rocksdb/db/obsolete_files_test.cc

316 lines
11 KiB
C++
Raw Normal View History

Fix race condition causing double deletion of ssts Summary: Possible interleaved execution of background compaction thread calling `FindObsoleteFiles (no full scan) / PurgeObsoleteFiles` and user thread calling `FindObsoleteFiles (full scan) / PurgeObsoleteFiles` can lead to race condition on which RocksDB attempts to delete a file twice. The second attempt will fail and return `IO error`. This may occur to other files, but this PR targets sst. Also add a unit test to verify that this PR fixes the issue. The newly added unit test `obsolete_files_test` has a test case for this scenario, implemented in `ObsoleteFilesTest#RaceForObsoleteFileDeletion`. `TestSyncPoint`s are used to coordinate the interleaving the `user_thread` and background compaction thread. They execute as follows ``` timeline user_thread background_compaction thread t1 | FindObsoleteFiles(full_scan=false) t2 | FindObsoleteFiles(full_scan=true) t3 | PurgeObsoleteFiles t4 | PurgeObsoleteFiles V ``` When `user_thread` invokes `FindObsoleteFiles` with full scan, it collects ALL files in RocksDB directory, including the ones that background compaction thread have collected in its job context. Then `user_thread` will see an IO error when trying to delete these files in `PurgeObsoleteFiles` because background compaction thread has already deleted the file in `PurgeObsoleteFiles`. To fix this, we make RocksDB remember which (SST) files have been found by threads after calling `FindObsoleteFiles` (see `DBImpl#files_grabbed_for_purge_`). Therefore, when another thread calls `FindObsoleteFiles` with full scan, it will not collect such files. ajkr could you take a look and comment? Thanks! Closes https://github.com/facebook/rocksdb/pull/3638 Differential Revision: D7384372 Pulled By: riversand963 fbshipit-source-id: 01489516d60012e722ee65a80e1449e589ce26d3
2018-03-28 17:23:31 +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.
#include <algorithm>
#include <cstdlib>
Fix race condition causing double deletion of ssts Summary: Possible interleaved execution of background compaction thread calling `FindObsoleteFiles (no full scan) / PurgeObsoleteFiles` and user thread calling `FindObsoleteFiles (full scan) / PurgeObsoleteFiles` can lead to race condition on which RocksDB attempts to delete a file twice. The second attempt will fail and return `IO error`. This may occur to other files, but this PR targets sst. Also add a unit test to verify that this PR fixes the issue. The newly added unit test `obsolete_files_test` has a test case for this scenario, implemented in `ObsoleteFilesTest#RaceForObsoleteFileDeletion`. `TestSyncPoint`s are used to coordinate the interleaving the `user_thread` and background compaction thread. They execute as follows ``` timeline user_thread background_compaction thread t1 | FindObsoleteFiles(full_scan=false) t2 | FindObsoleteFiles(full_scan=true) t3 | PurgeObsoleteFiles t4 | PurgeObsoleteFiles V ``` When `user_thread` invokes `FindObsoleteFiles` with full scan, it collects ALL files in RocksDB directory, including the ones that background compaction thread have collected in its job context. Then `user_thread` will see an IO error when trying to delete these files in `PurgeObsoleteFiles` because background compaction thread has already deleted the file in `PurgeObsoleteFiles`. To fix this, we make RocksDB remember which (SST) files have been found by threads after calling `FindObsoleteFiles` (see `DBImpl#files_grabbed_for_purge_`). Therefore, when another thread calls `FindObsoleteFiles` with full scan, it will not collect such files. ajkr could you take a look and comment? Thanks! Closes https://github.com/facebook/rocksdb/pull/3638 Differential Revision: D7384372 Pulled By: riversand963 fbshipit-source-id: 01489516d60012e722ee65a80e1449e589ce26d3
2018-03-28 17:23:31 +00:00
#include <map>
#include <string>
#include <vector>
#include "db/db_impl/db_impl.h"
#include "db/db_test_util.h"
Fix race condition causing double deletion of ssts Summary: Possible interleaved execution of background compaction thread calling `FindObsoleteFiles (no full scan) / PurgeObsoleteFiles` and user thread calling `FindObsoleteFiles (full scan) / PurgeObsoleteFiles` can lead to race condition on which RocksDB attempts to delete a file twice. The second attempt will fail and return `IO error`. This may occur to other files, but this PR targets sst. Also add a unit test to verify that this PR fixes the issue. The newly added unit test `obsolete_files_test` has a test case for this scenario, implemented in `ObsoleteFilesTest#RaceForObsoleteFileDeletion`. `TestSyncPoint`s are used to coordinate the interleaving the `user_thread` and background compaction thread. They execute as follows ``` timeline user_thread background_compaction thread t1 | FindObsoleteFiles(full_scan=false) t2 | FindObsoleteFiles(full_scan=true) t3 | PurgeObsoleteFiles t4 | PurgeObsoleteFiles V ``` When `user_thread` invokes `FindObsoleteFiles` with full scan, it collects ALL files in RocksDB directory, including the ones that background compaction thread have collected in its job context. Then `user_thread` will see an IO error when trying to delete these files in `PurgeObsoleteFiles` because background compaction thread has already deleted the file in `PurgeObsoleteFiles`. To fix this, we make RocksDB remember which (SST) files have been found by threads after calling `FindObsoleteFiles` (see `DBImpl#files_grabbed_for_purge_`). Therefore, when another thread calls `FindObsoleteFiles` with full scan, it will not collect such files. ajkr could you take a look and comment? Thanks! Closes https://github.com/facebook/rocksdb/pull/3638 Differential Revision: D7384372 Pulled By: riversand963 fbshipit-source-id: 01489516d60012e722ee65a80e1449e589ce26d3
2018-03-28 17:23:31 +00:00
#include "db/version_set.h"
#include "db/write_batch_internal.h"
#include "file/filename.h"
#include "port/stack_trace.h"
Fix race condition causing double deletion of ssts Summary: Possible interleaved execution of background compaction thread calling `FindObsoleteFiles (no full scan) / PurgeObsoleteFiles` and user thread calling `FindObsoleteFiles (full scan) / PurgeObsoleteFiles` can lead to race condition on which RocksDB attempts to delete a file twice. The second attempt will fail and return `IO error`. This may occur to other files, but this PR targets sst. Also add a unit test to verify that this PR fixes the issue. The newly added unit test `obsolete_files_test` has a test case for this scenario, implemented in `ObsoleteFilesTest#RaceForObsoleteFileDeletion`. `TestSyncPoint`s are used to coordinate the interleaving the `user_thread` and background compaction thread. They execute as follows ``` timeline user_thread background_compaction thread t1 | FindObsoleteFiles(full_scan=false) t2 | FindObsoleteFiles(full_scan=true) t3 | PurgeObsoleteFiles t4 | PurgeObsoleteFiles V ``` When `user_thread` invokes `FindObsoleteFiles` with full scan, it collects ALL files in RocksDB directory, including the ones that background compaction thread have collected in its job context. Then `user_thread` will see an IO error when trying to delete these files in `PurgeObsoleteFiles` because background compaction thread has already deleted the file in `PurgeObsoleteFiles`. To fix this, we make RocksDB remember which (SST) files have been found by threads after calling `FindObsoleteFiles` (see `DBImpl#files_grabbed_for_purge_`). Therefore, when another thread calls `FindObsoleteFiles` with full scan, it will not collect such files. ajkr could you take a look and comment? Thanks! Closes https://github.com/facebook/rocksdb/pull/3638 Differential Revision: D7384372 Pulled By: riversand963 fbshipit-source-id: 01489516d60012e722ee65a80e1449e589ce26d3
2018-03-28 17:23:31 +00:00
#include "rocksdb/db.h"
#include "rocksdb/env.h"
#include "rocksdb/transaction_log.h"
#include "test_util/sync_point.h"
#include "test_util/testharness.h"
#include "test_util/testutil.h"
#include "util/string_util.h"
Fix race condition causing double deletion of ssts Summary: Possible interleaved execution of background compaction thread calling `FindObsoleteFiles (no full scan) / PurgeObsoleteFiles` and user thread calling `FindObsoleteFiles (full scan) / PurgeObsoleteFiles` can lead to race condition on which RocksDB attempts to delete a file twice. The second attempt will fail and return `IO error`. This may occur to other files, but this PR targets sst. Also add a unit test to verify that this PR fixes the issue. The newly added unit test `obsolete_files_test` has a test case for this scenario, implemented in `ObsoleteFilesTest#RaceForObsoleteFileDeletion`. `TestSyncPoint`s are used to coordinate the interleaving the `user_thread` and background compaction thread. They execute as follows ``` timeline user_thread background_compaction thread t1 | FindObsoleteFiles(full_scan=false) t2 | FindObsoleteFiles(full_scan=true) t3 | PurgeObsoleteFiles t4 | PurgeObsoleteFiles V ``` When `user_thread` invokes `FindObsoleteFiles` with full scan, it collects ALL files in RocksDB directory, including the ones that background compaction thread have collected in its job context. Then `user_thread` will see an IO error when trying to delete these files in `PurgeObsoleteFiles` because background compaction thread has already deleted the file in `PurgeObsoleteFiles`. To fix this, we make RocksDB remember which (SST) files have been found by threads after calling `FindObsoleteFiles` (see `DBImpl#files_grabbed_for_purge_`). Therefore, when another thread calls `FindObsoleteFiles` with full scan, it will not collect such files. ajkr could you take a look and comment? Thanks! Closes https://github.com/facebook/rocksdb/pull/3638 Differential Revision: D7384372 Pulled By: riversand963 fbshipit-source-id: 01489516d60012e722ee65a80e1449e589ce26d3
2018-03-28 17:23:31 +00:00
namespace ROCKSDB_NAMESPACE {
Fix race condition causing double deletion of ssts Summary: Possible interleaved execution of background compaction thread calling `FindObsoleteFiles (no full scan) / PurgeObsoleteFiles` and user thread calling `FindObsoleteFiles (full scan) / PurgeObsoleteFiles` can lead to race condition on which RocksDB attempts to delete a file twice. The second attempt will fail and return `IO error`. This may occur to other files, but this PR targets sst. Also add a unit test to verify that this PR fixes the issue. The newly added unit test `obsolete_files_test` has a test case for this scenario, implemented in `ObsoleteFilesTest#RaceForObsoleteFileDeletion`. `TestSyncPoint`s are used to coordinate the interleaving the `user_thread` and background compaction thread. They execute as follows ``` timeline user_thread background_compaction thread t1 | FindObsoleteFiles(full_scan=false) t2 | FindObsoleteFiles(full_scan=true) t3 | PurgeObsoleteFiles t4 | PurgeObsoleteFiles V ``` When `user_thread` invokes `FindObsoleteFiles` with full scan, it collects ALL files in RocksDB directory, including the ones that background compaction thread have collected in its job context. Then `user_thread` will see an IO error when trying to delete these files in `PurgeObsoleteFiles` because background compaction thread has already deleted the file in `PurgeObsoleteFiles`. To fix this, we make RocksDB remember which (SST) files have been found by threads after calling `FindObsoleteFiles` (see `DBImpl#files_grabbed_for_purge_`). Therefore, when another thread calls `FindObsoleteFiles` with full scan, it will not collect such files. ajkr could you take a look and comment? Thanks! Closes https://github.com/facebook/rocksdb/pull/3638 Differential Revision: D7384372 Pulled By: riversand963 fbshipit-source-id: 01489516d60012e722ee65a80e1449e589ce26d3
2018-03-28 17:23:31 +00:00
class ObsoleteFilesTest : public DBTestBase {
Fix race condition causing double deletion of ssts Summary: Possible interleaved execution of background compaction thread calling `FindObsoleteFiles (no full scan) / PurgeObsoleteFiles` and user thread calling `FindObsoleteFiles (full scan) / PurgeObsoleteFiles` can lead to race condition on which RocksDB attempts to delete a file twice. The second attempt will fail and return `IO error`. This may occur to other files, but this PR targets sst. Also add a unit test to verify that this PR fixes the issue. The newly added unit test `obsolete_files_test` has a test case for this scenario, implemented in `ObsoleteFilesTest#RaceForObsoleteFileDeletion`. `TestSyncPoint`s are used to coordinate the interleaving the `user_thread` and background compaction thread. They execute as follows ``` timeline user_thread background_compaction thread t1 | FindObsoleteFiles(full_scan=false) t2 | FindObsoleteFiles(full_scan=true) t3 | PurgeObsoleteFiles t4 | PurgeObsoleteFiles V ``` When `user_thread` invokes `FindObsoleteFiles` with full scan, it collects ALL files in RocksDB directory, including the ones that background compaction thread have collected in its job context. Then `user_thread` will see an IO error when trying to delete these files in `PurgeObsoleteFiles` because background compaction thread has already deleted the file in `PurgeObsoleteFiles`. To fix this, we make RocksDB remember which (SST) files have been found by threads after calling `FindObsoleteFiles` (see `DBImpl#files_grabbed_for_purge_`). Therefore, when another thread calls `FindObsoleteFiles` with full scan, it will not collect such files. ajkr could you take a look and comment? Thanks! Closes https://github.com/facebook/rocksdb/pull/3638 Differential Revision: D7384372 Pulled By: riversand963 fbshipit-source-id: 01489516d60012e722ee65a80e1449e589ce26d3
2018-03-28 17:23:31 +00:00
public:
ObsoleteFilesTest()
: DBTestBase("obsolete_files_test", /*env_do_fsync=*/true),
wal_dir_(dbname_ + "/wal_files") {}
Fix race condition causing double deletion of ssts Summary: Possible interleaved execution of background compaction thread calling `FindObsoleteFiles (no full scan) / PurgeObsoleteFiles` and user thread calling `FindObsoleteFiles (full scan) / PurgeObsoleteFiles` can lead to race condition on which RocksDB attempts to delete a file twice. The second attempt will fail and return `IO error`. This may occur to other files, but this PR targets sst. Also add a unit test to verify that this PR fixes the issue. The newly added unit test `obsolete_files_test` has a test case for this scenario, implemented in `ObsoleteFilesTest#RaceForObsoleteFileDeletion`. `TestSyncPoint`s are used to coordinate the interleaving the `user_thread` and background compaction thread. They execute as follows ``` timeline user_thread background_compaction thread t1 | FindObsoleteFiles(full_scan=false) t2 | FindObsoleteFiles(full_scan=true) t3 | PurgeObsoleteFiles t4 | PurgeObsoleteFiles V ``` When `user_thread` invokes `FindObsoleteFiles` with full scan, it collects ALL files in RocksDB directory, including the ones that background compaction thread have collected in its job context. Then `user_thread` will see an IO error when trying to delete these files in `PurgeObsoleteFiles` because background compaction thread has already deleted the file in `PurgeObsoleteFiles`. To fix this, we make RocksDB remember which (SST) files have been found by threads after calling `FindObsoleteFiles` (see `DBImpl#files_grabbed_for_purge_`). Therefore, when another thread calls `FindObsoleteFiles` with full scan, it will not collect such files. ajkr could you take a look and comment? Thanks! Closes https://github.com/facebook/rocksdb/pull/3638 Differential Revision: D7384372 Pulled By: riversand963 fbshipit-source-id: 01489516d60012e722ee65a80e1449e589ce26d3
2018-03-28 17:23:31 +00:00
void AddKeys(int numkeys, int startkey) {
WriteOptions options;
options.sync = false;
for (int i = startkey; i < (numkeys + startkey); i++) {
std::string temp = std::to_string(i);
Fix race condition causing double deletion of ssts Summary: Possible interleaved execution of background compaction thread calling `FindObsoleteFiles (no full scan) / PurgeObsoleteFiles` and user thread calling `FindObsoleteFiles (full scan) / PurgeObsoleteFiles` can lead to race condition on which RocksDB attempts to delete a file twice. The second attempt will fail and return `IO error`. This may occur to other files, but this PR targets sst. Also add a unit test to verify that this PR fixes the issue. The newly added unit test `obsolete_files_test` has a test case for this scenario, implemented in `ObsoleteFilesTest#RaceForObsoleteFileDeletion`. `TestSyncPoint`s are used to coordinate the interleaving the `user_thread` and background compaction thread. They execute as follows ``` timeline user_thread background_compaction thread t1 | FindObsoleteFiles(full_scan=false) t2 | FindObsoleteFiles(full_scan=true) t3 | PurgeObsoleteFiles t4 | PurgeObsoleteFiles V ``` When `user_thread` invokes `FindObsoleteFiles` with full scan, it collects ALL files in RocksDB directory, including the ones that background compaction thread have collected in its job context. Then `user_thread` will see an IO error when trying to delete these files in `PurgeObsoleteFiles` because background compaction thread has already deleted the file in `PurgeObsoleteFiles`. To fix this, we make RocksDB remember which (SST) files have been found by threads after calling `FindObsoleteFiles` (see `DBImpl#files_grabbed_for_purge_`). Therefore, when another thread calls `FindObsoleteFiles` with full scan, it will not collect such files. ajkr could you take a look and comment? Thanks! Closes https://github.com/facebook/rocksdb/pull/3638 Differential Revision: D7384372 Pulled By: riversand963 fbshipit-source-id: 01489516d60012e722ee65a80e1449e589ce26d3
2018-03-28 17:23:31 +00:00
Slice key(temp);
Slice value(temp);
ASSERT_OK(db_->Put(options, key, value));
}
}
void createLevel0Files(int numFiles, int numKeysPerFile) {
int startKey = 0;
for (int i = 0; i < numFiles; i++) {
AddKeys(numKeysPerFile, startKey);
startKey += numKeysPerFile;
ASSERT_OK(dbfull()->TEST_FlushMemTable());
ASSERT_OK(
dbfull()->TEST_WaitForCompact()); // wait for background flush (flush
// is also a kind of compaction).
Fix race condition causing double deletion of ssts Summary: Possible interleaved execution of background compaction thread calling `FindObsoleteFiles (no full scan) / PurgeObsoleteFiles` and user thread calling `FindObsoleteFiles (full scan) / PurgeObsoleteFiles` can lead to race condition on which RocksDB attempts to delete a file twice. The second attempt will fail and return `IO error`. This may occur to other files, but this PR targets sst. Also add a unit test to verify that this PR fixes the issue. The newly added unit test `obsolete_files_test` has a test case for this scenario, implemented in `ObsoleteFilesTest#RaceForObsoleteFileDeletion`. `TestSyncPoint`s are used to coordinate the interleaving the `user_thread` and background compaction thread. They execute as follows ``` timeline user_thread background_compaction thread t1 | FindObsoleteFiles(full_scan=false) t2 | FindObsoleteFiles(full_scan=true) t3 | PurgeObsoleteFiles t4 | PurgeObsoleteFiles V ``` When `user_thread` invokes `FindObsoleteFiles` with full scan, it collects ALL files in RocksDB directory, including the ones that background compaction thread have collected in its job context. Then `user_thread` will see an IO error when trying to delete these files in `PurgeObsoleteFiles` because background compaction thread has already deleted the file in `PurgeObsoleteFiles`. To fix this, we make RocksDB remember which (SST) files have been found by threads after calling `FindObsoleteFiles` (see `DBImpl#files_grabbed_for_purge_`). Therefore, when another thread calls `FindObsoleteFiles` with full scan, it will not collect such files. ajkr could you take a look and comment? Thanks! Closes https://github.com/facebook/rocksdb/pull/3638 Differential Revision: D7384372 Pulled By: riversand963 fbshipit-source-id: 01489516d60012e722ee65a80e1449e589ce26d3
2018-03-28 17:23:31 +00:00
}
}
void CheckFileTypeCounts(const std::string& dir, int required_log,
int required_sst, int required_manifest) {
Fix race condition causing double deletion of ssts Summary: Possible interleaved execution of background compaction thread calling `FindObsoleteFiles (no full scan) / PurgeObsoleteFiles` and user thread calling `FindObsoleteFiles (full scan) / PurgeObsoleteFiles` can lead to race condition on which RocksDB attempts to delete a file twice. The second attempt will fail and return `IO error`. This may occur to other files, but this PR targets sst. Also add a unit test to verify that this PR fixes the issue. The newly added unit test `obsolete_files_test` has a test case for this scenario, implemented in `ObsoleteFilesTest#RaceForObsoleteFileDeletion`. `TestSyncPoint`s are used to coordinate the interleaving the `user_thread` and background compaction thread. They execute as follows ``` timeline user_thread background_compaction thread t1 | FindObsoleteFiles(full_scan=false) t2 | FindObsoleteFiles(full_scan=true) t3 | PurgeObsoleteFiles t4 | PurgeObsoleteFiles V ``` When `user_thread` invokes `FindObsoleteFiles` with full scan, it collects ALL files in RocksDB directory, including the ones that background compaction thread have collected in its job context. Then `user_thread` will see an IO error when trying to delete these files in `PurgeObsoleteFiles` because background compaction thread has already deleted the file in `PurgeObsoleteFiles`. To fix this, we make RocksDB remember which (SST) files have been found by threads after calling `FindObsoleteFiles` (see `DBImpl#files_grabbed_for_purge_`). Therefore, when another thread calls `FindObsoleteFiles` with full scan, it will not collect such files. ajkr could you take a look and comment? Thanks! Closes https://github.com/facebook/rocksdb/pull/3638 Differential Revision: D7384372 Pulled By: riversand963 fbshipit-source-id: 01489516d60012e722ee65a80e1449e589ce26d3
2018-03-28 17:23:31 +00:00
std::vector<std::string> filenames;
ASSERT_OK(env_->GetChildren(dir, &filenames));
Fix race condition causing double deletion of ssts Summary: Possible interleaved execution of background compaction thread calling `FindObsoleteFiles (no full scan) / PurgeObsoleteFiles` and user thread calling `FindObsoleteFiles (full scan) / PurgeObsoleteFiles` can lead to race condition on which RocksDB attempts to delete a file twice. The second attempt will fail and return `IO error`. This may occur to other files, but this PR targets sst. Also add a unit test to verify that this PR fixes the issue. The newly added unit test `obsolete_files_test` has a test case for this scenario, implemented in `ObsoleteFilesTest#RaceForObsoleteFileDeletion`. `TestSyncPoint`s are used to coordinate the interleaving the `user_thread` and background compaction thread. They execute as follows ``` timeline user_thread background_compaction thread t1 | FindObsoleteFiles(full_scan=false) t2 | FindObsoleteFiles(full_scan=true) t3 | PurgeObsoleteFiles t4 | PurgeObsoleteFiles V ``` When `user_thread` invokes `FindObsoleteFiles` with full scan, it collects ALL files in RocksDB directory, including the ones that background compaction thread have collected in its job context. Then `user_thread` will see an IO error when trying to delete these files in `PurgeObsoleteFiles` because background compaction thread has already deleted the file in `PurgeObsoleteFiles`. To fix this, we make RocksDB remember which (SST) files have been found by threads after calling `FindObsoleteFiles` (see `DBImpl#files_grabbed_for_purge_`). Therefore, when another thread calls `FindObsoleteFiles` with full scan, it will not collect such files. ajkr could you take a look and comment? Thanks! Closes https://github.com/facebook/rocksdb/pull/3638 Differential Revision: D7384372 Pulled By: riversand963 fbshipit-source-id: 01489516d60012e722ee65a80e1449e589ce26d3
2018-03-28 17:23:31 +00:00
int log_cnt = 0;
int sst_cnt = 0;
int manifest_cnt = 0;
for (const auto& file : filenames) {
Fix race condition causing double deletion of ssts Summary: Possible interleaved execution of background compaction thread calling `FindObsoleteFiles (no full scan) / PurgeObsoleteFiles` and user thread calling `FindObsoleteFiles (full scan) / PurgeObsoleteFiles` can lead to race condition on which RocksDB attempts to delete a file twice. The second attempt will fail and return `IO error`. This may occur to other files, but this PR targets sst. Also add a unit test to verify that this PR fixes the issue. The newly added unit test `obsolete_files_test` has a test case for this scenario, implemented in `ObsoleteFilesTest#RaceForObsoleteFileDeletion`. `TestSyncPoint`s are used to coordinate the interleaving the `user_thread` and background compaction thread. They execute as follows ``` timeline user_thread background_compaction thread t1 | FindObsoleteFiles(full_scan=false) t2 | FindObsoleteFiles(full_scan=true) t3 | PurgeObsoleteFiles t4 | PurgeObsoleteFiles V ``` When `user_thread` invokes `FindObsoleteFiles` with full scan, it collects ALL files in RocksDB directory, including the ones that background compaction thread have collected in its job context. Then `user_thread` will see an IO error when trying to delete these files in `PurgeObsoleteFiles` because background compaction thread has already deleted the file in `PurgeObsoleteFiles`. To fix this, we make RocksDB remember which (SST) files have been found by threads after calling `FindObsoleteFiles` (see `DBImpl#files_grabbed_for_purge_`). Therefore, when another thread calls `FindObsoleteFiles` with full scan, it will not collect such files. ajkr could you take a look and comment? Thanks! Closes https://github.com/facebook/rocksdb/pull/3638 Differential Revision: D7384372 Pulled By: riversand963 fbshipit-source-id: 01489516d60012e722ee65a80e1449e589ce26d3
2018-03-28 17:23:31 +00:00
uint64_t number;
FileType type;
if (ParseFileName(file, &number, &type)) {
log_cnt += (type == kWalFile);
Fix race condition causing double deletion of ssts Summary: Possible interleaved execution of background compaction thread calling `FindObsoleteFiles (no full scan) / PurgeObsoleteFiles` and user thread calling `FindObsoleteFiles (full scan) / PurgeObsoleteFiles` can lead to race condition on which RocksDB attempts to delete a file twice. The second attempt will fail and return `IO error`. This may occur to other files, but this PR targets sst. Also add a unit test to verify that this PR fixes the issue. The newly added unit test `obsolete_files_test` has a test case for this scenario, implemented in `ObsoleteFilesTest#RaceForObsoleteFileDeletion`. `TestSyncPoint`s are used to coordinate the interleaving the `user_thread` and background compaction thread. They execute as follows ``` timeline user_thread background_compaction thread t1 | FindObsoleteFiles(full_scan=false) t2 | FindObsoleteFiles(full_scan=true) t3 | PurgeObsoleteFiles t4 | PurgeObsoleteFiles V ``` When `user_thread` invokes `FindObsoleteFiles` with full scan, it collects ALL files in RocksDB directory, including the ones that background compaction thread have collected in its job context. Then `user_thread` will see an IO error when trying to delete these files in `PurgeObsoleteFiles` because background compaction thread has already deleted the file in `PurgeObsoleteFiles`. To fix this, we make RocksDB remember which (SST) files have been found by threads after calling `FindObsoleteFiles` (see `DBImpl#files_grabbed_for_purge_`). Therefore, when another thread calls `FindObsoleteFiles` with full scan, it will not collect such files. ajkr could you take a look and comment? Thanks! Closes https://github.com/facebook/rocksdb/pull/3638 Differential Revision: D7384372 Pulled By: riversand963 fbshipit-source-id: 01489516d60012e722ee65a80e1449e589ce26d3
2018-03-28 17:23:31 +00:00
sst_cnt += (type == kTableFile);
manifest_cnt += (type == kDescriptorFile);
}
}
ASSERT_EQ(required_log, log_cnt);
ASSERT_EQ(required_sst, sst_cnt);
ASSERT_EQ(required_manifest, manifest_cnt);
}
void ReopenDB() {
Options options = CurrentOptions();
// Trigger compaction when the number of level 0 files reaches 2.
options.create_if_missing = true;
options.level0_file_num_compaction_trigger = 2;
options.disable_auto_compactions = false;
options.delete_obsolete_files_period_micros = 0; // always do full purge
options.enable_thread_tracking = true;
options.write_buffer_size = 1024 * 1024 * 1000;
options.target_file_size_base = 1024 * 1024 * 1000;
options.max_bytes_for_level_base = 1024 * 1024 * 1000;
options.WAL_ttl_seconds = 300; // Used to test log files
options.WAL_size_limit_MB = 1024; // Used to test log files
options.wal_dir = wal_dir_;
// Note: the following prevents an otherwise harmless data race between the
// test setup code (AddBlobFile) in ObsoleteFilesTest.BlobFiles and the
// periodic stat dumping thread.
options.stats_dump_period_sec = 0;
Destroy(options);
Reopen(options);
}
const std::string wal_dir_;
Fix race condition causing double deletion of ssts Summary: Possible interleaved execution of background compaction thread calling `FindObsoleteFiles (no full scan) / PurgeObsoleteFiles` and user thread calling `FindObsoleteFiles (full scan) / PurgeObsoleteFiles` can lead to race condition on which RocksDB attempts to delete a file twice. The second attempt will fail and return `IO error`. This may occur to other files, but this PR targets sst. Also add a unit test to verify that this PR fixes the issue. The newly added unit test `obsolete_files_test` has a test case for this scenario, implemented in `ObsoleteFilesTest#RaceForObsoleteFileDeletion`. `TestSyncPoint`s are used to coordinate the interleaving the `user_thread` and background compaction thread. They execute as follows ``` timeline user_thread background_compaction thread t1 | FindObsoleteFiles(full_scan=false) t2 | FindObsoleteFiles(full_scan=true) t3 | PurgeObsoleteFiles t4 | PurgeObsoleteFiles V ``` When `user_thread` invokes `FindObsoleteFiles` with full scan, it collects ALL files in RocksDB directory, including the ones that background compaction thread have collected in its job context. Then `user_thread` will see an IO error when trying to delete these files in `PurgeObsoleteFiles` because background compaction thread has already deleted the file in `PurgeObsoleteFiles`. To fix this, we make RocksDB remember which (SST) files have been found by threads after calling `FindObsoleteFiles` (see `DBImpl#files_grabbed_for_purge_`). Therefore, when another thread calls `FindObsoleteFiles` with full scan, it will not collect such files. ajkr could you take a look and comment? Thanks! Closes https://github.com/facebook/rocksdb/pull/3638 Differential Revision: D7384372 Pulled By: riversand963 fbshipit-source-id: 01489516d60012e722ee65a80e1449e589ce26d3
2018-03-28 17:23:31 +00:00
};
TEST_F(ObsoleteFilesTest, RaceForObsoleteFileDeletion) {
ReopenDB();
SyncPoint::GetInstance()->DisableProcessing();
Fix race condition causing double deletion of ssts Summary: Possible interleaved execution of background compaction thread calling `FindObsoleteFiles (no full scan) / PurgeObsoleteFiles` and user thread calling `FindObsoleteFiles (full scan) / PurgeObsoleteFiles` can lead to race condition on which RocksDB attempts to delete a file twice. The second attempt will fail and return `IO error`. This may occur to other files, but this PR targets sst. Also add a unit test to verify that this PR fixes the issue. The newly added unit test `obsolete_files_test` has a test case for this scenario, implemented in `ObsoleteFilesTest#RaceForObsoleteFileDeletion`. `TestSyncPoint`s are used to coordinate the interleaving the `user_thread` and background compaction thread. They execute as follows ``` timeline user_thread background_compaction thread t1 | FindObsoleteFiles(full_scan=false) t2 | FindObsoleteFiles(full_scan=true) t3 | PurgeObsoleteFiles t4 | PurgeObsoleteFiles V ``` When `user_thread` invokes `FindObsoleteFiles` with full scan, it collects ALL files in RocksDB directory, including the ones that background compaction thread have collected in its job context. Then `user_thread` will see an IO error when trying to delete these files in `PurgeObsoleteFiles` because background compaction thread has already deleted the file in `PurgeObsoleteFiles`. To fix this, we make RocksDB remember which (SST) files have been found by threads after calling `FindObsoleteFiles` (see `DBImpl#files_grabbed_for_purge_`). Therefore, when another thread calls `FindObsoleteFiles` with full scan, it will not collect such files. ajkr could you take a look and comment? Thanks! Closes https://github.com/facebook/rocksdb/pull/3638 Differential Revision: D7384372 Pulled By: riversand963 fbshipit-source-id: 01489516d60012e722ee65a80e1449e589ce26d3
2018-03-28 17:23:31 +00:00
SyncPoint::GetInstance()->LoadDependency({
{"DBImpl::BackgroundCallCompaction:FoundObsoleteFiles",
"ObsoleteFilesTest::RaceForObsoleteFileDeletion:1"},
{"DBImpl::BackgroundCallCompaction:PurgedObsoleteFiles",
"ObsoleteFilesTest::RaceForObsoleteFileDeletion:2"},
});
Fix race condition causing double deletion of ssts Summary: Possible interleaved execution of background compaction thread calling `FindObsoleteFiles (no full scan) / PurgeObsoleteFiles` and user thread calling `FindObsoleteFiles (full scan) / PurgeObsoleteFiles` can lead to race condition on which RocksDB attempts to delete a file twice. The second attempt will fail and return `IO error`. This may occur to other files, but this PR targets sst. Also add a unit test to verify that this PR fixes the issue. The newly added unit test `obsolete_files_test` has a test case for this scenario, implemented in `ObsoleteFilesTest#RaceForObsoleteFileDeletion`. `TestSyncPoint`s are used to coordinate the interleaving the `user_thread` and background compaction thread. They execute as follows ``` timeline user_thread background_compaction thread t1 | FindObsoleteFiles(full_scan=false) t2 | FindObsoleteFiles(full_scan=true) t3 | PurgeObsoleteFiles t4 | PurgeObsoleteFiles V ``` When `user_thread` invokes `FindObsoleteFiles` with full scan, it collects ALL files in RocksDB directory, including the ones that background compaction thread have collected in its job context. Then `user_thread` will see an IO error when trying to delete these files in `PurgeObsoleteFiles` because background compaction thread has already deleted the file in `PurgeObsoleteFiles`. To fix this, we make RocksDB remember which (SST) files have been found by threads after calling `FindObsoleteFiles` (see `DBImpl#files_grabbed_for_purge_`). Therefore, when another thread calls `FindObsoleteFiles` with full scan, it will not collect such files. ajkr could you take a look and comment? Thanks! Closes https://github.com/facebook/rocksdb/pull/3638 Differential Revision: D7384372 Pulled By: riversand963 fbshipit-source-id: 01489516d60012e722ee65a80e1449e589ce26d3
2018-03-28 17:23:31 +00:00
SyncPoint::GetInstance()->SetCallBack(
"DBImpl::DeleteObsoleteFileImpl:AfterDeletion", [&](void* arg) {
Prefer static_cast in place of most reinterpret_cast (#12308) Summary: The following are risks associated with pointer-to-pointer reinterpret_cast: * Can produce the "wrong result" (crash or memory corruption). IIRC, in theory this can happen for any up-cast or down-cast for a non-standard-layout type, though in practice would only happen for multiple inheritance cases (where the base class pointer might be "inside" the derived object). We don't use multiple inheritance a lot, but we do. * Can mask useful compiler errors upon code change, including converting between unrelated pointer types that you are expecting to be related, and converting between pointer and scalar types unintentionally. I can only think of some obscure cases where static_cast could be troublesome when it compiles as a replacement: * Going through `void*` could plausibly cause unnecessary or broken pointer arithmetic. Suppose we have `struct Derived: public Base1, public Base2`. If we have `Derived*` -> `void*` -> `Base2*` -> `Derived*` through reinterpret casts, this could plausibly work (though technical UB) assuming the `Base2*` is not dereferenced. Changing to static cast could introduce breaking pointer arithmetic. * Unnecessary (but safe) pointer arithmetic could arise in a case like `Derived*` -> `Base2*` -> `Derived*` where before the Base2 pointer might not have been dereferenced. This could potentially affect performance. With some light scripting, I tried replacing pointer-to-pointer reinterpret_casts with static_cast and kept the cases that still compile. Most occurrences of reinterpret_cast have successfully been changed (except for java/ and third-party/). 294 changed, 257 remain. A couple of related interventions included here: * Previously Cache::Handle was not actually derived from in the implementations and just used as a `void*` stand-in with reinterpret_cast. Now there is a relationship to allow static_cast. In theory, this could introduce pointer arithmetic (as described above) but is unlikely without multiple inheritance AND non-empty Cache::Handle. * Remove some unnecessary casts to void* as this is allowed to be implicit (for better or worse). Most of the remaining reinterpret_casts are for converting to/from raw bytes of objects. We could consider better idioms for these patterns in follow-up work. I wish there were a way to implement a template variant of static_cast that would only compile if no pointer arithmetic is generated, but best I can tell, this is not possible. AFAIK the best you could do is a dynamic check that the void* conversion after the static cast is unchanged. Pull Request resolved: https://github.com/facebook/rocksdb/pull/12308 Test Plan: existing tests, CI Reviewed By: ltamasi Differential Revision: D53204947 Pulled By: pdillinger fbshipit-source-id: 9de23e618263b0d5b9820f4e15966876888a16e2
2024-02-07 18:44:11 +00:00
Status* p_status = static_cast<Status*>(arg);
ASSERT_OK(*p_status);
Fix race condition causing double deletion of ssts Summary: Possible interleaved execution of background compaction thread calling `FindObsoleteFiles (no full scan) / PurgeObsoleteFiles` and user thread calling `FindObsoleteFiles (full scan) / PurgeObsoleteFiles` can lead to race condition on which RocksDB attempts to delete a file twice. The second attempt will fail and return `IO error`. This may occur to other files, but this PR targets sst. Also add a unit test to verify that this PR fixes the issue. The newly added unit test `obsolete_files_test` has a test case for this scenario, implemented in `ObsoleteFilesTest#RaceForObsoleteFileDeletion`. `TestSyncPoint`s are used to coordinate the interleaving the `user_thread` and background compaction thread. They execute as follows ``` timeline user_thread background_compaction thread t1 | FindObsoleteFiles(full_scan=false) t2 | FindObsoleteFiles(full_scan=true) t3 | PurgeObsoleteFiles t4 | PurgeObsoleteFiles V ``` When `user_thread` invokes `FindObsoleteFiles` with full scan, it collects ALL files in RocksDB directory, including the ones that background compaction thread have collected in its job context. Then `user_thread` will see an IO error when trying to delete these files in `PurgeObsoleteFiles` because background compaction thread has already deleted the file in `PurgeObsoleteFiles`. To fix this, we make RocksDB remember which (SST) files have been found by threads after calling `FindObsoleteFiles` (see `DBImpl#files_grabbed_for_purge_`). Therefore, when another thread calls `FindObsoleteFiles` with full scan, it will not collect such files. ajkr could you take a look and comment? Thanks! Closes https://github.com/facebook/rocksdb/pull/3638 Differential Revision: D7384372 Pulled By: riversand963 fbshipit-source-id: 01489516d60012e722ee65a80e1449e589ce26d3
2018-03-28 17:23:31 +00:00
});
SyncPoint::GetInstance()->SetCallBack(
"DBImpl::CloseHelper:PendingPurgeFinished", [&](void* arg) {
std::unordered_set<uint64_t>* files_grabbed_for_purge_ptr =
reinterpret_cast<std::unordered_set<uint64_t>*>(arg);
Fix race condition causing double deletion of ssts Summary: Possible interleaved execution of background compaction thread calling `FindObsoleteFiles (no full scan) / PurgeObsoleteFiles` and user thread calling `FindObsoleteFiles (full scan) / PurgeObsoleteFiles` can lead to race condition on which RocksDB attempts to delete a file twice. The second attempt will fail and return `IO error`. This may occur to other files, but this PR targets sst. Also add a unit test to verify that this PR fixes the issue. The newly added unit test `obsolete_files_test` has a test case for this scenario, implemented in `ObsoleteFilesTest#RaceForObsoleteFileDeletion`. `TestSyncPoint`s are used to coordinate the interleaving the `user_thread` and background compaction thread. They execute as follows ``` timeline user_thread background_compaction thread t1 | FindObsoleteFiles(full_scan=false) t2 | FindObsoleteFiles(full_scan=true) t3 | PurgeObsoleteFiles t4 | PurgeObsoleteFiles V ``` When `user_thread` invokes `FindObsoleteFiles` with full scan, it collects ALL files in RocksDB directory, including the ones that background compaction thread have collected in its job context. Then `user_thread` will see an IO error when trying to delete these files in `PurgeObsoleteFiles` because background compaction thread has already deleted the file in `PurgeObsoleteFiles`. To fix this, we make RocksDB remember which (SST) files have been found by threads after calling `FindObsoleteFiles` (see `DBImpl#files_grabbed_for_purge_`). Therefore, when another thread calls `FindObsoleteFiles` with full scan, it will not collect such files. ajkr could you take a look and comment? Thanks! Closes https://github.com/facebook/rocksdb/pull/3638 Differential Revision: D7384372 Pulled By: riversand963 fbshipit-source-id: 01489516d60012e722ee65a80e1449e589ce26d3
2018-03-28 17:23:31 +00:00
ASSERT_TRUE(files_grabbed_for_purge_ptr->empty());
});
SyncPoint::GetInstance()->EnableProcessing();
createLevel0Files(2, 50000);
CheckFileTypeCounts(wal_dir_, 1, 0, 0);
port::Thread user_thread([this]() {
Fix race condition causing double deletion of ssts Summary: Possible interleaved execution of background compaction thread calling `FindObsoleteFiles (no full scan) / PurgeObsoleteFiles` and user thread calling `FindObsoleteFiles (full scan) / PurgeObsoleteFiles` can lead to race condition on which RocksDB attempts to delete a file twice. The second attempt will fail and return `IO error`. This may occur to other files, but this PR targets sst. Also add a unit test to verify that this PR fixes the issue. The newly added unit test `obsolete_files_test` has a test case for this scenario, implemented in `ObsoleteFilesTest#RaceForObsoleteFileDeletion`. `TestSyncPoint`s are used to coordinate the interleaving the `user_thread` and background compaction thread. They execute as follows ``` timeline user_thread background_compaction thread t1 | FindObsoleteFiles(full_scan=false) t2 | FindObsoleteFiles(full_scan=true) t3 | PurgeObsoleteFiles t4 | PurgeObsoleteFiles V ``` When `user_thread` invokes `FindObsoleteFiles` with full scan, it collects ALL files in RocksDB directory, including the ones that background compaction thread have collected in its job context. Then `user_thread` will see an IO error when trying to delete these files in `PurgeObsoleteFiles` because background compaction thread has already deleted the file in `PurgeObsoleteFiles`. To fix this, we make RocksDB remember which (SST) files have been found by threads after calling `FindObsoleteFiles` (see `DBImpl#files_grabbed_for_purge_`). Therefore, when another thread calls `FindObsoleteFiles` with full scan, it will not collect such files. ajkr could you take a look and comment? Thanks! Closes https://github.com/facebook/rocksdb/pull/3638 Differential Revision: D7384372 Pulled By: riversand963 fbshipit-source-id: 01489516d60012e722ee65a80e1449e589ce26d3
2018-03-28 17:23:31 +00:00
JobContext jobCxt(0);
TEST_SYNC_POINT("ObsoleteFilesTest::RaceForObsoleteFileDeletion:1");
dbfull()->TEST_LockMutex();
dbfull()->FindObsoleteFiles(&jobCxt, true /* force=true */,
false /* no_full_scan=false */);
dbfull()->TEST_UnlockMutex();
Fix race condition causing double deletion of ssts Summary: Possible interleaved execution of background compaction thread calling `FindObsoleteFiles (no full scan) / PurgeObsoleteFiles` and user thread calling `FindObsoleteFiles (full scan) / PurgeObsoleteFiles` can lead to race condition on which RocksDB attempts to delete a file twice. The second attempt will fail and return `IO error`. This may occur to other files, but this PR targets sst. Also add a unit test to verify that this PR fixes the issue. The newly added unit test `obsolete_files_test` has a test case for this scenario, implemented in `ObsoleteFilesTest#RaceForObsoleteFileDeletion`. `TestSyncPoint`s are used to coordinate the interleaving the `user_thread` and background compaction thread. They execute as follows ``` timeline user_thread background_compaction thread t1 | FindObsoleteFiles(full_scan=false) t2 | FindObsoleteFiles(full_scan=true) t3 | PurgeObsoleteFiles t4 | PurgeObsoleteFiles V ``` When `user_thread` invokes `FindObsoleteFiles` with full scan, it collects ALL files in RocksDB directory, including the ones that background compaction thread have collected in its job context. Then `user_thread` will see an IO error when trying to delete these files in `PurgeObsoleteFiles` because background compaction thread has already deleted the file in `PurgeObsoleteFiles`. To fix this, we make RocksDB remember which (SST) files have been found by threads after calling `FindObsoleteFiles` (see `DBImpl#files_grabbed_for_purge_`). Therefore, when another thread calls `FindObsoleteFiles` with full scan, it will not collect such files. ajkr could you take a look and comment? Thanks! Closes https://github.com/facebook/rocksdb/pull/3638 Differential Revision: D7384372 Pulled By: riversand963 fbshipit-source-id: 01489516d60012e722ee65a80e1449e589ce26d3
2018-03-28 17:23:31 +00:00
TEST_SYNC_POINT("ObsoleteFilesTest::RaceForObsoleteFileDeletion:2");
dbfull()->PurgeObsoleteFiles(jobCxt);
Fix race condition causing double deletion of ssts Summary: Possible interleaved execution of background compaction thread calling `FindObsoleteFiles (no full scan) / PurgeObsoleteFiles` and user thread calling `FindObsoleteFiles (full scan) / PurgeObsoleteFiles` can lead to race condition on which RocksDB attempts to delete a file twice. The second attempt will fail and return `IO error`. This may occur to other files, but this PR targets sst. Also add a unit test to verify that this PR fixes the issue. The newly added unit test `obsolete_files_test` has a test case for this scenario, implemented in `ObsoleteFilesTest#RaceForObsoleteFileDeletion`. `TestSyncPoint`s are used to coordinate the interleaving the `user_thread` and background compaction thread. They execute as follows ``` timeline user_thread background_compaction thread t1 | FindObsoleteFiles(full_scan=false) t2 | FindObsoleteFiles(full_scan=true) t3 | PurgeObsoleteFiles t4 | PurgeObsoleteFiles V ``` When `user_thread` invokes `FindObsoleteFiles` with full scan, it collects ALL files in RocksDB directory, including the ones that background compaction thread have collected in its job context. Then `user_thread` will see an IO error when trying to delete these files in `PurgeObsoleteFiles` because background compaction thread has already deleted the file in `PurgeObsoleteFiles`. To fix this, we make RocksDB remember which (SST) files have been found by threads after calling `FindObsoleteFiles` (see `DBImpl#files_grabbed_for_purge_`). Therefore, when another thread calls `FindObsoleteFiles` with full scan, it will not collect such files. ajkr could you take a look and comment? Thanks! Closes https://github.com/facebook/rocksdb/pull/3638 Differential Revision: D7384372 Pulled By: riversand963 fbshipit-source-id: 01489516d60012e722ee65a80e1449e589ce26d3
2018-03-28 17:23:31 +00:00
jobCxt.Clean();
});
user_thread.join();
}
TEST_F(ObsoleteFilesTest, DeleteObsoleteOptionsFile) {
ReopenDB();
createLevel0Files(2, 50000);
CheckFileTypeCounts(wal_dir_, 1, 0, 0);
ASSERT_OK(dbfull()->DisableFileDeletions());
for (int i = 0; i != 4; ++i) {
if (i % 2) {
ASSERT_OK(dbfull()->SetOptions(dbfull()->DefaultColumnFamily(),
{{"paranoid_file_checks", "false"}}));
} else {
ASSERT_OK(dbfull()->SetOptions(dbfull()->DefaultColumnFamily(),
{{"paranoid_file_checks", "true"}}));
}
}
ASSERT_OK(dbfull()->EnableFileDeletions());
Close();
std::vector<std::string> files;
int opts_file_count = 0;
ASSERT_OK(env_->GetChildren(dbname_, &files));
for (const auto& file : files) {
uint64_t file_num;
Slice dummy_info_log_name_prefix;
FileType type;
WalFileType log_type;
if (ParseFileName(file, &file_num, dummy_info_log_name_prefix, &type,
&log_type) &&
type == kOptionsFile) {
opts_file_count++;
}
}
ASSERT_EQ(2, opts_file_count);
}
TEST_F(ObsoleteFilesTest, BlobFiles) {
ReopenDB();
VersionSet* const versions = dbfull()->GetVersionSet();
assert(versions);
assert(versions->GetColumnFamilySet());
ColumnFamilyData* const cfd = versions->GetColumnFamilySet()->GetDefault();
assert(cfd);
Clean up blob files based on the linked SST set (#7001) Summary: The earlier `VersionBuilder` code only cleaned up blob files that were marked as entirely consisting of garbage using `VersionEdits` with `BlobFileGarbage`. This covers the cases when table files go through regular compaction, where we iterate through the KVs and thus have an opportunity to calculate the amount of garbage (that is, most cases). However, it does not help when table files are simply dropped (e.g. deletion compactions or the `DeleteFile` API). To deal with such cases, the patch adds logic that cleans up all blob files at the head of the list until the first one with linked SSTs is found. (As an example, let's assume we have blob files with numbers 1..10, and the first one with any linked SSTs is number 8. This means that SSTs in the `Version` only rely on blob files with numbers >= 8, and thus 1..7 are no longer needed.) The code change itself is pretty small; however, changing the logic like this necessitated changes to some tests that have been added recently (namely to the ones that use blob files in isolation, i.e. without any table files referring to them). Some of these cases were fixed by bypassing `VersionBuilder` altogether in order to keep the tests simple (which actually makes them more proper unit tests as well), while the `VersionBuilder` unit tests were fixed by adding dummy table files to the test cases as needed. Pull Request resolved: https://github.com/facebook/rocksdb/pull/7001 Test Plan: `make check` Reviewed By: riversand963 Differential Revision: D22119474 Pulled By: ltamasi fbshipit-source-id: c6547141355667d4291d9661d6518eb741e7b54a
2020-06-30 22:30:01 +00:00
const ImmutableCFOptions* const ioptions = cfd->ioptions();
assert(ioptions);
assert(!ioptions->cf_paths.empty());
Clean up blob files based on the linked SST set (#7001) Summary: The earlier `VersionBuilder` code only cleaned up blob files that were marked as entirely consisting of garbage using `VersionEdits` with `BlobFileGarbage`. This covers the cases when table files go through regular compaction, where we iterate through the KVs and thus have an opportunity to calculate the amount of garbage (that is, most cases). However, it does not help when table files are simply dropped (e.g. deletion compactions or the `DeleteFile` API). To deal with such cases, the patch adds logic that cleans up all blob files at the head of the list until the first one with linked SSTs is found. (As an example, let's assume we have blob files with numbers 1..10, and the first one with any linked SSTs is number 8. This means that SSTs in the `Version` only rely on blob files with numbers >= 8, and thus 1..7 are no longer needed.) The code change itself is pretty small; however, changing the logic like this necessitated changes to some tests that have been added recently (namely to the ones that use blob files in isolation, i.e. without any table files referring to them). Some of these cases were fixed by bypassing `VersionBuilder` altogether in order to keep the tests simple (which actually makes them more proper unit tests as well), while the `VersionBuilder` unit tests were fixed by adding dummy table files to the test cases as needed. Pull Request resolved: https://github.com/facebook/rocksdb/pull/7001 Test Plan: `make check` Reviewed By: riversand963 Differential Revision: D22119474 Pulled By: ltamasi fbshipit-source-id: c6547141355667d4291d9661d6518eb741e7b54a
2020-06-30 22:30:01 +00:00
const std::string& path = ioptions->cf_paths.front().path;
// Add an obsolete blob file.
constexpr uint64_t first_blob_file_number = 234;
Clean up blob files based on the linked SST set (#7001) Summary: The earlier `VersionBuilder` code only cleaned up blob files that were marked as entirely consisting of garbage using `VersionEdits` with `BlobFileGarbage`. This covers the cases when table files go through regular compaction, where we iterate through the KVs and thus have an opportunity to calculate the amount of garbage (that is, most cases). However, it does not help when table files are simply dropped (e.g. deletion compactions or the `DeleteFile` API). To deal with such cases, the patch adds logic that cleans up all blob files at the head of the list until the first one with linked SSTs is found. (As an example, let's assume we have blob files with numbers 1..10, and the first one with any linked SSTs is number 8. This means that SSTs in the `Version` only rely on blob files with numbers >= 8, and thus 1..7 are no longer needed.) The code change itself is pretty small; however, changing the logic like this necessitated changes to some tests that have been added recently (namely to the ones that use blob files in isolation, i.e. without any table files referring to them). Some of these cases were fixed by bypassing `VersionBuilder` altogether in order to keep the tests simple (which actually makes them more proper unit tests as well), while the `VersionBuilder` unit tests were fixed by adding dummy table files to the test cases as needed. Pull Request resolved: https://github.com/facebook/rocksdb/pull/7001 Test Plan: `make check` Reviewed By: riversand963 Differential Revision: D22119474 Pulled By: ltamasi fbshipit-source-id: c6547141355667d4291d9661d6518eb741e7b54a
2020-06-30 22:30:01 +00:00
versions->AddObsoleteBlobFile(first_blob_file_number, path);
// Add a live blob file.
Version* const version = cfd->current();
assert(version);
Clean up blob files based on the linked SST set (#7001) Summary: The earlier `VersionBuilder` code only cleaned up blob files that were marked as entirely consisting of garbage using `VersionEdits` with `BlobFileGarbage`. This covers the cases when table files go through regular compaction, where we iterate through the KVs and thus have an opportunity to calculate the amount of garbage (that is, most cases). However, it does not help when table files are simply dropped (e.g. deletion compactions or the `DeleteFile` API). To deal with such cases, the patch adds logic that cleans up all blob files at the head of the list until the first one with linked SSTs is found. (As an example, let's assume we have blob files with numbers 1..10, and the first one with any linked SSTs is number 8. This means that SSTs in the `Version` only rely on blob files with numbers >= 8, and thus 1..7 are no longer needed.) The code change itself is pretty small; however, changing the logic like this necessitated changes to some tests that have been added recently (namely to the ones that use blob files in isolation, i.e. without any table files referring to them). Some of these cases were fixed by bypassing `VersionBuilder` altogether in order to keep the tests simple (which actually makes them more proper unit tests as well), while the `VersionBuilder` unit tests were fixed by adding dummy table files to the test cases as needed. Pull Request resolved: https://github.com/facebook/rocksdb/pull/7001 Test Plan: `make check` Reviewed By: riversand963 Differential Revision: D22119474 Pulled By: ltamasi fbshipit-source-id: c6547141355667d4291d9661d6518eb741e7b54a
2020-06-30 22:30:01 +00:00
VersionStorageInfo* const storage_info = version->storage_info();
assert(storage_info);
constexpr uint64_t second_blob_file_number = 456;
constexpr uint64_t second_total_blob_count = 100;
constexpr uint64_t second_total_blob_bytes = 2000000;
constexpr char second_checksum_method[] = "CRC32B";
constexpr char second_checksum_value[] = "\x6d\xbd\xf2\x3a";
Clean up blob files based on the linked SST set (#7001) Summary: The earlier `VersionBuilder` code only cleaned up blob files that were marked as entirely consisting of garbage using `VersionEdits` with `BlobFileGarbage`. This covers the cases when table files go through regular compaction, where we iterate through the KVs and thus have an opportunity to calculate the amount of garbage (that is, most cases). However, it does not help when table files are simply dropped (e.g. deletion compactions or the `DeleteFile` API). To deal with such cases, the patch adds logic that cleans up all blob files at the head of the list until the first one with linked SSTs is found. (As an example, let's assume we have blob files with numbers 1..10, and the first one with any linked SSTs is number 8. This means that SSTs in the `Version` only rely on blob files with numbers >= 8, and thus 1..7 are no longer needed.) The code change itself is pretty small; however, changing the logic like this necessitated changes to some tests that have been added recently (namely to the ones that use blob files in isolation, i.e. without any table files referring to them). Some of these cases were fixed by bypassing `VersionBuilder` altogether in order to keep the tests simple (which actually makes them more proper unit tests as well), while the `VersionBuilder` unit tests were fixed by adding dummy table files to the test cases as needed. Pull Request resolved: https://github.com/facebook/rocksdb/pull/7001 Test Plan: `make check` Reviewed By: riversand963 Differential Revision: D22119474 Pulled By: ltamasi fbshipit-source-id: c6547141355667d4291d9661d6518eb741e7b54a
2020-06-30 22:30:01 +00:00
auto shared_meta = SharedBlobFileMetaData::Create(
second_blob_file_number, second_total_blob_count, second_total_blob_bytes,
second_checksum_method, second_checksum_value);
Clean up blob files based on the linked SST set (#7001) Summary: The earlier `VersionBuilder` code only cleaned up blob files that were marked as entirely consisting of garbage using `VersionEdits` with `BlobFileGarbage`. This covers the cases when table files go through regular compaction, where we iterate through the KVs and thus have an opportunity to calculate the amount of garbage (that is, most cases). However, it does not help when table files are simply dropped (e.g. deletion compactions or the `DeleteFile` API). To deal with such cases, the patch adds logic that cleans up all blob files at the head of the list until the first one with linked SSTs is found. (As an example, let's assume we have blob files with numbers 1..10, and the first one with any linked SSTs is number 8. This means that SSTs in the `Version` only rely on blob files with numbers >= 8, and thus 1..7 are no longer needed.) The code change itself is pretty small; however, changing the logic like this necessitated changes to some tests that have been added recently (namely to the ones that use blob files in isolation, i.e. without any table files referring to them). Some of these cases were fixed by bypassing `VersionBuilder` altogether in order to keep the tests simple (which actually makes them more proper unit tests as well), while the `VersionBuilder` unit tests were fixed by adding dummy table files to the test cases as needed. Pull Request resolved: https://github.com/facebook/rocksdb/pull/7001 Test Plan: `make check` Reviewed By: riversand963 Differential Revision: D22119474 Pulled By: ltamasi fbshipit-source-id: c6547141355667d4291d9661d6518eb741e7b54a
2020-06-30 22:30:01 +00:00
constexpr uint64_t second_garbage_blob_count = 0;
constexpr uint64_t second_garbage_blob_bytes = 0;
auto meta = BlobFileMetaData::Create(
std::move(shared_meta), BlobFileMetaData::LinkedSsts(),
second_garbage_blob_count, second_garbage_blob_bytes);
storage_info->AddBlobFile(std::move(meta));
// Check for obsolete files and make sure the first blob file is picked up
// and grabbed for purge. The second blob file should be on the live list.
constexpr int job_id = 0;
JobContext job_context{job_id};
dbfull()->TEST_LockMutex();
constexpr bool force_full_scan = false;
dbfull()->FindObsoleteFiles(&job_context, force_full_scan);
dbfull()->TEST_UnlockMutex();
ASSERT_TRUE(job_context.HaveSomethingToDelete());
ASSERT_EQ(job_context.blob_delete_files.size(), 1);
ASSERT_EQ(job_context.blob_delete_files[0].GetBlobFileNumber(),
first_blob_file_number);
const auto& files_grabbed_for_purge =
dbfull()->TEST_GetFilesGrabbedForPurge();
ASSERT_NE(files_grabbed_for_purge.find(first_blob_file_number),
files_grabbed_for_purge.end());
ASSERT_EQ(job_context.blob_live.size(), 1);
ASSERT_EQ(job_context.blob_live[0], second_blob_file_number);
// Hack the job context a bit by adding a few files to the full scan
// list and adjusting the pending file number. We add the two files
// above as well as two additional ones, where one is old
// and should be cleaned up, and the other is still pending.
constexpr uint64_t old_blob_file_number = 123;
constexpr uint64_t pending_blob_file_number = 567;
job_context.full_scan_candidate_files.emplace_back(
BlobFileName(old_blob_file_number), path);
job_context.full_scan_candidate_files.emplace_back(
BlobFileName(first_blob_file_number), path);
job_context.full_scan_candidate_files.emplace_back(
BlobFileName(second_blob_file_number), path);
job_context.full_scan_candidate_files.emplace_back(
BlobFileName(pending_blob_file_number), path);
job_context.min_pending_output = pending_blob_file_number;
// Purge obsolete files and make sure we purge the old file and the first file
// (and keep the second file and the pending file).
std::vector<std::string> deleted_files;
SyncPoint::GetInstance()->SetCallBack(
"DBImpl::DeleteObsoleteFileImpl::BeforeDeletion", [&](void* arg) {
const std::string* file = static_cast<std::string*>(arg);
assert(file);
constexpr char blob_extension[] = ".blob";
if (file->find(blob_extension) != std::string::npos) {
deleted_files.emplace_back(*file);
}
});
SyncPoint::GetInstance()->EnableProcessing();
dbfull()->PurgeObsoleteFiles(job_context);
job_context.Clean();
SyncPoint::GetInstance()->DisableProcessing();
SyncPoint::GetInstance()->ClearAllCallBacks();
ASSERT_EQ(files_grabbed_for_purge.find(first_blob_file_number),
files_grabbed_for_purge.end());
std::sort(deleted_files.begin(), deleted_files.end());
const std::vector<std::string> expected_deleted_files{
BlobFileName(path, old_blob_file_number),
BlobFileName(path, first_blob_file_number)};
ASSERT_EQ(deleted_files, expected_deleted_files);
}
} // namespace ROCKSDB_NAMESPACE
Fix race condition causing double deletion of ssts Summary: Possible interleaved execution of background compaction thread calling `FindObsoleteFiles (no full scan) / PurgeObsoleteFiles` and user thread calling `FindObsoleteFiles (full scan) / PurgeObsoleteFiles` can lead to race condition on which RocksDB attempts to delete a file twice. The second attempt will fail and return `IO error`. This may occur to other files, but this PR targets sst. Also add a unit test to verify that this PR fixes the issue. The newly added unit test `obsolete_files_test` has a test case for this scenario, implemented in `ObsoleteFilesTest#RaceForObsoleteFileDeletion`. `TestSyncPoint`s are used to coordinate the interleaving the `user_thread` and background compaction thread. They execute as follows ``` timeline user_thread background_compaction thread t1 | FindObsoleteFiles(full_scan=false) t2 | FindObsoleteFiles(full_scan=true) t3 | PurgeObsoleteFiles t4 | PurgeObsoleteFiles V ``` When `user_thread` invokes `FindObsoleteFiles` with full scan, it collects ALL files in RocksDB directory, including the ones that background compaction thread have collected in its job context. Then `user_thread` will see an IO error when trying to delete these files in `PurgeObsoleteFiles` because background compaction thread has already deleted the file in `PurgeObsoleteFiles`. To fix this, we make RocksDB remember which (SST) files have been found by threads after calling `FindObsoleteFiles` (see `DBImpl#files_grabbed_for_purge_`). Therefore, when another thread calls `FindObsoleteFiles` with full scan, it will not collect such files. ajkr could you take a look and comment? Thanks! Closes https://github.com/facebook/rocksdb/pull/3638 Differential Revision: D7384372 Pulled By: riversand963 fbshipit-source-id: 01489516d60012e722ee65a80e1449e589ce26d3
2018-03-28 17:23:31 +00:00
int main(int argc, char** argv) {
ROCKSDB_NAMESPACE::port::InstallStackTraceHandler();
Fix race condition causing double deletion of ssts Summary: Possible interleaved execution of background compaction thread calling `FindObsoleteFiles (no full scan) / PurgeObsoleteFiles` and user thread calling `FindObsoleteFiles (full scan) / PurgeObsoleteFiles` can lead to race condition on which RocksDB attempts to delete a file twice. The second attempt will fail and return `IO error`. This may occur to other files, but this PR targets sst. Also add a unit test to verify that this PR fixes the issue. The newly added unit test `obsolete_files_test` has a test case for this scenario, implemented in `ObsoleteFilesTest#RaceForObsoleteFileDeletion`. `TestSyncPoint`s are used to coordinate the interleaving the `user_thread` and background compaction thread. They execute as follows ``` timeline user_thread background_compaction thread t1 | FindObsoleteFiles(full_scan=false) t2 | FindObsoleteFiles(full_scan=true) t3 | PurgeObsoleteFiles t4 | PurgeObsoleteFiles V ``` When `user_thread` invokes `FindObsoleteFiles` with full scan, it collects ALL files in RocksDB directory, including the ones that background compaction thread have collected in its job context. Then `user_thread` will see an IO error when trying to delete these files in `PurgeObsoleteFiles` because background compaction thread has already deleted the file in `PurgeObsoleteFiles`. To fix this, we make RocksDB remember which (SST) files have been found by threads after calling `FindObsoleteFiles` (see `DBImpl#files_grabbed_for_purge_`). Therefore, when another thread calls `FindObsoleteFiles` with full scan, it will not collect such files. ajkr could you take a look and comment? Thanks! Closes https://github.com/facebook/rocksdb/pull/3638 Differential Revision: D7384372 Pulled By: riversand963 fbshipit-source-id: 01489516d60012e722ee65a80e1449e589ce26d3
2018-03-28 17:23:31 +00:00
::testing::InitGoogleTest(&argc, argv);
RegisterCustomObjects(argc, argv);
Fix race condition causing double deletion of ssts Summary: Possible interleaved execution of background compaction thread calling `FindObsoleteFiles (no full scan) / PurgeObsoleteFiles` and user thread calling `FindObsoleteFiles (full scan) / PurgeObsoleteFiles` can lead to race condition on which RocksDB attempts to delete a file twice. The second attempt will fail and return `IO error`. This may occur to other files, but this PR targets sst. Also add a unit test to verify that this PR fixes the issue. The newly added unit test `obsolete_files_test` has a test case for this scenario, implemented in `ObsoleteFilesTest#RaceForObsoleteFileDeletion`. `TestSyncPoint`s are used to coordinate the interleaving the `user_thread` and background compaction thread. They execute as follows ``` timeline user_thread background_compaction thread t1 | FindObsoleteFiles(full_scan=false) t2 | FindObsoleteFiles(full_scan=true) t3 | PurgeObsoleteFiles t4 | PurgeObsoleteFiles V ``` When `user_thread` invokes `FindObsoleteFiles` with full scan, it collects ALL files in RocksDB directory, including the ones that background compaction thread have collected in its job context. Then `user_thread` will see an IO error when trying to delete these files in `PurgeObsoleteFiles` because background compaction thread has already deleted the file in `PurgeObsoleteFiles`. To fix this, we make RocksDB remember which (SST) files have been found by threads after calling `FindObsoleteFiles` (see `DBImpl#files_grabbed_for_purge_`). Therefore, when another thread calls `FindObsoleteFiles` with full scan, it will not collect such files. ajkr could you take a look and comment? Thanks! Closes https://github.com/facebook/rocksdb/pull/3638 Differential Revision: D7384372 Pulled By: riversand963 fbshipit-source-id: 01489516d60012e722ee65a80e1449e589ce26d3
2018-03-28 17:23:31 +00:00
return RUN_ALL_TESTS();
}