteRetryableErrorAutoRecover0"}});
|
|
|
|
|
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->SetCallBack(
|
|
|
|
|
"DBImpl::BackgroundCompaction:Start",
|
|
|
|
|
[&](void*) { fault_fs_->SetFilesystemActive(true); });
|
|
|
|
|
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->SetCallBack(
|
|
|
|
|
"BackgroundCallCompaction:0", [&](void*) { fail_first.store(true); });
|
|
|
|
|
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->SetCallBack(
|
|
|
|
|
"CompactionJob::Run():PausingManualCompaction:2", [&](void*) {
|
|
|
|
|
if (fail_first.load() && fail_second.load()) {
|
|
|
|
|
fault_fs_->SetFilesystemActive(false, error_msg);
|
|
|
|
|
fail_second.store(false);
|
|
|
|
|
}
|
|
|
|
|
});
|
|
|
|
|
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->EnableProcessing();
|
|
|
|
|
|
|
|
|
|
ASSERT_OK(Put(Key(1), "val"));
|
|
|
|
|
s = Flush();
|
|
|
|
|
ASSERT_OK(s);
|
|
|
|
|
|
|
|
|
|
s = dbfull()->TEST_WaitForCompact();
|
|
|
|
|
ASSERT_OK(s);
|
|
|
|
|
TEST_SYNC_POINT("CompactionWriteRetryableErrorAutoRecover0");
|
|
|
|
|
SyncPoint::GetInstance()->ClearAllCallBacks();
|
|
|
|
|
SyncPoint::GetInstance()->DisableProcessing();
|
|
|
|
|
|
|
|
|
|
Reopen(GetDefaultOptions());
|
|
|
|
|
}
|
|
|
|
|
|
2020-09-14 23:03:32 +00:00
|
|
|
class DBErrorHandlingFencingTest : public DBErrorHandlingFSTest,
|
|
|
|
|
public testing::WithParamInterface<bool> {};
|
|
|
|
|
|
2023-10-31 23:13:36 +00:00
|
|
|
TEST_P(DBErrorHandlingFencingTest, FlushWriteFenced) {
|
|
|
|
|
std::shared_ptr<ErrorHandlerFSListener> listener =
|
|
|
|
|
std::make_shared<ErrorHandlerFSListener>();
|
2020-09-14 23:03:32 +00:00
|
|
|
Options options = GetDefaultOptions();
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
options.env = fault_env_.get();
|
2020-09-14 23:03:32 +00:00
|
|
|
options.create_if_missing = true;
|
|
|
|
|
options.listeners.emplace_back(listener);
|
|
|
|
|
options.paranoid_checks = GetParam();
|
|
|
|
|
Status s;
|
|
|
|
|
|
|
|
|
|
listener->EnableAutoRecovery(true);
|
|
|
|
|
DestroyAndReopen(options);
|
|
|
|
|
|
2020-10-02 23:39:17 +00:00
|
|
|
ASSERT_OK(Put(Key(0), "val"));
|
2020-09-14 23:03:32 +00:00
|
|
|
SyncPoint::GetInstance()->SetCallBack("FlushJob::Start", [&](void*) {
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
fault_fs_->SetFilesystemActive(false, IOStatus::IOFenced("IO fenced"));
|
2020-09-14 23:03:32 +00:00
|
|
|
});
|
|
|
|
|
SyncPoint::GetInstance()->EnableProcessing();
|
|
|
|
|
s = Flush();
|
|
|
|
|
ASSERT_EQ(s.severity(), ROCKSDB_NAMESPACE::Status::Severity::kFatalError);
|
|
|
|
|
ASSERT_TRUE(s.IsIOFenced());
|
|
|
|
|
SyncPoint::GetInstance()->DisableProcessing();
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
fault_fs_->SetFilesystemActive(true);
|
2020-09-14 23:03:32 +00:00
|
|
|
s = dbfull()->Resume();
|
|
|
|
|
ASSERT_TRUE(s.IsIOFenced());
|
|
|
|
|
Destroy(options);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
TEST_P(DBErrorHandlingFencingTest, ManifestWriteFenced) {
|
2023-10-31 23:13:36 +00:00
|
|
|
std::shared_ptr<ErrorHandlerFSListener> listener =
|
|
|
|
|
std::make_shared<ErrorHandlerFSListener>();
|
2020-09-14 23:03:32 +00:00
|
|
|
Options options = GetDefaultOptions();
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
options.env = fault_env_.get();
|
2020-09-14 23:03:32 +00:00
|
|
|
options.create_if_missing = true;
|
|
|
|
|
options.listeners.emplace_back(listener);
|
|
|
|
|
options.paranoid_checks = GetParam();
|
|
|
|
|
Status s;
|
|
|
|
|
std::string old_manifest;
|
|
|
|
|
std::string new_manifest;
|
|
|
|
|
|
|
|
|
|
listener->EnableAutoRecovery(true);
|
|
|
|
|
DestroyAndReopen(options);
|
|
|
|
|
old_manifest = GetManifestNameFromLiveFiles();
|
|
|
|
|
|
2020-10-02 23:39:17 +00:00
|
|
|
ASSERT_OK(Put(Key(0), "val"));
|
2020-12-24 00:54:05 +00:00
|
|
|
ASSERT_OK(Flush());
|
2020-10-02 23:39:17 +00:00
|
|
|
ASSERT_OK(Put(Key(1), "val"));
|
2020-09-14 23:03:32 +00:00
|
|
|
SyncPoint::GetInstance()->SetCallBack(
|
|
|
|
|
"VersionSet::LogAndApply:WriteManifest", [&](void*) {
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
fault_fs_->SetFilesystemActive(false, IOStatus::IOFenced("IO fenced"));
|
2020-09-14 23:03:32 +00:00
|
|
|
});
|
|
|
|
|
SyncPoint::GetInstance()->EnableProcessing();
|
|
|
|
|
s = Flush();
|
|
|
|
|
ASSERT_EQ(s.severity(), ROCKSDB_NAMESPACE::Status::Severity::kFatalError);
|
|
|
|
|
ASSERT_TRUE(s.IsIOFenced());
|
|
|
|
|
SyncPoint::GetInstance()->ClearAllCallBacks();
|
|
|
|
|
SyncPoint::GetInstance()->DisableProcessing();
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
fault_fs_->SetFilesystemActive(true);
|
2020-09-14 23:03:32 +00:00
|
|
|
s = dbfull()->Resume();
|
|
|
|
|
ASSERT_TRUE(s.IsIOFenced());
|
|
|
|
|
Close();
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
TEST_P(DBErrorHandlingFencingTest, CompactionWriteFenced) {
|
2023-10-31 23:13:36 +00:00
|
|
|
std::shared_ptr<ErrorHandlerFSListener> listener =
|
|
|
|
|
std::make_shared<ErrorHandlerFSListener>();
|
2020-09-14 23:03:32 +00:00
|
|
|
Options options = GetDefaultOptions();
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
options.env = fault_env_.get();
|
2020-09-14 23:03:32 +00:00
|
|
|
options.create_if_missing = true;
|
|
|
|
|
options.level0_file_num_compaction_trigger = 2;
|
|
|
|
|
options.listeners.emplace_back(listener);
|
|
|
|
|
options.paranoid_checks = GetParam();
|
|
|
|
|
Status s;
|
|
|
|
|
DestroyAndReopen(options);
|
|
|
|
|
|
2020-10-02 23:39:17 +00:00
|
|
|
ASSERT_OK(Put(Key(0), "va;"));
|
|
|
|
|
ASSERT_OK(Put(Key(2), "va;"));
|
2020-09-14 23:03:32 +00:00
|
|
|
s = Flush();
|
2020-12-24 00:54:05 +00:00
|
|
|
ASSERT_OK(s);
|
2020-09-14 23:03:32 +00:00
|
|
|
|
|
|
|
|
listener->EnableAutoRecovery(true);
|
|
|
|
|
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->LoadDependency(
|
|
|
|
|
{{"DBImpl::FlushMemTable:FlushMemTableFinished",
|
|
|
|
|
"BackgroundCallCompaction:0"}});
|
|
|
|
|
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->SetCallBack(
|
|
|
|
|
"BackgroundCallCompaction:0", [&](void*) {
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
fault_fs_->SetFilesystemActive(false, IOStatus::IOFenced("IO fenced"));
|
2020-09-14 23:03:32 +00:00
|
|
|
});
|
|
|
|
|
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->EnableProcessing();
|
|
|
|
|
|
2020-10-02 23:39:17 +00:00
|
|
|
ASSERT_OK(Put(Key(1), "val"));
|
2020-09-14 23:03:32 +00:00
|
|
|
s = Flush();
|
2020-12-24 00:54:05 +00:00
|
|
|
ASSERT_OK(s);
|
2020-09-14 23:03:32 +00:00
|
|
|
|
|
|
|
|
s = dbfull()->TEST_WaitForCompact();
|
|
|
|
|
ASSERT_EQ(s.severity(), ROCKSDB_NAMESPACE::Status::Severity::kFatalError);
|
|
|
|
|
ASSERT_TRUE(s.IsIOFenced());
|
|
|
|
|
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
fault_fs_->SetFilesystemActive(true);
|
2020-09-14 23:03:32 +00:00
|
|
|
s = dbfull()->Resume();
|
|
|
|
|
ASSERT_TRUE(s.IsIOFenced());
|
|
|
|
|
Destroy(options);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
TEST_P(DBErrorHandlingFencingTest, WALWriteFenced) {
|
2023-10-31 23:13:36 +00:00
|
|
|
std::shared_ptr<ErrorHandlerFSListener> listener =
|
|
|
|
|
std::make_shared<ErrorHandlerFSListener>();
|
2020-09-14 23:03:32 +00:00
|
|
|
Options options = GetDefaultOptions();
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
options.env = fault_env_.get();
|
2020-09-14 23:03:32 +00:00
|
|
|
options.create_if_missing = true;
|
|
|
|
|
options.writable_file_max_buffer_size = 32768;
|
|
|
|
|
options.listeners.emplace_back(listener);
|
|
|
|
|
options.paranoid_checks = GetParam();
|
|
|
|
|
Status s;
|
|
|
|
|
Random rnd(301);
|
|
|
|
|
|
|
|
|
|
listener->EnableAutoRecovery(true);
|
|
|
|
|
DestroyAndReopen(options);
|
|
|
|
|
|
|
|
|
|
{
|
|
|
|
|
WriteBatch batch;
|
|
|
|
|
|
|
|
|
|
for (auto i = 0; i < 100; ++i) {
|
2020-10-02 23:39:17 +00:00
|
|
|
ASSERT_OK(batch.Put(Key(i), rnd.RandomString(1024)));
|
2020-09-14 23:03:32 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
WriteOptions wopts;
|
|
|
|
|
wopts.sync = true;
|
2020-12-24 00:54:05 +00:00
|
|
|
ASSERT_OK(dbfull()->Write(wopts, &batch));
|
2020-09-14 23:03:32 +00:00
|
|
|
};
|
|
|
|
|
|
|
|
|
|
{
|
|
|
|
|
WriteBatch batch;
|
|
|
|
|
int write_error = 0;
|
|
|
|
|
|
|
|
|
|
for (auto i = 100; i < 199; ++i) {
|
2020-10-02 23:39:17 +00:00
|
|
|
ASSERT_OK(batch.Put(Key(i), rnd.RandomString(1024)));
|
2020-09-14 23:03:32 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
SyncPoint::GetInstance()->SetCallBack(
|
|
|
|
|
"WritableFileWriter::Append:BeforePrepareWrite", [&](void*) {
|
|
|
|
|
write_error++;
|
|
|
|
|
if (write_error > 2) {
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
fault_fs_->SetFilesystemActive(false,
|
|
|
|
|
IOStatus::IOFenced("IO fenced"));
|
2020-09-14 23:03:32 +00:00
|
|
|
}
|
|
|
|
|
});
|
|
|
|
|
SyncPoint::GetInstance()->EnableProcessing();
|
|
|
|
|
WriteOptions wopts;
|
|
|
|
|
wopts.sync = true;
|
|
|
|
|
s = dbfull()->Write(wopts, &batch);
|
|
|
|
|
ASSERT_TRUE(s.IsIOFenced());
|
|
|
|
|
}
|
|
|
|
|
SyncPoint::GetInstance()->DisableProcessing();
|
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566)
Summary:
This PR does a few things:
1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation.
2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed:
- The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated
- The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory).
3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10).
I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged.
Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently.
Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :)
Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566
Reviewed By: zhichao-cao
Differential Revision: D24408980
Pulled By: jay-zhuang
fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
2020-10-27 17:31:34 +00:00
|
|
|
fault_fs_->SetFilesystemActive(true);
|
2020-09-14 23:03:32 +00:00
|
|
|
{
|
|
|
|
|
WriteBatch batch;
|
|
|
|
|
|
|
|
|
|
for (auto i = 0; i < 100; ++i) {
|
2020-10-02 23:39:17 +00:00
|
|
|
ASSERT_OK(batch.Put(Key(i), rnd.RandomString(1024)));
|
2020-09-14 23:03:32 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
WriteOptions wopts;
|
|
|
|
|
wopts.sync = true;
|
|
|
|
|
s = dbfull()->Write(wopts, &batch);
|
|
|
|
|
ASSERT_TRUE(s.IsIOFenced());
|
|
|
|
|
}
|
|
|
|
|
Close();
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
INSTANTIATE_TEST_CASE_P(DBErrorHandlingFSTest, DBErrorHandlingFencingTest,
|
|
|
|
|
::testing::Bool());
|
|
|
|
|
|
2020-02-20 20:07:53 +00:00
|
|
|
} // namespace ROCKSDB_NAMESPACE
|
2018-06-28 19:23:57 +00:00
|
|
|
|
|
|
|
|
int main(int argc, char** argv) {
|
2020-02-20 20:07:53 +00:00
|
|
|
ROCKSDB_NAMESPACE::port::InstallStackTraceHandler();
|
2018-06-28 19:23:57 +00:00
|
|
|
::testing::InitGoogleTest(&argc, argv);
|
|
|
|
|
return RUN_ALL_TESTS();
|
|
|
|
|
}
|