2016-02-09 23:12:00 +00:00
|
|
|
// Copyright (c) 2011-present, Facebook, Inc. All rights reserved.
|
2014-01-02 17:08:12 +00:00
|
|
|
// This source code is licensed under the BSD-style license found in the
|
|
|
|
// LICENSE file in the root directory of this source tree. An additional grant
|
|
|
|
// of patent rights can be found in the PATENTS file in the same 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.
|
|
|
|
|
2014-02-26 18:03:34 +00:00
|
|
|
#include <algorithm>
|
|
|
|
#include <vector>
|
|
|
|
#include <string>
|
LogAndApply() should fail if the column family has been dropped
Summary:
This patch finally fixes the ColumnFamilyTest.ReadDroppedColumnFamily test. The test has been failing very sporadically and it was hard to repro. However, I managed to write a new tests that reproes the failure deterministically.
Here's what happens:
1. We start the flush for the column family
2. We check if the column family was dropped here: https://github.com/facebook/rocksdb/blob/a3fc49bfddcdb1ff29409aacd06c04df56c7a1d7/db/flush_job.cc#L149
3. This check goes through, ends up in InstallMemtableFlushResults() and it goes into LogAndApply()
4. At about this time, we start dropping the column family. Dropping the column family process gets to LogAndApply() at about the same time as LogAndApply() from flush process
5. Drop column family goes through LogAndApply() first, marking the column family as dropped.
6. Flush process gets woken up and gets a chance to write to the MANIFEST. However, this is where it gets stuck: https://github.com/facebook/rocksdb/blob/a3fc49bfddcdb1ff29409aacd06c04df56c7a1d7/db/version_set.cc#L1975
7. We see that the column family was dropped, so there is no need to write to the MANIFEST. We return OK.
8. Flush gets OK back from LogAndApply() and it deletes the memtable, thinking that the data is now safely persisted to sst file.
The fix is pretty simple. Instead of OK, we return ShutdownInProgress. This is not really true, but we have been using this status code to also mean "this operation was canceled because the column family has been dropped".
The fix is only one LOC. All other code is related to tests. I added a new test that reproes the failure. I also moved SleepingBackgroundTask to util/testutil.h (because I needed it in column_family_test for my new test). There's plenty of other places where we reimplement SleepingBackgroundTask, but I'll address that in a separate commit.
Test Plan:
1. new test
2. make check
3. Make sure the ColumnFamilyTest.ReadDroppedColumnFamily doesn't fail on Travis: https://travis-ci.org/facebook/rocksdb/jobs/79952386
Reviewers: yhchiang, anthony, IslamAbdelRahman, kradhakrishnan, rven, sdong
Reviewed By: sdong
Subscribers: dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D46773
2015-09-15 18:28:44 +00:00
|
|
|
#include <thread>
|
2014-02-26 18:03:34 +00:00
|
|
|
|
2014-01-02 17:08:12 +00:00
|
|
|
#include "db/db_impl.h"
|
2016-06-06 21:40:36 +00:00
|
|
|
#include "db/db_test_util.h"
|
2014-01-02 17:08:12 +00:00
|
|
|
#include "rocksdb/db.h"
|
2015-03-20 00:04:29 +00:00
|
|
|
#include "rocksdb/env.h"
|
|
|
|
#include "rocksdb/iterator.h"
|
2016-01-07 02:14:01 +00:00
|
|
|
#include "util/coding.h"
|
2016-07-19 22:12:46 +00:00
|
|
|
#include "util/fault_injection_test_env.h"
|
2016-01-07 02:14:01 +00:00
|
|
|
#include "util/options_parser.h"
|
2015-03-20 00:29:37 +00:00
|
|
|
#include "util/string_util.h"
|
2016-01-07 02:14:01 +00:00
|
|
|
#include "util/sync_point.h"
|
2014-01-02 17:08:12 +00:00
|
|
|
#include "util/testharness.h"
|
2014-02-25 18:38:04 +00:00
|
|
|
#include "util/testutil.h"
|
2014-01-28 19:05:04 +00:00
|
|
|
#include "utilities/merge_operators.h"
|
2014-01-02 17:08:12 +00:00
|
|
|
|
|
|
|
namespace rocksdb {
|
|
|
|
|
2016-06-06 21:40:36 +00:00
|
|
|
static const int kValueSize = 1000;
|
|
|
|
|
2014-02-25 18:38:04 +00:00
|
|
|
namespace {
|
|
|
|
std::string RandomString(Random* rnd, int len) {
|
|
|
|
std::string r;
|
|
|
|
test::RandomString(rnd, len, &r);
|
|
|
|
return r;
|
|
|
|
}
|
|
|
|
} // anonymous namespace
|
|
|
|
|
2014-04-15 16:57:25 +00:00
|
|
|
// counts how many operations were performed
|
|
|
|
class EnvCounter : public EnvWrapper {
|
|
|
|
public:
|
|
|
|
explicit EnvCounter(Env* base)
|
|
|
|
: EnvWrapper(base), num_new_writable_file_(0) {}
|
|
|
|
int GetNumberOfNewWritableFileCalls() {
|
|
|
|
return num_new_writable_file_;
|
|
|
|
}
|
|
|
|
Status NewWritableFile(const std::string& f, unique_ptr<WritableFile>* r,
|
2015-02-26 19:28:41 +00:00
|
|
|
const EnvOptions& soptions) override {
|
2014-04-15 16:57:25 +00:00
|
|
|
++num_new_writable_file_;
|
|
|
|
return EnvWrapper::NewWritableFile(f, r, soptions);
|
|
|
|
}
|
|
|
|
|
|
|
|
private:
|
|
|
|
int num_new_writable_file_;
|
|
|
|
};
|
|
|
|
|
2015-03-17 21:08:00 +00:00
|
|
|
class ColumnFamilyTest : public testing::Test {
|
2014-01-02 17:08:12 +00:00
|
|
|
public:
|
2014-02-25 18:38:04 +00:00
|
|
|
ColumnFamilyTest() : rnd_(139) {
|
2014-04-15 16:57:25 +00:00
|
|
|
env_ = new EnvCounter(Env::Default());
|
2014-01-02 17:08:12 +00:00
|
|
|
dbname_ = test::TmpDir() + "/column_family_test";
|
|
|
|
db_options_.create_if_missing = true;
|
2015-11-11 06:58:01 +00:00
|
|
|
db_options_.fail_if_options_file_error = true;
|
2014-04-15 16:57:25 +00:00
|
|
|
db_options_.env = env_;
|
2014-01-28 19:05:04 +00:00
|
|
|
DestroyDB(dbname_, Options(db_options_, column_family_options_));
|
2014-01-02 17:08:12 +00:00
|
|
|
}
|
|
|
|
|
2014-04-16 00:12:18 +00:00
|
|
|
~ColumnFamilyTest() {
|
2015-12-28 19:23:06 +00:00
|
|
|
Close();
|
2016-02-01 20:45:45 +00:00
|
|
|
rocksdb::SyncPoint::GetInstance()->DisableProcessing();
|
2015-12-28 19:23:06 +00:00
|
|
|
Destroy();
|
2014-04-16 00:12:18 +00:00
|
|
|
delete env_;
|
|
|
|
}
|
|
|
|
|
2016-06-06 21:40:36 +00:00
|
|
|
// Return the value to associate with the specified key
|
|
|
|
Slice Value(int k, std::string* storage) {
|
|
|
|
if (k == 0) {
|
|
|
|
// Ugh. Random seed of 0 used to produce no entropy. This code
|
|
|
|
// preserves the implementation that was in place when all of the
|
|
|
|
// magic values in this file were picked.
|
|
|
|
*storage = std::string(kValueSize, ' ');
|
|
|
|
return Slice(*storage);
|
|
|
|
} else {
|
|
|
|
Random r(k);
|
|
|
|
return test::RandomString(&r, kValueSize, storage);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
void Build(int base, int n, int flush_every = 0) {
|
|
|
|
std::string key_space, value_space;
|
|
|
|
WriteBatch batch;
|
|
|
|
|
|
|
|
for (int i = 0; i < n; i++) {
|
|
|
|
if (flush_every != 0 && i != 0 && i % flush_every == 0) {
|
|
|
|
DBImpl* dbi = reinterpret_cast<DBImpl*>(db_);
|
|
|
|
dbi->TEST_FlushMemTable();
|
|
|
|
}
|
|
|
|
|
|
|
|
int keyi = base + i;
|
|
|
|
Slice key(DBTestBase::Key(keyi));
|
|
|
|
|
|
|
|
batch.Clear();
|
|
|
|
batch.Put(handles_[0], key, Value(keyi, &value_space));
|
|
|
|
batch.Put(handles_[1], key, Value(keyi, &value_space));
|
|
|
|
batch.Put(handles_[2], key, Value(keyi, &value_space));
|
|
|
|
ASSERT_OK(db_->Write(WriteOptions(), &batch));
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
void CheckMissed() {
|
|
|
|
uint64_t next_expected = 0;
|
|
|
|
uint64_t missed = 0;
|
|
|
|
int bad_keys = 0;
|
|
|
|
int bad_values = 0;
|
|
|
|
int correct = 0;
|
|
|
|
std::string value_space;
|
|
|
|
for (int cf = 0; cf < 3; cf++) {
|
|
|
|
next_expected = 0;
|
|
|
|
Iterator* iter = db_->NewIterator(ReadOptions(false, true), handles_[cf]);
|
|
|
|
for (iter->SeekToFirst(); iter->Valid(); iter->Next()) {
|
|
|
|
uint64_t key;
|
|
|
|
Slice in(iter->key());
|
|
|
|
in.remove_prefix(3);
|
|
|
|
if (!ConsumeDecimalNumber(&in, &key) || !in.empty() ||
|
|
|
|
key < next_expected) {
|
|
|
|
bad_keys++;
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
missed += (key - next_expected);
|
|
|
|
next_expected = key + 1;
|
|
|
|
if (iter->value() != Value(static_cast<int>(key), &value_space)) {
|
|
|
|
bad_values++;
|
|
|
|
} else {
|
|
|
|
correct++;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
delete iter;
|
|
|
|
}
|
|
|
|
|
|
|
|
ASSERT_EQ(0, bad_keys);
|
|
|
|
ASSERT_EQ(0, bad_values);
|
|
|
|
ASSERT_EQ(0, missed);
|
|
|
|
(void)correct;
|
|
|
|
}
|
|
|
|
|
2014-01-02 17:08:12 +00:00
|
|
|
void Close() {
|
2014-02-11 01:04:44 +00:00
|
|
|
for (auto h : handles_) {
|
2015-12-09 01:01:02 +00:00
|
|
|
if (h) {
|
2016-07-14 00:59:25 +00:00
|
|
|
db_->DestroyColumnFamilyHandle(h);
|
2015-12-09 01:01:02 +00:00
|
|
|
}
|
2014-02-11 01:04:44 +00:00
|
|
|
}
|
|
|
|
handles_.clear();
|
2014-02-26 22:16:23 +00:00
|
|
|
names_.clear();
|
2014-01-02 17:08:12 +00:00
|
|
|
delete db_;
|
|
|
|
db_ = nullptr;
|
|
|
|
}
|
|
|
|
|
2014-02-26 22:16:23 +00:00
|
|
|
Status TryOpen(std::vector<std::string> cf,
|
|
|
|
std::vector<ColumnFamilyOptions> options = {}) {
|
2014-02-26 18:03:34 +00:00
|
|
|
std::vector<ColumnFamilyDescriptor> column_families;
|
2014-02-26 22:16:23 +00:00
|
|
|
names_.clear();
|
|
|
|
for (size_t i = 0; i < cf.size(); ++i) {
|
|
|
|
column_families.push_back(ColumnFamilyDescriptor(
|
|
|
|
cf[i], options.size() == 0 ? column_family_options_ : options[i]));
|
|
|
|
names_.push_back(cf[i]);
|
2014-01-06 21:31:06 +00:00
|
|
|
}
|
2014-02-07 22:48:48 +00:00
|
|
|
return DB::Open(db_options_, dbname_, column_families, &handles_, &db_);
|
2014-01-02 17:08:12 +00:00
|
|
|
}
|
|
|
|
|
2014-04-09 16:56:17 +00:00
|
|
|
Status OpenReadOnly(std::vector<std::string> cf,
|
|
|
|
std::vector<ColumnFamilyOptions> options = {}) {
|
|
|
|
std::vector<ColumnFamilyDescriptor> column_families;
|
|
|
|
names_.clear();
|
|
|
|
for (size_t i = 0; i < cf.size(); ++i) {
|
|
|
|
column_families.push_back(ColumnFamilyDescriptor(
|
|
|
|
cf[i], options.size() == 0 ? column_family_options_ : options[i]));
|
|
|
|
names_.push_back(cf[i]);
|
|
|
|
}
|
|
|
|
return DB::OpenForReadOnly(db_options_, dbname_, column_families, &handles_,
|
|
|
|
&db_);
|
|
|
|
}
|
|
|
|
|
2015-06-29 21:39:01 +00:00
|
|
|
#ifndef ROCKSDB_LITE // ReadOnlyDB is not supported
|
2014-04-09 16:56:17 +00:00
|
|
|
void AssertOpenReadOnly(std::vector<std::string> cf,
|
|
|
|
std::vector<ColumnFamilyOptions> options = {}) {
|
|
|
|
ASSERT_OK(OpenReadOnly(cf, options));
|
|
|
|
}
|
2015-06-29 21:39:01 +00:00
|
|
|
#endif // !ROCKSDB_LITE
|
2014-04-09 16:56:17 +00:00
|
|
|
|
|
|
|
|
2014-02-26 22:16:23 +00:00
|
|
|
void Open(std::vector<std::string> cf,
|
|
|
|
std::vector<ColumnFamilyOptions> options = {}) {
|
|
|
|
ASSERT_OK(TryOpen(cf, options));
|
|
|
|
}
|
|
|
|
|
|
|
|
void Open() {
|
|
|
|
Open({"default"});
|
|
|
|
}
|
|
|
|
|
2014-02-25 18:38:04 +00:00
|
|
|
DBImpl* dbfull() { return reinterpret_cast<DBImpl*>(db_); }
|
|
|
|
|
2014-02-26 22:16:23 +00:00
|
|
|
int GetProperty(int cf, std::string property) {
|
|
|
|
std::string value;
|
rocksdb: Replace ASSERT* with EXPECT* in functions that does not return void value
Summary:
gtest does not use exceptions to fail a unit test by design, and `ASSERT*`s are implemented using `return`. As a consequence we cannot use `ASSERT*` in a function that does not return `void` value ([[ https://code.google.com/p/googletest/wiki/AdvancedGuide#Assertion_Placement | 1]]), and have to fix our existing code. This diff does this in a generic way, with no manual changes.
In order to detect all existing `ASSERT*` that are used in functions that doesn't return void value, I change the code to generate compile errors for such cases.
In `util/testharness.h` I defined `EXPECT*` assertions, the same way as `ASSERT*`, and redefined `ASSERT*` to return `void`. Then executed:
```lang=bash
% USE_CLANG=1 make all -j55 -k 2> build.log
% perl -naF: -e 'print "-- -number=".$F[1]." ".$F[0]."\n" if /: error:/' \
build.log | xargs -L 1 perl -spi -e 's/ASSERT/EXPECT/g if $. == $number'
% make format
```
After that I reverted back change to `ASSERT*` in `util/testharness.h`. But preserved introduced `EXPECT*`, which is the same as `ASSERT*`. This will be deleted once switched to gtest.
This diff is independent and contains manual changes only in `util/testharness.h`.
Test Plan:
Make sure all tests are passing.
```lang=bash
% USE_CLANG=1 make check
```
Reviewers: igor, lgalanis, sdong, yufei.zhu, rven, meyering
Reviewed By: meyering
Subscribers: dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D33333
2015-03-17 03:52:32 +00:00
|
|
|
EXPECT_TRUE(dbfull()->GetProperty(handles_[cf], property, &value));
|
2015-04-24 02:17:57 +00:00
|
|
|
#ifndef CYGWIN
|
2014-02-26 22:16:23 +00:00
|
|
|
return std::stoi(value);
|
2015-04-24 02:17:57 +00:00
|
|
|
#else
|
2015-06-08 18:43:55 +00:00
|
|
|
return std::strtol(value.c_str(), 0 /* off */, 10 /* base */);
|
2015-04-24 02:17:57 +00:00
|
|
|
#endif
|
2014-02-26 22:16:23 +00:00
|
|
|
}
|
|
|
|
|
2014-01-28 19:05:04 +00:00
|
|
|
void Destroy() {
|
2015-12-28 19:23:06 +00:00
|
|
|
Close();
|
2014-01-28 19:05:04 +00:00
|
|
|
ASSERT_OK(DestroyDB(dbname_, Options(db_options_, column_family_options_)));
|
|
|
|
}
|
|
|
|
|
2014-02-26 22:16:23 +00:00
|
|
|
void CreateColumnFamilies(
|
|
|
|
const std::vector<std::string>& cfs,
|
|
|
|
const std::vector<ColumnFamilyOptions> options = {}) {
|
2014-11-11 21:47:22 +00:00
|
|
|
int cfi = static_cast<int>(handles_.size());
|
2014-01-28 19:05:04 +00:00
|
|
|
handles_.resize(cfi + cfs.size());
|
2014-02-26 22:16:23 +00:00
|
|
|
names_.resize(cfi + cfs.size());
|
|
|
|
for (size_t i = 0; i < cfs.size(); ++i) {
|
2016-01-07 02:14:01 +00:00
|
|
|
const auto& current_cf_opt =
|
|
|
|
options.size() == 0 ? column_family_options_ : options[i];
|
|
|
|
ASSERT_OK(
|
|
|
|
db_->CreateColumnFamily(current_cf_opt, cfs[i], &handles_[cfi]));
|
2014-02-26 22:16:23 +00:00
|
|
|
names_[cfi] = cfs[i];
|
2016-01-07 02:14:01 +00:00
|
|
|
|
|
|
|
#ifndef ROCKSDB_LITE // RocksDBLite does not support GetDescriptor
|
|
|
|
// Verify the CF options of the returned CF handle.
|
|
|
|
ColumnFamilyDescriptor desc;
|
|
|
|
ASSERT_OK(handles_[cfi]->GetDescriptor(&desc));
|
|
|
|
RocksDBOptionsParser::VerifyCFOptions(desc.options, current_cf_opt);
|
|
|
|
#endif // !ROCKSDB_LITE
|
2014-02-26 22:16:23 +00:00
|
|
|
cfi++;
|
2014-01-28 19:05:04 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2014-02-26 22:16:23 +00:00
|
|
|
void Reopen(const std::vector<ColumnFamilyOptions> options = {}) {
|
|
|
|
std::vector<std::string> names;
|
|
|
|
for (auto name : names_) {
|
|
|
|
if (name != "") {
|
|
|
|
names.push_back(name);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
Close();
|
|
|
|
assert(options.size() == 0 || names.size() == options.size());
|
|
|
|
Open(names, options);
|
|
|
|
}
|
|
|
|
|
|
|
|
void CreateColumnFamiliesAndReopen(const std::vector<std::string>& cfs) {
|
|
|
|
CreateColumnFamilies(cfs);
|
|
|
|
Reopen();
|
|
|
|
}
|
|
|
|
|
2014-02-26 18:03:34 +00:00
|
|
|
void DropColumnFamilies(const std::vector<int>& cfs) {
|
2014-02-11 01:04:44 +00:00
|
|
|
for (auto cf : cfs) {
|
|
|
|
ASSERT_OK(db_->DropColumnFamily(handles_[cf]));
|
2016-07-14 00:59:25 +00:00
|
|
|
db_->DestroyColumnFamilyHandle(handles_[cf]);
|
2014-02-11 01:04:44 +00:00
|
|
|
handles_[cf] = nullptr;
|
2014-02-26 22:16:23 +00:00
|
|
|
names_[cf] = "";
|
2014-02-11 01:04:44 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
Running manual compactions in parallel with other automatic or manual compactions in restricted cases
Summary:
This diff provides a framework for doing manual
compactions in parallel with other compactions. We now have a deque of manual compactions. We also pass manual compactions as an argument from RunManualCompactions down to
BackgroundCompactions, so that RunManualCompactions can be reentrant.
Parallelism is controlled by the two routines
ConflictingManualCompaction to allow/disallow new parallel/manual
compactions based on already existing ManualCompactions. In this diff, by default manual compactions still have to run exclusive of other compactions. However, by setting the compaction option, exclusive_manual_compaction to false, it is possible to run other compactions in parallel with a manual compaction. However, we are still restricted to one manual compaction per column family at a time. All of these restrictions will be relaxed in future diffs.
I will be adding more tests later.
Test Plan: Rocksdb regression + new tests + valgrind
Reviewers: igor, anthony, IslamAbdelRahman, kradhakrishnan, yhchiang, sdong
Reviewed By: sdong
Subscribers: yoshinorim, dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D47973
2015-12-14 19:20:34 +00:00
|
|
|
void PutRandomData(int cf, int num, int key_value_size, bool save = false) {
|
2014-02-26 22:16:23 +00:00
|
|
|
for (int i = 0; i < num; ++i) {
|
|
|
|
// 10 bytes for key, rest is value
|
Running manual compactions in parallel with other automatic or manual compactions in restricted cases
Summary:
This diff provides a framework for doing manual
compactions in parallel with other compactions. We now have a deque of manual compactions. We also pass manual compactions as an argument from RunManualCompactions down to
BackgroundCompactions, so that RunManualCompactions can be reentrant.
Parallelism is controlled by the two routines
ConflictingManualCompaction to allow/disallow new parallel/manual
compactions based on already existing ManualCompactions. In this diff, by default manual compactions still have to run exclusive of other compactions. However, by setting the compaction option, exclusive_manual_compaction to false, it is possible to run other compactions in parallel with a manual compaction. However, we are still restricted to one manual compaction per column family at a time. All of these restrictions will be relaxed in future diffs.
I will be adding more tests later.
Test Plan: Rocksdb regression + new tests + valgrind
Reviewers: igor, anthony, IslamAbdelRahman, kradhakrishnan, yhchiang, sdong
Reviewed By: sdong
Subscribers: yoshinorim, dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D47973
2015-12-14 19:20:34 +00:00
|
|
|
if (!save) {
|
|
|
|
ASSERT_OK(Put(cf, test::RandomKey(&rnd_, 11),
|
|
|
|
RandomString(&rnd_, key_value_size - 10)));
|
|
|
|
} else {
|
|
|
|
std::string key = test::RandomKey(&rnd_, 11);
|
|
|
|
keys_.insert(key);
|
|
|
|
ASSERT_OK(Put(cf, key, RandomString(&rnd_, key_value_size - 10)));
|
|
|
|
}
|
2014-02-25 18:38:04 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
void WaitForFlush(int cf) {
|
2015-06-29 21:39:01 +00:00
|
|
|
#ifndef ROCKSDB_LITE // TEST functions are not supported in lite
|
2014-02-25 18:38:04 +00:00
|
|
|
ASSERT_OK(dbfull()->TEST_WaitForFlushMemTable(handles_[cf]));
|
2015-06-29 21:39:01 +00:00
|
|
|
#endif // !ROCKSDB_LITE
|
2014-02-25 18:38:04 +00:00
|
|
|
}
|
|
|
|
|
2015-06-29 21:39:01 +00:00
|
|
|
void WaitForCompaction() {
|
|
|
|
#ifndef ROCKSDB_LITE // TEST functions are not supported in lite
|
|
|
|
ASSERT_OK(dbfull()->TEST_WaitForCompact());
|
|
|
|
#endif // !ROCKSDB_LITE
|
|
|
|
}
|
|
|
|
|
|
|
|
uint64_t MaxTotalInMemoryState() {
|
|
|
|
#ifndef ROCKSDB_LITE
|
|
|
|
return dbfull()->TEST_MaxTotalInMemoryState();
|
|
|
|
#else
|
|
|
|
return 0;
|
|
|
|
#endif // !ROCKSDB_LITE
|
|
|
|
}
|
|
|
|
|
|
|
|
void AssertMaxTotalInMemoryState(uint64_t value) {
|
|
|
|
ASSERT_EQ(value, MaxTotalInMemoryState());
|
|
|
|
}
|
2014-02-27 00:05:24 +00:00
|
|
|
|
2014-02-26 18:03:34 +00:00
|
|
|
Status Put(int cf, const std::string& key, const std::string& value) {
|
2014-01-28 19:05:04 +00:00
|
|
|
return db_->Put(WriteOptions(), handles_[cf], Slice(key), Slice(value));
|
|
|
|
}
|
2014-02-26 18:03:34 +00:00
|
|
|
Status Merge(int cf, const std::string& key, const std::string& value) {
|
2014-01-28 19:05:04 +00:00
|
|
|
return db_->Merge(WriteOptions(), handles_[cf], Slice(key), Slice(value));
|
|
|
|
}
|
2014-01-31 01:48:42 +00:00
|
|
|
Status Flush(int cf) {
|
|
|
|
return db_->Flush(FlushOptions(), handles_[cf]);
|
|
|
|
}
|
2014-01-28 19:05:04 +00:00
|
|
|
|
2014-02-26 18:03:34 +00:00
|
|
|
std::string Get(int cf, const std::string& key) {
|
2014-01-28 19:05:04 +00:00
|
|
|
ReadOptions options;
|
|
|
|
options.verify_checksums = true;
|
2014-02-26 18:03:34 +00:00
|
|
|
std::string result;
|
2014-01-28 19:05:04 +00:00
|
|
|
Status s = db_->Get(options, handles_[cf], Slice(key), &result);
|
|
|
|
if (s.IsNotFound()) {
|
|
|
|
result = "NOT_FOUND";
|
|
|
|
} else if (!s.ok()) {
|
|
|
|
result = s.ToString();
|
|
|
|
}
|
|
|
|
return result;
|
|
|
|
}
|
|
|
|
|
2014-02-27 00:05:24 +00:00
|
|
|
void CompactAll(int cf) {
|
2015-06-17 21:36:14 +00:00
|
|
|
ASSERT_OK(db_->CompactRange(CompactRangeOptions(), handles_[cf], nullptr,
|
|
|
|
nullptr));
|
2014-02-27 00:05:24 +00:00
|
|
|
}
|
|
|
|
|
2014-02-01 00:45:20 +00:00
|
|
|
void Compact(int cf, const Slice& start, const Slice& limit) {
|
2015-06-17 21:36:14 +00:00
|
|
|
ASSERT_OK(
|
|
|
|
db_->CompactRange(CompactRangeOptions(), handles_[cf], &start, &limit));
|
2014-02-01 00:45:20 +00:00
|
|
|
}
|
|
|
|
|
2014-02-27 00:05:24 +00:00
|
|
|
int NumTableFilesAtLevel(int level, int cf) {
|
|
|
|
return GetProperty(cf,
|
2014-11-25 04:44:49 +00:00
|
|
|
"rocksdb.num-files-at-level" + ToString(level));
|
2014-02-01 00:45:20 +00:00
|
|
|
}
|
|
|
|
|
2015-06-29 21:39:01 +00:00
|
|
|
#ifndef ROCKSDB_LITE
|
2014-02-01 00:45:20 +00:00
|
|
|
// Return spread of files per level
|
2014-02-26 18:03:34 +00:00
|
|
|
std::string FilesPerLevel(int cf) {
|
|
|
|
std::string result;
|
2014-02-01 00:45:20 +00:00
|
|
|
int last_non_zero_offset = 0;
|
2014-02-27 00:05:24 +00:00
|
|
|
for (int level = 0; level < dbfull()->NumberLevels(handles_[cf]); level++) {
|
|
|
|
int f = NumTableFilesAtLevel(level, cf);
|
2014-02-01 00:45:20 +00:00
|
|
|
char buf[100];
|
|
|
|
snprintf(buf, sizeof(buf), "%s%d", (level ? "," : ""), f);
|
|
|
|
result += buf;
|
|
|
|
if (f > 0) {
|
2014-11-11 21:47:22 +00:00
|
|
|
last_non_zero_offset = static_cast<int>(result.size());
|
2014-02-01 00:45:20 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
result.resize(last_non_zero_offset);
|
|
|
|
return result;
|
|
|
|
}
|
2015-06-29 21:39:01 +00:00
|
|
|
#endif
|
2014-02-01 00:45:20 +00:00
|
|
|
|
2015-06-29 21:39:01 +00:00
|
|
|
void AssertFilesPerLevel(const std::string& value, int cf) {
|
|
|
|
#ifndef ROCKSDB_LITE
|
|
|
|
ASSERT_EQ(value, FilesPerLevel(cf));
|
|
|
|
#endif
|
|
|
|
}
|
|
|
|
|
|
|
|
#ifndef ROCKSDB_LITE // GetLiveFilesMetaData is not supported
|
2014-04-30 18:33:40 +00:00
|
|
|
int CountLiveFiles() {
|
2014-02-11 01:04:44 +00:00
|
|
|
std::vector<LiveFileMetaData> metadata;
|
|
|
|
db_->GetLiveFilesMetaData(&metadata);
|
|
|
|
return static_cast<int>(metadata.size());
|
|
|
|
}
|
2015-06-29 21:39:01 +00:00
|
|
|
#endif // !ROCKSDB_LITE
|
|
|
|
|
|
|
|
void AssertCountLiveFiles(int expected_value) {
|
|
|
|
#ifndef ROCKSDB_LITE
|
|
|
|
ASSERT_EQ(expected_value, CountLiveFiles());
|
|
|
|
#endif
|
|
|
|
}
|
2014-02-11 01:04:44 +00:00
|
|
|
|
2014-02-01 00:45:20 +00:00
|
|
|
// Do n memtable flushes, each of which produces an sstable
|
|
|
|
// covering the range [small,large].
|
2014-02-26 18:03:34 +00:00
|
|
|
void MakeTables(int cf, int n, const std::string& small,
|
|
|
|
const std::string& large) {
|
2014-02-01 00:45:20 +00:00
|
|
|
for (int i = 0; i < n; i++) {
|
|
|
|
ASSERT_OK(Put(cf, small, "begin"));
|
|
|
|
ASSERT_OK(Put(cf, large, "end"));
|
|
|
|
ASSERT_OK(db_->Flush(FlushOptions(), handles_[cf]));
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2015-06-29 21:39:01 +00:00
|
|
|
#ifndef ROCKSDB_LITE // GetSortedWalFiles is not supported
|
2014-02-25 18:38:04 +00:00
|
|
|
int CountLiveLogFiles() {
|
2014-02-27 18:29:13 +00:00
|
|
|
int micros_wait_for_log_deletion = 20000;
|
|
|
|
env_->SleepForMicroseconds(micros_wait_for_log_deletion);
|
2014-02-25 18:38:04 +00:00
|
|
|
int ret = 0;
|
|
|
|
VectorLogPtr wal_files;
|
2014-02-26 22:41:18 +00:00
|
|
|
Status s;
|
|
|
|
// GetSortedWalFiles is a flakey function -- it gets all the wal_dir
|
2015-04-25 09:14:27 +00:00
|
|
|
// children files and then later checks for their existence. if some of the
|
2014-02-26 22:41:18 +00:00
|
|
|
// log files doesn't exist anymore, it reports an error. it does all of this
|
|
|
|
// without DB mutex held, so if a background process deletes the log file
|
|
|
|
// while the function is being executed, it returns an error. We retry the
|
|
|
|
// function 10 times to avoid the error failing the test
|
|
|
|
for (int retries = 0; retries < 10; ++retries) {
|
|
|
|
wal_files.clear();
|
|
|
|
s = db_->GetSortedWalFiles(wal_files);
|
|
|
|
if (s.ok()) {
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
rocksdb: Replace ASSERT* with EXPECT* in functions that does not return void value
Summary:
gtest does not use exceptions to fail a unit test by design, and `ASSERT*`s are implemented using `return`. As a consequence we cannot use `ASSERT*` in a function that does not return `void` value ([[ https://code.google.com/p/googletest/wiki/AdvancedGuide#Assertion_Placement | 1]]), and have to fix our existing code. This diff does this in a generic way, with no manual changes.
In order to detect all existing `ASSERT*` that are used in functions that doesn't return void value, I change the code to generate compile errors for such cases.
In `util/testharness.h` I defined `EXPECT*` assertions, the same way as `ASSERT*`, and redefined `ASSERT*` to return `void`. Then executed:
```lang=bash
% USE_CLANG=1 make all -j55 -k 2> build.log
% perl -naF: -e 'print "-- -number=".$F[1]." ".$F[0]."\n" if /: error:/' \
build.log | xargs -L 1 perl -spi -e 's/ASSERT/EXPECT/g if $. == $number'
% make format
```
After that I reverted back change to `ASSERT*` in `util/testharness.h`. But preserved introduced `EXPECT*`, which is the same as `ASSERT*`. This will be deleted once switched to gtest.
This diff is independent and contains manual changes only in `util/testharness.h`.
Test Plan:
Make sure all tests are passing.
```lang=bash
% USE_CLANG=1 make check
```
Reviewers: igor, lgalanis, sdong, yufei.zhu, rven, meyering
Reviewed By: meyering
Subscribers: dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D33333
2015-03-17 03:52:32 +00:00
|
|
|
EXPECT_OK(s);
|
2014-02-25 18:38:04 +00:00
|
|
|
for (const auto& wal : wal_files) {
|
|
|
|
if (wal->Type() == kAliveLogFile) {
|
|
|
|
++ret;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
return ret;
|
2015-06-29 21:39:01 +00:00
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
#endif // !ROCKSDB_LITE
|
|
|
|
|
|
|
|
void AssertCountLiveLogFiles(int value) {
|
|
|
|
#ifndef ROCKSDB_LITE // GetSortedWalFiles is not supported
|
|
|
|
ASSERT_EQ(value, CountLiveLogFiles());
|
|
|
|
#endif // !ROCKSDB_LITE
|
2014-02-25 18:38:04 +00:00
|
|
|
}
|
|
|
|
|
2014-02-26 22:16:23 +00:00
|
|
|
void AssertNumberOfImmutableMemtables(std::vector<int> num_per_cf) {
|
|
|
|
assert(num_per_cf.size() == handles_.size());
|
|
|
|
|
2015-06-29 21:39:01 +00:00
|
|
|
#ifndef ROCKSDB_LITE // GetProperty is not supported in lite
|
2014-02-26 22:16:23 +00:00
|
|
|
for (size_t i = 0; i < num_per_cf.size(); ++i) {
|
2014-11-11 21:47:22 +00:00
|
|
|
ASSERT_EQ(num_per_cf[i], GetProperty(static_cast<int>(i),
|
|
|
|
"rocksdb.num-immutable-mem-table"));
|
2014-02-26 22:16:23 +00:00
|
|
|
}
|
2015-06-29 21:39:01 +00:00
|
|
|
#endif // !ROCKSDB_LITE
|
2014-02-26 22:16:23 +00:00
|
|
|
}
|
|
|
|
|
2014-02-26 18:03:34 +00:00
|
|
|
void CopyFile(const std::string& source, const std::string& destination,
|
2014-01-28 19:05:04 +00:00
|
|
|
uint64_t size = 0) {
|
|
|
|
const EnvOptions soptions;
|
|
|
|
unique_ptr<SequentialFile> srcfile;
|
|
|
|
ASSERT_OK(env_->NewSequentialFile(source, &srcfile, soptions));
|
|
|
|
unique_ptr<WritableFile> destfile;
|
|
|
|
ASSERT_OK(env_->NewWritableFile(destination, &destfile, soptions));
|
|
|
|
|
|
|
|
if (size == 0) {
|
|
|
|
// default argument means copy everything
|
|
|
|
ASSERT_OK(env_->GetFileSize(source, &size));
|
|
|
|
}
|
|
|
|
|
|
|
|
char buffer[4096];
|
|
|
|
Slice slice;
|
|
|
|
while (size > 0) {
|
2014-02-26 18:03:34 +00:00
|
|
|
uint64_t one = std::min(uint64_t(sizeof(buffer)), size);
|
2014-01-28 19:05:04 +00:00
|
|
|
ASSERT_OK(srcfile->Read(one, &slice, buffer));
|
|
|
|
ASSERT_OK(destfile->Append(slice));
|
|
|
|
size -= slice.size();
|
|
|
|
}
|
|
|
|
ASSERT_OK(destfile->Close());
|
|
|
|
}
|
|
|
|
|
2014-02-26 18:03:34 +00:00
|
|
|
std::vector<ColumnFamilyHandle*> handles_;
|
2014-02-26 22:16:23 +00:00
|
|
|
std::vector<std::string> names_;
|
Running manual compactions in parallel with other automatic or manual compactions in restricted cases
Summary:
This diff provides a framework for doing manual
compactions in parallel with other compactions. We now have a deque of manual compactions. We also pass manual compactions as an argument from RunManualCompactions down to
BackgroundCompactions, so that RunManualCompactions can be reentrant.
Parallelism is controlled by the two routines
ConflictingManualCompaction to allow/disallow new parallel/manual
compactions based on already existing ManualCompactions. In this diff, by default manual compactions still have to run exclusive of other compactions. However, by setting the compaction option, exclusive_manual_compaction to false, it is possible to run other compactions in parallel with a manual compaction. However, we are still restricted to one manual compaction per column family at a time. All of these restrictions will be relaxed in future diffs.
I will be adding more tests later.
Test Plan: Rocksdb regression + new tests + valgrind
Reviewers: igor, anthony, IslamAbdelRahman, kradhakrishnan, yhchiang, sdong
Reviewed By: sdong
Subscribers: yoshinorim, dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D47973
2015-12-14 19:20:34 +00:00
|
|
|
std::set<std::string> keys_;
|
2014-01-02 17:08:12 +00:00
|
|
|
ColumnFamilyOptions column_family_options_;
|
|
|
|
DBOptions db_options_;
|
2014-02-26 18:03:34 +00:00
|
|
|
std::string dbname_;
|
2014-02-01 00:45:20 +00:00
|
|
|
DB* db_ = nullptr;
|
2014-04-15 16:57:25 +00:00
|
|
|
EnvCounter* env_;
|
2014-02-25 18:38:04 +00:00
|
|
|
Random rnd_;
|
2014-01-02 17:08:12 +00:00
|
|
|
};
|
|
|
|
|
2015-03-17 21:08:00 +00:00
|
|
|
TEST_F(ColumnFamilyTest, DontReuseColumnFamilyID) {
|
2014-03-05 20:13:44 +00:00
|
|
|
for (int iter = 0; iter < 3; ++iter) {
|
|
|
|
Open();
|
|
|
|
CreateColumnFamilies({"one", "two", "three"});
|
|
|
|
for (size_t i = 0; i < handles_.size(); ++i) {
|
2014-03-14 18:26:13 +00:00
|
|
|
auto cfh = reinterpret_cast<ColumnFamilyHandleImpl*>(handles_[i]);
|
|
|
|
ASSERT_EQ(i, cfh->GetID());
|
2014-03-05 20:13:44 +00:00
|
|
|
}
|
|
|
|
if (iter == 1) {
|
|
|
|
Reopen();
|
|
|
|
}
|
|
|
|
DropColumnFamilies({3});
|
|
|
|
Reopen();
|
|
|
|
if (iter == 2) {
|
|
|
|
// this tests if max_column_family is correctly persisted with
|
|
|
|
// WriteSnapshot()
|
|
|
|
Reopen();
|
|
|
|
}
|
|
|
|
CreateColumnFamilies({"three2"});
|
|
|
|
// ID 3 that was used for dropped column family "three" should not be reused
|
2014-03-14 18:26:13 +00:00
|
|
|
auto cfh3 = reinterpret_cast<ColumnFamilyHandleImpl*>(handles_[3]);
|
2014-04-29 20:25:23 +00:00
|
|
|
ASSERT_EQ(4U, cfh3->GetID());
|
2014-03-05 20:13:44 +00:00
|
|
|
Close();
|
|
|
|
Destroy();
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
Ignore stale logs while restarting DBs
Summary:
Stale log files can be deleted out of order. This can happen for various reasons. One of the reason is that no data is ever inserted to a column family and we have an optimization to update its log number, but not all the old log files are cleaned up (the case shown in the unit tests added). It can also happen when we simply delete multiple log files out of order.
This causes data corruption because we simply increase seqID after processing the next row and we may end up with writing data with smaller seqID than what is already flushed to memtables.
In DB recovery, for the oldest files we are replaying, if there it contains no data for any column family, we ignore the sequence IDs in the file.
Test Plan: Add two unit tests that fail without the fix.
Reviewers: IslamAbdelRahman, igor, yiwu
Reviewed By: yiwu
Subscribers: hermanlee4, yoshinorim, leveldb, andrewkr, dhruba
Differential Revision: https://reviews.facebook.net/D60891
2016-07-19 18:48:00 +00:00
|
|
|
class FlushEmptyCFTestWithParam : public ColumnFamilyTest,
|
|
|
|
public testing::WithParamInterface<bool> {
|
|
|
|
public:
|
|
|
|
FlushEmptyCFTestWithParam() { allow_2pc_ = GetParam(); }
|
|
|
|
|
|
|
|
// Required if inheriting from testing::WithParamInterface<>
|
|
|
|
static void SetUpTestCase() {}
|
|
|
|
static void TearDownTestCase() {}
|
|
|
|
|
|
|
|
bool allow_2pc_;
|
|
|
|
};
|
|
|
|
|
|
|
|
TEST_P(FlushEmptyCFTestWithParam, FlushEmptyCFTest) {
|
|
|
|
std::unique_ptr<FaultInjectionTestEnv> fault_env(
|
|
|
|
new FaultInjectionTestEnv(env_));
|
|
|
|
db_options_.env = fault_env.get();
|
|
|
|
db_options_.allow_2pc = allow_2pc_;
|
|
|
|
Open();
|
|
|
|
CreateColumnFamilies({"one", "two"});
|
|
|
|
// Generate log file A.
|
|
|
|
ASSERT_OK(Put(1, "foo", "v1")); // seqID 1
|
|
|
|
|
|
|
|
Reopen();
|
|
|
|
// Log file A is not dropped after reopening because default column family's
|
|
|
|
// min log number is 0.
|
|
|
|
// It flushes to SST file X
|
|
|
|
ASSERT_OK(Put(1, "foo", "v1")); // seqID 2
|
|
|
|
ASSERT_OK(Put(1, "bar", "v2")); // seqID 3
|
|
|
|
// Current log file is file B now. While flushing, a new log file C is created
|
|
|
|
// and is set to current. Boths' min log number is set to file C in memory, so
|
|
|
|
// after flushing file B is deleted. At the same time, the min log number of
|
|
|
|
// default CF is not written to manifest. Log file A still remains.
|
|
|
|
// Flushed to SST file Y.
|
|
|
|
Flush(1);
|
|
|
|
Flush(0);
|
|
|
|
ASSERT_OK(Put(1, "bar", "v3")); // seqID 4
|
|
|
|
ASSERT_OK(Put(1, "foo", "v4")); // seqID 5
|
|
|
|
|
|
|
|
// Preserve file system state up to here to simulate a crash condition.
|
|
|
|
fault_env->SetFilesystemActive(false);
|
|
|
|
std::vector<std::string> names;
|
|
|
|
for (auto name : names_) {
|
|
|
|
if (name != "") {
|
|
|
|
names.push_back(name);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
Close();
|
|
|
|
fault_env->ResetState();
|
|
|
|
|
|
|
|
// Before opening, there are four files:
|
|
|
|
// Log file A contains seqID 1
|
|
|
|
// Log file C contains seqID 4, 5
|
|
|
|
// SST file X contains seqID 1
|
|
|
|
// SST file Y contains seqID 2, 3
|
|
|
|
// Min log number:
|
|
|
|
// default CF: 0
|
|
|
|
// CF one, two: C
|
|
|
|
// When opening the DB, all the seqID should be preserved.
|
|
|
|
Open(names, {});
|
|
|
|
ASSERT_EQ("v4", Get(1, "foo"));
|
|
|
|
ASSERT_EQ("v3", Get(1, "bar"));
|
|
|
|
Close();
|
|
|
|
|
|
|
|
db_options_.env = env_;
|
|
|
|
}
|
|
|
|
|
|
|
|
TEST_P(FlushEmptyCFTestWithParam, FlushEmptyCFTest2) {
|
|
|
|
std::unique_ptr<FaultInjectionTestEnv> fault_env(
|
|
|
|
new FaultInjectionTestEnv(env_));
|
|
|
|
db_options_.env = fault_env.get();
|
|
|
|
db_options_.allow_2pc = allow_2pc_;
|
|
|
|
Open();
|
|
|
|
CreateColumnFamilies({"one", "two"});
|
|
|
|
// Generate log file A.
|
|
|
|
ASSERT_OK(Put(1, "foo", "v1")); // seqID 1
|
|
|
|
|
|
|
|
Reopen();
|
|
|
|
// Log file A is not dropped after reopening because default column family's
|
|
|
|
// min log number is 0.
|
|
|
|
// It flushes to SST file X
|
|
|
|
ASSERT_OK(Put(1, "foo", "v1")); // seqID 2
|
|
|
|
ASSERT_OK(Put(1, "bar", "v2")); // seqID 3
|
|
|
|
// Current log file is file B now. While flushing, a new log file C is created
|
|
|
|
// and is set to current. Both CFs' min log number is set to file C so after
|
|
|
|
// flushing file B is deleted. Log file A still remains.
|
|
|
|
// Flushed to SST file Y.
|
|
|
|
Flush(1);
|
|
|
|
ASSERT_OK(Put(0, "bar", "v2")); // seqID 4
|
|
|
|
ASSERT_OK(Put(2, "bar", "v2")); // seqID 5
|
|
|
|
ASSERT_OK(Put(1, "bar", "v3")); // seqID 6
|
|
|
|
// Flushing all column families. This forces all CFs' min log to current. This
|
|
|
|
// is written to the manifest file. Log file C is cleared.
|
|
|
|
Flush(0);
|
|
|
|
Flush(1);
|
|
|
|
Flush(2);
|
|
|
|
// Write to log file D
|
|
|
|
ASSERT_OK(Put(1, "bar", "v4")); // seqID 7
|
|
|
|
ASSERT_OK(Put(1, "bar", "v5")); // seqID 8
|
|
|
|
// Preserve file system state up to here to simulate a crash condition.
|
|
|
|
fault_env->SetFilesystemActive(false);
|
|
|
|
std::vector<std::string> names;
|
|
|
|
for (auto name : names_) {
|
|
|
|
if (name != "") {
|
|
|
|
names.push_back(name);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
Close();
|
|
|
|
fault_env->ResetState();
|
|
|
|
// Before opening, there are two logfiles:
|
|
|
|
// Log file A contains seqID 1
|
|
|
|
// Log file D contains seqID 7, 8
|
|
|
|
// Min log number:
|
|
|
|
// default CF: D
|
|
|
|
// CF one, two: D
|
|
|
|
// When opening the DB, log file D should be replayed using the seqID
|
|
|
|
// specified in the file.
|
|
|
|
Open(names, {});
|
|
|
|
ASSERT_EQ("v1", Get(1, "foo"));
|
|
|
|
ASSERT_EQ("v5", Get(1, "bar"));
|
|
|
|
Close();
|
|
|
|
|
|
|
|
db_options_.env = env_;
|
|
|
|
}
|
|
|
|
|
|
|
|
INSTANTIATE_TEST_CASE_P(FlushEmptyCFTestWithParam, FlushEmptyCFTestWithParam,
|
|
|
|
::testing::Bool());
|
|
|
|
|
2015-03-17 21:08:00 +00:00
|
|
|
TEST_F(ColumnFamilyTest, AddDrop) {
|
2014-02-26 22:16:23 +00:00
|
|
|
Open();
|
2014-02-11 01:04:44 +00:00
|
|
|
CreateColumnFamilies({"one", "two", "three"});
|
2014-02-28 18:29:37 +00:00
|
|
|
ASSERT_EQ("NOT_FOUND", Get(1, "fodor"));
|
|
|
|
ASSERT_EQ("NOT_FOUND", Get(2, "fodor"));
|
2014-02-11 01:04:44 +00:00
|
|
|
DropColumnFamilies({2});
|
2014-02-28 18:29:37 +00:00
|
|
|
ASSERT_EQ("NOT_FOUND", Get(1, "fodor"));
|
2014-02-11 01:04:44 +00:00
|
|
|
CreateColumnFamilies({"four"});
|
2014-02-28 18:29:37 +00:00
|
|
|
ASSERT_EQ("NOT_FOUND", Get(3, "fodor"));
|
|
|
|
ASSERT_OK(Put(1, "fodor", "mirko"));
|
|
|
|
ASSERT_EQ("mirko", Get(1, "fodor"));
|
|
|
|
ASSERT_EQ("NOT_FOUND", Get(3, "fodor"));
|
2014-01-02 17:08:12 +00:00
|
|
|
Close();
|
2014-02-26 22:16:23 +00:00
|
|
|
ASSERT_TRUE(TryOpen({"default"}).IsInvalidArgument());
|
|
|
|
Open({"default", "one", "three", "four"});
|
|
|
|
DropColumnFamilies({1});
|
|
|
|
Reopen();
|
2014-01-02 17:08:12 +00:00
|
|
|
Close();
|
|
|
|
|
2014-02-26 18:03:34 +00:00
|
|
|
std::vector<std::string> families;
|
2014-01-22 19:44:53 +00:00
|
|
|
ASSERT_OK(DB::ListColumnFamilies(db_options_, dbname_, &families));
|
2016-05-20 14:42:18 +00:00
|
|
|
std::sort(families.begin(), families.end());
|
2014-02-26 18:03:34 +00:00
|
|
|
ASSERT_TRUE(families ==
|
2014-02-26 22:16:23 +00:00
|
|
|
std::vector<std::string>({"default", "four", "three"}));
|
2014-01-02 17:08:12 +00:00
|
|
|
}
|
|
|
|
|
2015-03-17 21:08:00 +00:00
|
|
|
TEST_F(ColumnFamilyTest, DropTest) {
|
2014-02-11 01:04:44 +00:00
|
|
|
// first iteration - dont reopen DB before dropping
|
|
|
|
// second iteration - reopen DB before dropping
|
|
|
|
for (int iter = 0; iter < 2; ++iter) {
|
2014-02-26 22:16:23 +00:00
|
|
|
Open({"default"});
|
|
|
|
CreateColumnFamiliesAndReopen({"pikachu"});
|
2014-02-11 01:04:44 +00:00
|
|
|
for (int i = 0; i < 100; ++i) {
|
2014-11-25 04:44:49 +00:00
|
|
|
ASSERT_OK(Put(1, ToString(i), "bar" + ToString(i)));
|
2014-02-11 01:04:44 +00:00
|
|
|
}
|
|
|
|
ASSERT_OK(Flush(1));
|
|
|
|
|
|
|
|
if (iter == 1) {
|
2014-02-26 22:16:23 +00:00
|
|
|
Reopen();
|
2014-02-11 01:04:44 +00:00
|
|
|
}
|
|
|
|
ASSERT_EQ("bar1", Get(1, "1"));
|
|
|
|
|
2015-06-29 21:39:01 +00:00
|
|
|
AssertCountLiveFiles(1);
|
2014-02-11 01:04:44 +00:00
|
|
|
DropColumnFamilies({1});
|
|
|
|
// make sure that all files are deleted when we drop the column family
|
2015-06-29 21:39:01 +00:00
|
|
|
AssertCountLiveFiles(0);
|
2014-02-11 01:04:44 +00:00
|
|
|
Destroy();
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2015-03-17 21:08:00 +00:00
|
|
|
TEST_F(ColumnFamilyTest, WriteBatchFailure) {
|
2014-02-26 01:30:54 +00:00
|
|
|
Open();
|
2014-03-14 18:26:13 +00:00
|
|
|
CreateColumnFamiliesAndReopen({"one", "two"});
|
2014-02-26 01:30:54 +00:00
|
|
|
WriteBatch batch;
|
2014-09-02 20:29:05 +00:00
|
|
|
batch.Put(handles_[0], Slice("existing"), Slice("column-family"));
|
2014-03-14 18:26:13 +00:00
|
|
|
batch.Put(handles_[1], Slice("non-existing"), Slice("column-family"));
|
|
|
|
ASSERT_OK(db_->Write(WriteOptions(), &batch));
|
|
|
|
DropColumnFamilies({1});
|
2014-09-02 20:29:05 +00:00
|
|
|
WriteOptions woptions_ignore_missing_cf;
|
|
|
|
woptions_ignore_missing_cf.ignore_missing_column_families = true;
|
|
|
|
batch.Put(handles_[0], Slice("still here"), Slice("column-family"));
|
|
|
|
ASSERT_OK(db_->Write(woptions_ignore_missing_cf, &batch));
|
|
|
|
ASSERT_EQ("column-family", Get(0, "still here"));
|
2014-02-26 01:30:54 +00:00
|
|
|
Status s = db_->Write(WriteOptions(), &batch);
|
|
|
|
ASSERT_TRUE(s.IsInvalidArgument());
|
|
|
|
Close();
|
|
|
|
}
|
|
|
|
|
2015-03-17 21:08:00 +00:00
|
|
|
TEST_F(ColumnFamilyTest, ReadWrite) {
|
2014-02-26 22:16:23 +00:00
|
|
|
Open();
|
|
|
|
CreateColumnFamiliesAndReopen({"one", "two"});
|
2014-01-28 19:05:04 +00:00
|
|
|
ASSERT_OK(Put(0, "foo", "v1"));
|
|
|
|
ASSERT_OK(Put(0, "bar", "v2"));
|
|
|
|
ASSERT_OK(Put(1, "mirko", "v3"));
|
|
|
|
ASSERT_OK(Put(0, "foo", "v2"));
|
|
|
|
ASSERT_OK(Put(2, "fodor", "v5"));
|
|
|
|
|
|
|
|
for (int iter = 0; iter <= 3; ++iter) {
|
|
|
|
ASSERT_EQ("v2", Get(0, "foo"));
|
|
|
|
ASSERT_EQ("v2", Get(0, "bar"));
|
|
|
|
ASSERT_EQ("v3", Get(1, "mirko"));
|
|
|
|
ASSERT_EQ("v5", Get(2, "fodor"));
|
|
|
|
ASSERT_EQ("NOT_FOUND", Get(0, "fodor"));
|
|
|
|
ASSERT_EQ("NOT_FOUND", Get(1, "fodor"));
|
|
|
|
ASSERT_EQ("NOT_FOUND", Get(2, "foo"));
|
|
|
|
if (iter <= 1) {
|
2014-02-26 22:16:23 +00:00
|
|
|
Reopen();
|
2014-01-28 19:05:04 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
Close();
|
|
|
|
}
|
|
|
|
|
2015-03-17 21:08:00 +00:00
|
|
|
TEST_F(ColumnFamilyTest, IgnoreRecoveredLog) {
|
2014-02-26 18:03:34 +00:00
|
|
|
std::string backup_logs = dbname_ + "/backup_logs";
|
2014-01-28 19:05:04 +00:00
|
|
|
|
|
|
|
// delete old files in backup_logs directory
|
2014-01-29 22:06:59 +00:00
|
|
|
ASSERT_OK(env_->CreateDirIfMissing(dbname_));
|
|
|
|
ASSERT_OK(env_->CreateDirIfMissing(backup_logs));
|
2014-02-26 18:03:34 +00:00
|
|
|
std::vector<std::string> old_files;
|
2014-01-28 19:05:04 +00:00
|
|
|
env_->GetChildren(backup_logs, &old_files);
|
|
|
|
for (auto& file : old_files) {
|
|
|
|
if (file != "." && file != "..") {
|
|
|
|
env_->DeleteFile(backup_logs + "/" + file);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
column_family_options_.merge_operator =
|
|
|
|
MergeOperators::CreateUInt64AddOperator();
|
|
|
|
db_options_.wal_dir = dbname_ + "/logs";
|
|
|
|
Destroy();
|
2014-02-26 22:16:23 +00:00
|
|
|
Open();
|
2014-01-28 19:05:04 +00:00
|
|
|
CreateColumnFamilies({"cf1", "cf2"});
|
|
|
|
|
|
|
|
// fill up the DB
|
2014-02-26 18:03:34 +00:00
|
|
|
std::string one, two, three;
|
2014-01-28 19:05:04 +00:00
|
|
|
PutFixed64(&one, 1);
|
|
|
|
PutFixed64(&two, 2);
|
|
|
|
PutFixed64(&three, 3);
|
|
|
|
ASSERT_OK(Merge(0, "foo", one));
|
|
|
|
ASSERT_OK(Merge(1, "mirko", one));
|
|
|
|
ASSERT_OK(Merge(0, "foo", one));
|
|
|
|
ASSERT_OK(Merge(2, "bla", one));
|
|
|
|
ASSERT_OK(Merge(2, "fodor", one));
|
|
|
|
ASSERT_OK(Merge(0, "bar", one));
|
|
|
|
ASSERT_OK(Merge(2, "bla", one));
|
|
|
|
ASSERT_OK(Merge(1, "mirko", two));
|
|
|
|
ASSERT_OK(Merge(1, "franjo", one));
|
|
|
|
|
|
|
|
// copy the logs to backup
|
2014-02-26 18:03:34 +00:00
|
|
|
std::vector<std::string> logs;
|
2014-01-28 19:05:04 +00:00
|
|
|
env_->GetChildren(db_options_.wal_dir, &logs);
|
|
|
|
for (auto& log : logs) {
|
|
|
|
if (log != ".." && log != ".") {
|
|
|
|
CopyFile(db_options_.wal_dir + "/" + log, backup_logs + "/" + log);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
// recover the DB
|
|
|
|
Close();
|
|
|
|
|
|
|
|
// 1. check consistency
|
|
|
|
// 2. copy the logs from backup back to WAL dir. if the recovery happens
|
|
|
|
// again on the same log files, this should lead to incorrect results
|
|
|
|
// due to applying merge operator twice
|
|
|
|
// 3. check consistency
|
|
|
|
for (int iter = 0; iter < 2; ++iter) {
|
|
|
|
// assert consistency
|
2014-02-26 22:16:23 +00:00
|
|
|
Open({"default", "cf1", "cf2"});
|
2014-01-28 19:05:04 +00:00
|
|
|
ASSERT_EQ(two, Get(0, "foo"));
|
|
|
|
ASSERT_EQ(one, Get(0, "bar"));
|
|
|
|
ASSERT_EQ(three, Get(1, "mirko"));
|
|
|
|
ASSERT_EQ(one, Get(1, "franjo"));
|
|
|
|
ASSERT_EQ(one, Get(2, "fodor"));
|
|
|
|
ASSERT_EQ(two, Get(2, "bla"));
|
|
|
|
Close();
|
|
|
|
|
|
|
|
if (iter == 0) {
|
|
|
|
// copy the logs from backup back to wal dir
|
|
|
|
for (auto& log : logs) {
|
|
|
|
if (log != ".." && log != ".") {
|
|
|
|
CopyFile(backup_logs + "/" + log, db_options_.wal_dir + "/" + log);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2015-03-17 21:08:00 +00:00
|
|
|
TEST_F(ColumnFamilyTest, FlushTest) {
|
2014-02-26 22:16:23 +00:00
|
|
|
Open();
|
|
|
|
CreateColumnFamiliesAndReopen({"one", "two"});
|
2014-01-31 01:48:42 +00:00
|
|
|
ASSERT_OK(Put(0, "foo", "v1"));
|
|
|
|
ASSERT_OK(Put(0, "bar", "v2"));
|
|
|
|
ASSERT_OK(Put(1, "mirko", "v3"));
|
|
|
|
ASSERT_OK(Put(0, "foo", "v2"));
|
|
|
|
ASSERT_OK(Put(2, "fodor", "v5"));
|
2014-12-08 20:52:18 +00:00
|
|
|
|
|
|
|
for (int j = 0; j < 2; j++) {
|
|
|
|
ReadOptions ro;
|
|
|
|
std::vector<Iterator*> iterators;
|
|
|
|
// Hold super version.
|
|
|
|
if (j == 0) {
|
|
|
|
ASSERT_OK(db_->NewIterators(ro, handles_, &iterators));
|
|
|
|
}
|
|
|
|
|
|
|
|
for (int i = 0; i < 3; ++i) {
|
|
|
|
uint64_t max_total_in_memory_state =
|
2015-06-29 21:39:01 +00:00
|
|
|
MaxTotalInMemoryState();
|
2014-12-08 20:52:18 +00:00
|
|
|
Flush(i);
|
2015-06-29 21:39:01 +00:00
|
|
|
AssertMaxTotalInMemoryState(max_total_in_memory_state);
|
2014-12-08 20:52:18 +00:00
|
|
|
}
|
|
|
|
ASSERT_OK(Put(1, "foofoo", "bar"));
|
|
|
|
ASSERT_OK(Put(0, "foofoo", "bar"));
|
|
|
|
|
|
|
|
for (auto* it : iterators) {
|
|
|
|
delete it;
|
|
|
|
}
|
2014-01-31 01:48:42 +00:00
|
|
|
}
|
2014-02-26 22:16:23 +00:00
|
|
|
Reopen();
|
2014-01-31 01:48:42 +00:00
|
|
|
|
|
|
|
for (int iter = 0; iter <= 2; ++iter) {
|
|
|
|
ASSERT_EQ("v2", Get(0, "foo"));
|
|
|
|
ASSERT_EQ("v2", Get(0, "bar"));
|
|
|
|
ASSERT_EQ("v3", Get(1, "mirko"));
|
|
|
|
ASSERT_EQ("v5", Get(2, "fodor"));
|
|
|
|
ASSERT_EQ("NOT_FOUND", Get(0, "fodor"));
|
|
|
|
ASSERT_EQ("NOT_FOUND", Get(1, "fodor"));
|
|
|
|
ASSERT_EQ("NOT_FOUND", Get(2, "foo"));
|
|
|
|
if (iter <= 1) {
|
2014-02-26 22:16:23 +00:00
|
|
|
Reopen();
|
2014-01-31 01:48:42 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
Close();
|
|
|
|
}
|
|
|
|
|
2014-02-25 18:38:04 +00:00
|
|
|
// Makes sure that obsolete log files get deleted
|
2015-03-17 21:08:00 +00:00
|
|
|
TEST_F(ColumnFamilyTest, LogDeletionTest) {
|
2014-04-30 18:33:40 +00:00
|
|
|
db_options_.max_total_wal_size = std::numeric_limits<uint64_t>::max();
|
2015-08-26 21:19:31 +00:00
|
|
|
column_family_options_.arena_block_size = 4 * 1024;
|
2014-02-25 18:38:04 +00:00
|
|
|
column_family_options_.write_buffer_size = 100000; // 100KB
|
2014-02-26 22:16:23 +00:00
|
|
|
Open();
|
2014-02-25 18:38:04 +00:00
|
|
|
CreateColumnFamilies({"one", "two", "three", "four"});
|
|
|
|
// Each bracket is one log file. if number is in (), it means
|
|
|
|
// we don't need it anymore (it's been flushed)
|
|
|
|
// []
|
2015-06-29 21:39:01 +00:00
|
|
|
AssertCountLiveLogFiles(0);
|
2014-02-26 22:16:23 +00:00
|
|
|
PutRandomData(0, 1, 100);
|
2014-02-25 18:38:04 +00:00
|
|
|
// [0]
|
2014-02-26 22:16:23 +00:00
|
|
|
PutRandomData(1, 1, 100);
|
2014-02-25 18:38:04 +00:00
|
|
|
// [0, 1]
|
2014-02-26 22:16:23 +00:00
|
|
|
PutRandomData(1, 1000, 100);
|
2014-02-25 18:38:04 +00:00
|
|
|
WaitForFlush(1);
|
|
|
|
// [0, (1)] [1]
|
2015-06-29 21:39:01 +00:00
|
|
|
AssertCountLiveLogFiles(2);
|
2014-02-26 22:16:23 +00:00
|
|
|
PutRandomData(0, 1, 100);
|
2014-02-25 18:38:04 +00:00
|
|
|
// [0, (1)] [0, 1]
|
2015-06-29 21:39:01 +00:00
|
|
|
AssertCountLiveLogFiles(2);
|
2014-02-26 22:16:23 +00:00
|
|
|
PutRandomData(2, 1, 100);
|
2014-02-25 18:38:04 +00:00
|
|
|
// [0, (1)] [0, 1, 2]
|
2014-02-26 22:16:23 +00:00
|
|
|
PutRandomData(2, 1000, 100);
|
2014-02-25 18:38:04 +00:00
|
|
|
WaitForFlush(2);
|
|
|
|
// [0, (1)] [0, 1, (2)] [2]
|
2015-06-29 21:39:01 +00:00
|
|
|
AssertCountLiveLogFiles(3);
|
2014-02-26 22:16:23 +00:00
|
|
|
PutRandomData(2, 1000, 100);
|
2014-02-25 18:38:04 +00:00
|
|
|
WaitForFlush(2);
|
|
|
|
// [0, (1)] [0, 1, (2)] [(2)] [2]
|
2015-06-29 21:39:01 +00:00
|
|
|
AssertCountLiveLogFiles(4);
|
2014-02-26 22:16:23 +00:00
|
|
|
PutRandomData(3, 1, 100);
|
2014-02-25 18:38:04 +00:00
|
|
|
// [0, (1)] [0, 1, (2)] [(2)] [2, 3]
|
2014-02-26 22:16:23 +00:00
|
|
|
PutRandomData(1, 1, 100);
|
2014-02-25 18:38:04 +00:00
|
|
|
// [0, (1)] [0, 1, (2)] [(2)] [1, 2, 3]
|
2015-06-29 21:39:01 +00:00
|
|
|
AssertCountLiveLogFiles(4);
|
2014-02-26 22:16:23 +00:00
|
|
|
PutRandomData(1, 1000, 100);
|
2014-02-25 18:38:04 +00:00
|
|
|
WaitForFlush(1);
|
|
|
|
// [0, (1)] [0, (1), (2)] [(2)] [(1), 2, 3] [1]
|
2015-06-29 21:39:01 +00:00
|
|
|
AssertCountLiveLogFiles(5);
|
2014-02-26 22:16:23 +00:00
|
|
|
PutRandomData(0, 1000, 100);
|
2014-02-25 18:38:04 +00:00
|
|
|
WaitForFlush(0);
|
|
|
|
// [(0), (1)] [(0), (1), (2)] [(2)] [(1), 2, 3] [1, (0)] [0]
|
|
|
|
// delete obsolete logs -->
|
|
|
|
// [(1), 2, 3] [1, (0)] [0]
|
2015-06-29 21:39:01 +00:00
|
|
|
AssertCountLiveLogFiles(3);
|
2014-02-26 22:16:23 +00:00
|
|
|
PutRandomData(0, 1000, 100);
|
2014-02-25 18:38:04 +00:00
|
|
|
WaitForFlush(0);
|
|
|
|
// [(1), 2, 3] [1, (0)], [(0)] [0]
|
2015-06-29 21:39:01 +00:00
|
|
|
AssertCountLiveLogFiles(4);
|
2014-02-26 22:16:23 +00:00
|
|
|
PutRandomData(1, 1000, 100);
|
2014-02-25 18:38:04 +00:00
|
|
|
WaitForFlush(1);
|
|
|
|
// [(1), 2, 3] [(1), (0)] [(0)] [0, (1)] [1]
|
2015-06-29 21:39:01 +00:00
|
|
|
AssertCountLiveLogFiles(5);
|
2014-02-26 22:16:23 +00:00
|
|
|
PutRandomData(2, 1000, 100);
|
2014-02-25 18:38:04 +00:00
|
|
|
WaitForFlush(2);
|
|
|
|
// [(1), (2), 3] [(1), (0)] [(0)] [0, (1)] [1, (2)], [2]
|
2015-06-29 21:39:01 +00:00
|
|
|
AssertCountLiveLogFiles(6);
|
2014-02-26 22:16:23 +00:00
|
|
|
PutRandomData(3, 1000, 100);
|
2014-02-25 18:38:04 +00:00
|
|
|
WaitForFlush(3);
|
|
|
|
// [(1), (2), (3)] [(1), (0)] [(0)] [0, (1)] [1, (2)], [2, (3)] [3]
|
|
|
|
// delete obsolete logs -->
|
|
|
|
// [0, (1)] [1, (2)], [2, (3)] [3]
|
2015-06-29 21:39:01 +00:00
|
|
|
AssertCountLiveLogFiles(4);
|
2014-02-25 18:38:04 +00:00
|
|
|
Close();
|
|
|
|
}
|
|
|
|
|
2016-07-19 22:12:46 +00:00
|
|
|
TEST_F(ColumnFamilyTest, CrashAfterFlush) {
|
|
|
|
std::unique_ptr<FaultInjectionTestEnv> fault_env(
|
|
|
|
new FaultInjectionTestEnv(env_));
|
|
|
|
db_options_.env = fault_env.get();
|
|
|
|
Open();
|
|
|
|
CreateColumnFamilies({"one"});
|
|
|
|
|
|
|
|
WriteBatch batch;
|
|
|
|
batch.Put(handles_[0], Slice("foo"), Slice("bar"));
|
|
|
|
batch.Put(handles_[1], Slice("foo"), Slice("bar"));
|
|
|
|
ASSERT_OK(db_->Write(WriteOptions(), &batch));
|
|
|
|
Flush(0);
|
|
|
|
fault_env->SetFilesystemActive(false);
|
|
|
|
|
|
|
|
std::vector<std::string> names;
|
|
|
|
for (auto name : names_) {
|
|
|
|
if (name != "") {
|
|
|
|
names.push_back(name);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
Close();
|
|
|
|
fault_env->DropUnsyncedFileData();
|
|
|
|
fault_env->ResetState();
|
|
|
|
Open(names, {});
|
|
|
|
|
|
|
|
// Write batch should be atomic.
|
|
|
|
ASSERT_EQ(Get(0, "foo"), Get(1, "foo"));
|
|
|
|
|
|
|
|
Close();
|
|
|
|
db_options_.env = env_;
|
|
|
|
}
|
|
|
|
|
2014-02-26 22:16:23 +00:00
|
|
|
// Makes sure that obsolete log files get deleted
|
2015-03-17 21:08:00 +00:00
|
|
|
TEST_F(ColumnFamilyTest, DifferentWriteBufferSizes) {
|
2014-04-30 18:33:40 +00:00
|
|
|
// disable flushing stale column families
|
|
|
|
db_options_.max_total_wal_size = std::numeric_limits<uint64_t>::max();
|
2014-02-26 22:16:23 +00:00
|
|
|
Open();
|
|
|
|
CreateColumnFamilies({"one", "two", "three"});
|
|
|
|
ColumnFamilyOptions default_cf, one, two, three;
|
|
|
|
// setup options. all column families have max_write_buffer_number setup to 10
|
|
|
|
// "default" -> 100KB memtable, start flushing immediatelly
|
|
|
|
// "one" -> 200KB memtable, start flushing with two immutable memtables
|
|
|
|
// "two" -> 1MB memtable, start flushing with three immutable memtables
|
|
|
|
// "three" -> 90KB memtable, start flushing with four immutable memtables
|
|
|
|
default_cf.write_buffer_size = 100000;
|
2015-08-26 21:19:31 +00:00
|
|
|
default_cf.arena_block_size = 4 * 4096;
|
2014-02-26 22:16:23 +00:00
|
|
|
default_cf.max_write_buffer_number = 10;
|
|
|
|
default_cf.min_write_buffer_number_to_merge = 1;
|
Support saving history in memtable_list
Summary:
For transactions, we are using the memtables to validate that there are no write conflicts. But after flushing, we don't have any memtables, and transactions could fail to commit. So we want to someone keep around some extra history to use for conflict checking. In addition, we want to provide a way to increase the size of this history if too many transactions fail to commit.
After chatting with people, it seems like everyone prefers just using Memtables to store this history (instead of a separate history structure). It seems like the best place for this is abstracted inside the memtable_list. I decide to create a separate list in MemtableListVersion as using the same list complicated the flush/installalflushresults logic too much.
This diff adds a new parameter to control how much memtable history to keep around after flushing. However, it sounds like people aren't too fond of adding new parameters. So I am making the default size of flushed+not-flushed memtables be set to max_write_buffers. This should not change the maximum amount of memory used, but make it more likely we're using closer the the limit. (We are now postponing deleting flushed memtables until the max_write_buffer limit is reached). So while we might use more memory on average, we are still obeying the limit set (and you could argue it's better to go ahead and use up memory now instead of waiting for a write stall to happen to test this limit).
However, if people are opposed to this default behavior, we can easily set it to 0 and require this parameter be set in order to use transactions.
Test Plan: Added a xfunc test to play around with setting different values of this parameter in all tests. Added testing in memtablelist_test and planning on adding more testing here.
Reviewers: sdong, rven, igor
Reviewed By: igor
Subscribers: dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D37443
2015-05-28 23:34:24 +00:00
|
|
|
default_cf.max_write_buffer_number_to_maintain = 0;
|
2014-02-26 22:16:23 +00:00
|
|
|
one.write_buffer_size = 200000;
|
2015-08-26 21:19:31 +00:00
|
|
|
one.arena_block_size = 4 * 4096;
|
2014-02-26 22:16:23 +00:00
|
|
|
one.max_write_buffer_number = 10;
|
|
|
|
one.min_write_buffer_number_to_merge = 2;
|
Support saving history in memtable_list
Summary:
For transactions, we are using the memtables to validate that there are no write conflicts. But after flushing, we don't have any memtables, and transactions could fail to commit. So we want to someone keep around some extra history to use for conflict checking. In addition, we want to provide a way to increase the size of this history if too many transactions fail to commit.
After chatting with people, it seems like everyone prefers just using Memtables to store this history (instead of a separate history structure). It seems like the best place for this is abstracted inside the memtable_list. I decide to create a separate list in MemtableListVersion as using the same list complicated the flush/installalflushresults logic too much.
This diff adds a new parameter to control how much memtable history to keep around after flushing. However, it sounds like people aren't too fond of adding new parameters. So I am making the default size of flushed+not-flushed memtables be set to max_write_buffers. This should not change the maximum amount of memory used, but make it more likely we're using closer the the limit. (We are now postponing deleting flushed memtables until the max_write_buffer limit is reached). So while we might use more memory on average, we are still obeying the limit set (and you could argue it's better to go ahead and use up memory now instead of waiting for a write stall to happen to test this limit).
However, if people are opposed to this default behavior, we can easily set it to 0 and require this parameter be set in order to use transactions.
Test Plan: Added a xfunc test to play around with setting different values of this parameter in all tests. Added testing in memtablelist_test and planning on adding more testing here.
Reviewers: sdong, rven, igor
Reviewed By: igor
Subscribers: dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D37443
2015-05-28 23:34:24 +00:00
|
|
|
one.max_write_buffer_number_to_maintain = 1;
|
2014-02-26 22:16:23 +00:00
|
|
|
two.write_buffer_size = 1000000;
|
2015-08-26 21:19:31 +00:00
|
|
|
two.arena_block_size = 4 * 4096;
|
2014-02-26 22:16:23 +00:00
|
|
|
two.max_write_buffer_number = 10;
|
|
|
|
two.min_write_buffer_number_to_merge = 3;
|
Support saving history in memtable_list
Summary:
For transactions, we are using the memtables to validate that there are no write conflicts. But after flushing, we don't have any memtables, and transactions could fail to commit. So we want to someone keep around some extra history to use for conflict checking. In addition, we want to provide a way to increase the size of this history if too many transactions fail to commit.
After chatting with people, it seems like everyone prefers just using Memtables to store this history (instead of a separate history structure). It seems like the best place for this is abstracted inside the memtable_list. I decide to create a separate list in MemtableListVersion as using the same list complicated the flush/installalflushresults logic too much.
This diff adds a new parameter to control how much memtable history to keep around after flushing. However, it sounds like people aren't too fond of adding new parameters. So I am making the default size of flushed+not-flushed memtables be set to max_write_buffers. This should not change the maximum amount of memory used, but make it more likely we're using closer the the limit. (We are now postponing deleting flushed memtables until the max_write_buffer limit is reached). So while we might use more memory on average, we are still obeying the limit set (and you could argue it's better to go ahead and use up memory now instead of waiting for a write stall to happen to test this limit).
However, if people are opposed to this default behavior, we can easily set it to 0 and require this parameter be set in order to use transactions.
Test Plan: Added a xfunc test to play around with setting different values of this parameter in all tests. Added testing in memtablelist_test and planning on adding more testing here.
Reviewers: sdong, rven, igor
Reviewed By: igor
Subscribers: dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D37443
2015-05-28 23:34:24 +00:00
|
|
|
two.max_write_buffer_number_to_maintain = 2;
|
2015-12-07 18:13:58 +00:00
|
|
|
three.write_buffer_size = 4096 * 22;
|
2015-08-26 21:19:31 +00:00
|
|
|
three.arena_block_size = 4096;
|
2014-02-26 22:16:23 +00:00
|
|
|
three.max_write_buffer_number = 10;
|
|
|
|
three.min_write_buffer_number_to_merge = 4;
|
Support saving history in memtable_list
Summary:
For transactions, we are using the memtables to validate that there are no write conflicts. But after flushing, we don't have any memtables, and transactions could fail to commit. So we want to someone keep around some extra history to use for conflict checking. In addition, we want to provide a way to increase the size of this history if too many transactions fail to commit.
After chatting with people, it seems like everyone prefers just using Memtables to store this history (instead of a separate history structure). It seems like the best place for this is abstracted inside the memtable_list. I decide to create a separate list in MemtableListVersion as using the same list complicated the flush/installalflushresults logic too much.
This diff adds a new parameter to control how much memtable history to keep around after flushing. However, it sounds like people aren't too fond of adding new parameters. So I am making the default size of flushed+not-flushed memtables be set to max_write_buffers. This should not change the maximum amount of memory used, but make it more likely we're using closer the the limit. (We are now postponing deleting flushed memtables until the max_write_buffer limit is reached). So while we might use more memory on average, we are still obeying the limit set (and you could argue it's better to go ahead and use up memory now instead of waiting for a write stall to happen to test this limit).
However, if people are opposed to this default behavior, we can easily set it to 0 and require this parameter be set in order to use transactions.
Test Plan: Added a xfunc test to play around with setting different values of this parameter in all tests. Added testing in memtablelist_test and planning on adding more testing here.
Reviewers: sdong, rven, igor
Reviewed By: igor
Subscribers: dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D37443
2015-05-28 23:34:24 +00:00
|
|
|
three.max_write_buffer_number_to_maintain = -1;
|
2014-02-26 22:16:23 +00:00
|
|
|
|
|
|
|
Reopen({default_cf, one, two, three});
|
|
|
|
|
2014-02-27 18:29:13 +00:00
|
|
|
int micros_wait_for_flush = 10000;
|
2014-02-26 22:16:23 +00:00
|
|
|
PutRandomData(0, 100, 1000);
|
2014-02-27 18:29:13 +00:00
|
|
|
WaitForFlush(0);
|
2014-02-26 22:16:23 +00:00
|
|
|
AssertNumberOfImmutableMemtables({0, 0, 0, 0});
|
2015-06-29 21:39:01 +00:00
|
|
|
AssertCountLiveLogFiles(1);
|
2014-02-26 22:16:23 +00:00
|
|
|
PutRandomData(1, 200, 1000);
|
|
|
|
env_->SleepForMicroseconds(micros_wait_for_flush);
|
|
|
|
AssertNumberOfImmutableMemtables({0, 1, 0, 0});
|
2015-06-29 21:39:01 +00:00
|
|
|
AssertCountLiveLogFiles(2);
|
2014-02-26 22:16:23 +00:00
|
|
|
PutRandomData(2, 1000, 1000);
|
|
|
|
env_->SleepForMicroseconds(micros_wait_for_flush);
|
|
|
|
AssertNumberOfImmutableMemtables({0, 1, 1, 0});
|
2015-06-29 21:39:01 +00:00
|
|
|
AssertCountLiveLogFiles(3);
|
2014-02-26 22:16:23 +00:00
|
|
|
PutRandomData(2, 1000, 1000);
|
|
|
|
env_->SleepForMicroseconds(micros_wait_for_flush);
|
|
|
|
AssertNumberOfImmutableMemtables({0, 1, 2, 0});
|
2015-06-29 21:39:01 +00:00
|
|
|
AssertCountLiveLogFiles(4);
|
2015-12-07 18:13:58 +00:00
|
|
|
PutRandomData(3, 93, 990);
|
2014-02-26 22:16:23 +00:00
|
|
|
env_->SleepForMicroseconds(micros_wait_for_flush);
|
|
|
|
AssertNumberOfImmutableMemtables({0, 1, 2, 1});
|
2015-06-29 21:39:01 +00:00
|
|
|
AssertCountLiveLogFiles(5);
|
2015-12-07 18:13:58 +00:00
|
|
|
PutRandomData(3, 88, 990);
|
2014-02-26 22:16:23 +00:00
|
|
|
env_->SleepForMicroseconds(micros_wait_for_flush);
|
|
|
|
AssertNumberOfImmutableMemtables({0, 1, 2, 2});
|
2015-06-29 21:39:01 +00:00
|
|
|
AssertCountLiveLogFiles(6);
|
2015-12-07 18:13:58 +00:00
|
|
|
PutRandomData(3, 88, 990);
|
2014-02-26 22:16:23 +00:00
|
|
|
env_->SleepForMicroseconds(micros_wait_for_flush);
|
|
|
|
AssertNumberOfImmutableMemtables({0, 1, 2, 3});
|
2015-06-29 21:39:01 +00:00
|
|
|
AssertCountLiveLogFiles(7);
|
2014-02-26 22:16:23 +00:00
|
|
|
PutRandomData(0, 100, 1000);
|
2014-02-27 18:29:13 +00:00
|
|
|
WaitForFlush(0);
|
2014-02-26 22:16:23 +00:00
|
|
|
AssertNumberOfImmutableMemtables({0, 1, 2, 3});
|
2015-06-29 21:39:01 +00:00
|
|
|
AssertCountLiveLogFiles(8);
|
2014-02-26 22:16:23 +00:00
|
|
|
PutRandomData(2, 100, 10000);
|
2014-02-27 18:29:13 +00:00
|
|
|
WaitForFlush(2);
|
2014-02-26 22:16:23 +00:00
|
|
|
AssertNumberOfImmutableMemtables({0, 1, 0, 3});
|
2015-06-29 21:39:01 +00:00
|
|
|
AssertCountLiveLogFiles(9);
|
2015-12-07 18:13:58 +00:00
|
|
|
PutRandomData(3, 88, 990);
|
2014-02-27 18:29:13 +00:00
|
|
|
WaitForFlush(3);
|
2014-02-26 22:16:23 +00:00
|
|
|
AssertNumberOfImmutableMemtables({0, 1, 0, 0});
|
2015-06-29 21:39:01 +00:00
|
|
|
AssertCountLiveLogFiles(10);
|
2015-12-07 18:13:58 +00:00
|
|
|
PutRandomData(3, 88, 990);
|
2014-02-26 22:16:23 +00:00
|
|
|
env_->SleepForMicroseconds(micros_wait_for_flush);
|
|
|
|
AssertNumberOfImmutableMemtables({0, 1, 0, 1});
|
2015-06-29 21:39:01 +00:00
|
|
|
AssertCountLiveLogFiles(11);
|
2014-02-26 22:16:23 +00:00
|
|
|
PutRandomData(1, 200, 1000);
|
2014-02-27 18:29:13 +00:00
|
|
|
WaitForFlush(1);
|
2014-02-26 22:16:23 +00:00
|
|
|
AssertNumberOfImmutableMemtables({0, 0, 0, 1});
|
2015-06-29 21:39:01 +00:00
|
|
|
AssertCountLiveLogFiles(5);
|
2015-12-07 18:13:58 +00:00
|
|
|
PutRandomData(3, 88 * 3, 990);
|
2015-05-05 22:59:02 +00:00
|
|
|
WaitForFlush(3);
|
2015-12-07 18:13:58 +00:00
|
|
|
PutRandomData(3, 88 * 4, 990);
|
2014-02-27 18:29:13 +00:00
|
|
|
WaitForFlush(3);
|
2014-02-26 22:16:23 +00:00
|
|
|
AssertNumberOfImmutableMemtables({0, 0, 0, 0});
|
2015-06-29 21:39:01 +00:00
|
|
|
AssertCountLiveLogFiles(12);
|
2014-02-26 22:16:23 +00:00
|
|
|
PutRandomData(0, 100, 1000);
|
2014-02-27 18:29:13 +00:00
|
|
|
WaitForFlush(0);
|
2014-02-26 22:16:23 +00:00
|
|
|
AssertNumberOfImmutableMemtables({0, 0, 0, 0});
|
2015-06-29 21:39:01 +00:00
|
|
|
AssertCountLiveLogFiles(12);
|
2015-08-26 21:19:31 +00:00
|
|
|
PutRandomData(2, 3 * 1000, 1000);
|
2014-02-27 18:29:13 +00:00
|
|
|
WaitForFlush(2);
|
2014-02-26 22:16:23 +00:00
|
|
|
AssertNumberOfImmutableMemtables({0, 0, 0, 0});
|
2015-06-29 21:39:01 +00:00
|
|
|
AssertCountLiveLogFiles(12);
|
2014-02-26 22:16:23 +00:00
|
|
|
PutRandomData(1, 2*200, 1000);
|
2014-02-27 18:29:13 +00:00
|
|
|
WaitForFlush(1);
|
2014-02-26 22:16:23 +00:00
|
|
|
AssertNumberOfImmutableMemtables({0, 0, 0, 0});
|
2015-06-29 21:39:01 +00:00
|
|
|
AssertCountLiveLogFiles(7);
|
2014-02-26 22:16:23 +00:00
|
|
|
Close();
|
|
|
|
}
|
|
|
|
|
2015-06-29 21:39:01 +00:00
|
|
|
#ifndef ROCKSDB_LITE // Cuckoo is not supported in lite
|
2015-03-17 21:08:00 +00:00
|
|
|
TEST_F(ColumnFamilyTest, MemtableNotSupportSnapshot) {
|
2014-12-11 02:39:09 +00:00
|
|
|
Open();
|
|
|
|
auto* s1 = dbfull()->GetSnapshot();
|
|
|
|
ASSERT_TRUE(s1 != nullptr);
|
|
|
|
dbfull()->ReleaseSnapshot(s1);
|
|
|
|
|
|
|
|
// Add a column family that doesn't support snapshot
|
|
|
|
ColumnFamilyOptions first;
|
|
|
|
first.memtable_factory.reset(NewHashCuckooRepFactory(1024 * 1024));
|
|
|
|
CreateColumnFamilies({"first"}, {first});
|
|
|
|
auto* s2 = dbfull()->GetSnapshot();
|
|
|
|
ASSERT_TRUE(s2 == nullptr);
|
|
|
|
|
|
|
|
// Add a column family that supports snapshot. Snapshot stays not supported.
|
|
|
|
ColumnFamilyOptions second;
|
|
|
|
CreateColumnFamilies({"second"}, {second});
|
|
|
|
auto* s3 = dbfull()->GetSnapshot();
|
|
|
|
ASSERT_TRUE(s3 == nullptr);
|
|
|
|
Close();
|
|
|
|
}
|
2015-06-29 21:39:01 +00:00
|
|
|
#endif // !ROCKSDB_LITE
|
2014-12-11 02:39:09 +00:00
|
|
|
|
2016-07-25 22:06:11 +00:00
|
|
|
class TestComparator : public Comparator {
|
|
|
|
int Compare(const rocksdb::Slice& a, const rocksdb::Slice& b) const override {
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
const char* Name() const override { return "Test"; }
|
|
|
|
void FindShortestSeparator(std::string* start,
|
|
|
|
const rocksdb::Slice& limit) const override {}
|
|
|
|
void FindShortSuccessor(std::string* key) const override {}
|
|
|
|
};
|
|
|
|
|
|
|
|
static TestComparator third_comparator;
|
|
|
|
static TestComparator fourth_comparator;
|
|
|
|
|
|
|
|
// Test that we can retrieve the comparator from a created CF
|
|
|
|
TEST_F(ColumnFamilyTest, GetComparator) {
|
|
|
|
Open();
|
|
|
|
// Add a column family with no comparator specified
|
|
|
|
CreateColumnFamilies({"first"});
|
|
|
|
const Comparator* comp = handles_[0]->GetComparator();
|
|
|
|
ASSERT_EQ(comp, BytewiseComparator());
|
|
|
|
|
|
|
|
// Add three column families - one with no comparator and two
|
|
|
|
// with comparators specified
|
|
|
|
ColumnFamilyOptions second, third, fourth;
|
|
|
|
second.comparator = &third_comparator;
|
|
|
|
third.comparator = &fourth_comparator;
|
|
|
|
CreateColumnFamilies({"second", "third", "fourth"}, {second, third, fourth});
|
|
|
|
ASSERT_EQ(handles_[1]->GetComparator(), BytewiseComparator());
|
|
|
|
ASSERT_EQ(handles_[2]->GetComparator(), &third_comparator);
|
|
|
|
ASSERT_EQ(handles_[3]->GetComparator(), &fourth_comparator);
|
|
|
|
Close();
|
|
|
|
}
|
|
|
|
|
2015-03-17 21:08:00 +00:00
|
|
|
TEST_F(ColumnFamilyTest, DifferentMergeOperators) {
|
2014-02-27 00:05:24 +00:00
|
|
|
Open();
|
|
|
|
CreateColumnFamilies({"first", "second"});
|
|
|
|
ColumnFamilyOptions default_cf, first, second;
|
|
|
|
first.merge_operator = MergeOperators::CreateUInt64AddOperator();
|
|
|
|
second.merge_operator = MergeOperators::CreateStringAppendOperator();
|
|
|
|
Reopen({default_cf, first, second});
|
|
|
|
|
|
|
|
std::string one, two, three;
|
|
|
|
PutFixed64(&one, 1);
|
|
|
|
PutFixed64(&two, 2);
|
|
|
|
PutFixed64(&three, 3);
|
|
|
|
|
|
|
|
ASSERT_OK(Put(0, "foo", two));
|
|
|
|
ASSERT_OK(Put(0, "foo", one));
|
|
|
|
ASSERT_TRUE(Merge(0, "foo", two).IsNotSupported());
|
|
|
|
ASSERT_EQ(Get(0, "foo"), one);
|
|
|
|
|
|
|
|
ASSERT_OK(Put(1, "foo", two));
|
|
|
|
ASSERT_OK(Put(1, "foo", one));
|
|
|
|
ASSERT_OK(Merge(1, "foo", two));
|
|
|
|
ASSERT_EQ(Get(1, "foo"), three);
|
|
|
|
|
|
|
|
ASSERT_OK(Put(2, "foo", two));
|
|
|
|
ASSERT_OK(Put(2, "foo", one));
|
|
|
|
ASSERT_OK(Merge(2, "foo", two));
|
|
|
|
ASSERT_EQ(Get(2, "foo"), one + "," + two);
|
|
|
|
Close();
|
|
|
|
}
|
|
|
|
|
2015-03-17 21:08:00 +00:00
|
|
|
TEST_F(ColumnFamilyTest, DifferentCompactionStyles) {
|
2014-02-27 00:05:24 +00:00
|
|
|
Open();
|
|
|
|
CreateColumnFamilies({"one", "two"});
|
|
|
|
ColumnFamilyOptions default_cf, one, two;
|
|
|
|
db_options_.max_open_files = 20; // only 10 files in file cache
|
2014-02-27 00:24:56 +00:00
|
|
|
db_options_.disableDataSync = true;
|
2014-02-27 00:05:24 +00:00
|
|
|
|
|
|
|
default_cf.compaction_style = kCompactionStyleLevel;
|
|
|
|
default_cf.num_levels = 3;
|
|
|
|
default_cf.write_buffer_size = 64 << 10; // 64KB
|
2014-02-27 00:24:56 +00:00
|
|
|
default_cf.target_file_size_base = 30 << 10;
|
|
|
|
default_cf.source_compaction_factor = 100;
|
2014-08-25 21:22:05 +00:00
|
|
|
BlockBasedTableOptions table_options;
|
|
|
|
table_options.no_block_cache = true;
|
|
|
|
default_cf.table_factory.reset(NewBlockBasedTableFactory(table_options));
|
2014-02-27 00:05:24 +00:00
|
|
|
|
|
|
|
one.compaction_style = kCompactionStyleUniversal;
|
2015-06-29 21:39:01 +00:00
|
|
|
|
2015-03-30 21:04:21 +00:00
|
|
|
one.num_levels = 1;
|
2014-02-27 00:05:24 +00:00
|
|
|
// trigger compaction if there are >= 4 files
|
|
|
|
one.level0_file_num_compaction_trigger = 4;
|
2015-08-26 21:19:31 +00:00
|
|
|
one.write_buffer_size = 120000;
|
2014-02-27 00:05:24 +00:00
|
|
|
|
|
|
|
two.compaction_style = kCompactionStyleLevel;
|
|
|
|
two.num_levels = 4;
|
|
|
|
two.level0_file_num_compaction_trigger = 3;
|
|
|
|
two.write_buffer_size = 100000;
|
|
|
|
|
|
|
|
Reopen({default_cf, one, two});
|
|
|
|
|
|
|
|
// SETUP column family "one" -- universal style
|
|
|
|
for (int i = 0; i < one.level0_file_num_compaction_trigger - 1; ++i) {
|
2015-08-26 21:19:31 +00:00
|
|
|
PutRandomData(1, 10, 12000);
|
|
|
|
PutRandomData(1, 1, 10);
|
2014-02-27 00:05:24 +00:00
|
|
|
WaitForFlush(1);
|
2015-06-29 21:39:01 +00:00
|
|
|
AssertFilesPerLevel(ToString(i + 1), 1);
|
2014-02-27 00:05:24 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
// SETUP column family "two" -- level style with 4 levels
|
|
|
|
for (int i = 0; i < two.level0_file_num_compaction_trigger - 1; ++i) {
|
2015-08-26 21:19:31 +00:00
|
|
|
PutRandomData(2, 10, 12000);
|
|
|
|
PutRandomData(2, 1, 10);
|
2014-02-27 00:05:24 +00:00
|
|
|
WaitForFlush(2);
|
2015-06-29 21:39:01 +00:00
|
|
|
AssertFilesPerLevel(ToString(i + 1), 2);
|
2014-02-27 00:05:24 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
// TRIGGER compaction "one"
|
2015-08-26 21:19:31 +00:00
|
|
|
PutRandomData(1, 10, 12000);
|
|
|
|
PutRandomData(1, 1, 10);
|
2014-02-27 00:05:24 +00:00
|
|
|
|
|
|
|
// TRIGGER compaction "two"
|
2015-08-26 21:19:31 +00:00
|
|
|
PutRandomData(2, 10, 12000);
|
|
|
|
PutRandomData(2, 1, 10);
|
2014-02-27 00:05:24 +00:00
|
|
|
|
|
|
|
// WAIT for compactions
|
|
|
|
WaitForCompaction();
|
|
|
|
|
|
|
|
// VERIFY compaction "one"
|
2015-06-29 21:39:01 +00:00
|
|
|
AssertFilesPerLevel("1", 1);
|
2014-02-27 00:05:24 +00:00
|
|
|
|
|
|
|
// VERIFY compaction "two"
|
2015-06-29 21:39:01 +00:00
|
|
|
AssertFilesPerLevel("0,1", 2);
|
2014-02-27 00:05:24 +00:00
|
|
|
CompactAll(2);
|
2015-06-29 21:39:01 +00:00
|
|
|
AssertFilesPerLevel("0,1", 2);
|
2014-02-27 00:05:24 +00:00
|
|
|
|
|
|
|
Close();
|
|
|
|
}
|
|
|
|
|
Running manual compactions in parallel with other automatic or manual compactions in restricted cases
Summary:
This diff provides a framework for doing manual
compactions in parallel with other compactions. We now have a deque of manual compactions. We also pass manual compactions as an argument from RunManualCompactions down to
BackgroundCompactions, so that RunManualCompactions can be reentrant.
Parallelism is controlled by the two routines
ConflictingManualCompaction to allow/disallow new parallel/manual
compactions based on already existing ManualCompactions. In this diff, by default manual compactions still have to run exclusive of other compactions. However, by setting the compaction option, exclusive_manual_compaction to false, it is possible to run other compactions in parallel with a manual compaction. However, we are still restricted to one manual compaction per column family at a time. All of these restrictions will be relaxed in future diffs.
I will be adding more tests later.
Test Plan: Rocksdb regression + new tests + valgrind
Reviewers: igor, anthony, IslamAbdelRahman, kradhakrishnan, yhchiang, sdong
Reviewed By: sdong
Subscribers: yoshinorim, dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D47973
2015-12-14 19:20:34 +00:00
|
|
|
#ifndef ROCKSDB_LITE
|
|
|
|
// Sync points not supported in RocksDB Lite
|
|
|
|
|
|
|
|
TEST_F(ColumnFamilyTest, MultipleManualCompactions) {
|
|
|
|
Open();
|
|
|
|
CreateColumnFamilies({"one", "two"});
|
|
|
|
ColumnFamilyOptions default_cf, one, two;
|
|
|
|
db_options_.max_open_files = 20; // only 10 files in file cache
|
|
|
|
db_options_.disableDataSync = true;
|
|
|
|
db_options_.max_background_compactions = 3;
|
|
|
|
|
|
|
|
default_cf.compaction_style = kCompactionStyleLevel;
|
|
|
|
default_cf.num_levels = 3;
|
|
|
|
default_cf.write_buffer_size = 64 << 10; // 64KB
|
|
|
|
default_cf.target_file_size_base = 30 << 10;
|
|
|
|
default_cf.source_compaction_factor = 100;
|
|
|
|
BlockBasedTableOptions table_options;
|
|
|
|
table_options.no_block_cache = true;
|
|
|
|
default_cf.table_factory.reset(NewBlockBasedTableFactory(table_options));
|
|
|
|
|
|
|
|
one.compaction_style = kCompactionStyleUniversal;
|
|
|
|
|
|
|
|
one.num_levels = 1;
|
|
|
|
// trigger compaction if there are >= 4 files
|
|
|
|
one.level0_file_num_compaction_trigger = 4;
|
|
|
|
one.write_buffer_size = 120000;
|
|
|
|
|
|
|
|
two.compaction_style = kCompactionStyleLevel;
|
|
|
|
two.num_levels = 4;
|
|
|
|
two.level0_file_num_compaction_trigger = 3;
|
|
|
|
two.write_buffer_size = 100000;
|
|
|
|
|
|
|
|
Reopen({default_cf, one, two});
|
|
|
|
|
|
|
|
// SETUP column family "one" -- universal style
|
|
|
|
for (int i = 0; i < one.level0_file_num_compaction_trigger - 2; ++i) {
|
|
|
|
PutRandomData(1, 10, 12000, true);
|
|
|
|
PutRandomData(1, 1, 10, true);
|
|
|
|
WaitForFlush(1);
|
|
|
|
AssertFilesPerLevel(ToString(i + 1), 1);
|
|
|
|
}
|
|
|
|
bool cf_1_1 = true;
|
|
|
|
rocksdb::SyncPoint::GetInstance()->LoadDependency(
|
|
|
|
{{"ColumnFamilyTest::MultiManual:4", "ColumnFamilyTest::MultiManual:1"},
|
|
|
|
{"ColumnFamilyTest::MultiManual:2", "ColumnFamilyTest::MultiManual:5"},
|
|
|
|
{"ColumnFamilyTest::MultiManual:2", "ColumnFamilyTest::MultiManual:3"}});
|
|
|
|
rocksdb::SyncPoint::GetInstance()->SetCallBack(
|
|
|
|
"DBImpl::BackgroundCompaction:NonTrivial:AfterRun", [&](void* arg) {
|
|
|
|
if (cf_1_1) {
|
|
|
|
TEST_SYNC_POINT("ColumnFamilyTest::MultiManual:4");
|
|
|
|
cf_1_1 = false;
|
|
|
|
TEST_SYNC_POINT("ColumnFamilyTest::MultiManual:3");
|
|
|
|
}
|
|
|
|
});
|
|
|
|
|
|
|
|
rocksdb::SyncPoint::GetInstance()->EnableProcessing();
|
|
|
|
std::vector<std::thread> threads;
|
|
|
|
threads.emplace_back([&] {
|
|
|
|
CompactRangeOptions compact_options;
|
|
|
|
compact_options.exclusive_manual_compaction = false;
|
|
|
|
ASSERT_OK(
|
|
|
|
db_->CompactRange(compact_options, handles_[1], nullptr, nullptr));
|
|
|
|
});
|
|
|
|
|
|
|
|
// SETUP column family "two" -- level style with 4 levels
|
|
|
|
for (int i = 0; i < two.level0_file_num_compaction_trigger - 2; ++i) {
|
|
|
|
PutRandomData(2, 10, 12000);
|
|
|
|
PutRandomData(2, 1, 10);
|
|
|
|
WaitForFlush(2);
|
|
|
|
AssertFilesPerLevel(ToString(i + 1), 2);
|
|
|
|
}
|
|
|
|
threads.emplace_back([&] {
|
|
|
|
TEST_SYNC_POINT("ColumnFamilyTest::MultiManual:1");
|
|
|
|
CompactRangeOptions compact_options;
|
|
|
|
compact_options.exclusive_manual_compaction = false;
|
|
|
|
ASSERT_OK(
|
|
|
|
db_->CompactRange(compact_options, handles_[2], nullptr, nullptr));
|
|
|
|
TEST_SYNC_POINT("ColumnFamilyTest::MultiManual:2");
|
|
|
|
});
|
|
|
|
|
|
|
|
TEST_SYNC_POINT("ColumnFamilyTest::MultiManual:5");
|
|
|
|
for (auto& t : threads) {
|
|
|
|
t.join();
|
|
|
|
}
|
|
|
|
|
|
|
|
// VERIFY compaction "one"
|
|
|
|
AssertFilesPerLevel("1", 1);
|
|
|
|
|
|
|
|
// VERIFY compaction "two"
|
|
|
|
AssertFilesPerLevel("0,1", 2);
|
|
|
|
CompactAll(2);
|
|
|
|
AssertFilesPerLevel("0,1", 2);
|
|
|
|
// Compare against saved keys
|
|
|
|
std::set<std::string>::iterator key_iter = keys_.begin();
|
|
|
|
while (key_iter != keys_.end()) {
|
|
|
|
ASSERT_NE("NOT_FOUND", Get(1, *key_iter));
|
|
|
|
key_iter++;
|
|
|
|
}
|
|
|
|
Close();
|
|
|
|
}
|
|
|
|
|
|
|
|
TEST_F(ColumnFamilyTest, AutomaticAndManualCompactions) {
|
|
|
|
Open();
|
|
|
|
CreateColumnFamilies({"one", "two"});
|
|
|
|
ColumnFamilyOptions default_cf, one, two;
|
|
|
|
db_options_.max_open_files = 20; // only 10 files in file cache
|
|
|
|
db_options_.disableDataSync = true;
|
|
|
|
db_options_.max_background_compactions = 3;
|
2016-04-08 00:40:42 +00:00
|
|
|
db_options_.base_background_compactions = 3;
|
Running manual compactions in parallel with other automatic or manual compactions in restricted cases
Summary:
This diff provides a framework for doing manual
compactions in parallel with other compactions. We now have a deque of manual compactions. We also pass manual compactions as an argument from RunManualCompactions down to
BackgroundCompactions, so that RunManualCompactions can be reentrant.
Parallelism is controlled by the two routines
ConflictingManualCompaction to allow/disallow new parallel/manual
compactions based on already existing ManualCompactions. In this diff, by default manual compactions still have to run exclusive of other compactions. However, by setting the compaction option, exclusive_manual_compaction to false, it is possible to run other compactions in parallel with a manual compaction. However, we are still restricted to one manual compaction per column family at a time. All of these restrictions will be relaxed in future diffs.
I will be adding more tests later.
Test Plan: Rocksdb regression + new tests + valgrind
Reviewers: igor, anthony, IslamAbdelRahman, kradhakrishnan, yhchiang, sdong
Reviewed By: sdong
Subscribers: yoshinorim, dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D47973
2015-12-14 19:20:34 +00:00
|
|
|
|
|
|
|
default_cf.compaction_style = kCompactionStyleLevel;
|
|
|
|
default_cf.num_levels = 3;
|
|
|
|
default_cf.write_buffer_size = 64 << 10; // 64KB
|
|
|
|
default_cf.target_file_size_base = 30 << 10;
|
|
|
|
default_cf.source_compaction_factor = 100;
|
|
|
|
BlockBasedTableOptions table_options;
|
|
|
|
table_options.no_block_cache = true;
|
|
|
|
default_cf.table_factory.reset(NewBlockBasedTableFactory(table_options));
|
|
|
|
|
|
|
|
one.compaction_style = kCompactionStyleUniversal;
|
|
|
|
|
|
|
|
one.num_levels = 1;
|
|
|
|
// trigger compaction if there are >= 4 files
|
|
|
|
one.level0_file_num_compaction_trigger = 4;
|
|
|
|
one.write_buffer_size = 120000;
|
|
|
|
|
|
|
|
two.compaction_style = kCompactionStyleLevel;
|
|
|
|
two.num_levels = 4;
|
|
|
|
two.level0_file_num_compaction_trigger = 3;
|
|
|
|
two.write_buffer_size = 100000;
|
|
|
|
|
|
|
|
Reopen({default_cf, one, two});
|
|
|
|
|
|
|
|
bool cf_1_1 = true;
|
|
|
|
rocksdb::SyncPoint::GetInstance()->LoadDependency(
|
|
|
|
{{"ColumnFamilyTest::AutoManual:4", "ColumnFamilyTest::AutoManual:1"},
|
|
|
|
{"ColumnFamilyTest::AutoManual:2", "ColumnFamilyTest::AutoManual:5"},
|
|
|
|
{"ColumnFamilyTest::AutoManual:2", "ColumnFamilyTest::AutoManual:3"}});
|
|
|
|
rocksdb::SyncPoint::GetInstance()->SetCallBack(
|
|
|
|
"DBImpl::BackgroundCompaction:NonTrivial:AfterRun", [&](void* arg) {
|
|
|
|
if (cf_1_1) {
|
|
|
|
cf_1_1 = false;
|
|
|
|
TEST_SYNC_POINT("ColumnFamilyTest::AutoManual:4");
|
|
|
|
TEST_SYNC_POINT("ColumnFamilyTest::AutoManual:3");
|
|
|
|
}
|
|
|
|
});
|
|
|
|
rocksdb::SyncPoint::GetInstance()->EnableProcessing();
|
|
|
|
// SETUP column family "one" -- universal style
|
|
|
|
for (int i = 0; i < one.level0_file_num_compaction_trigger; ++i) {
|
|
|
|
PutRandomData(1, 10, 12000, true);
|
|
|
|
PutRandomData(1, 1, 10, true);
|
|
|
|
WaitForFlush(1);
|
|
|
|
AssertFilesPerLevel(ToString(i + 1), 1);
|
|
|
|
}
|
|
|
|
|
|
|
|
TEST_SYNC_POINT("ColumnFamilyTest::AutoManual:1");
|
|
|
|
|
|
|
|
// SETUP column family "two" -- level style with 4 levels
|
|
|
|
for (int i = 0; i < two.level0_file_num_compaction_trigger - 2; ++i) {
|
|
|
|
PutRandomData(2, 10, 12000);
|
|
|
|
PutRandomData(2, 1, 10);
|
|
|
|
WaitForFlush(2);
|
|
|
|
AssertFilesPerLevel(ToString(i + 1), 2);
|
|
|
|
}
|
|
|
|
std::thread threads([&] {
|
|
|
|
CompactRangeOptions compact_options;
|
|
|
|
compact_options.exclusive_manual_compaction = false;
|
|
|
|
ASSERT_OK(
|
|
|
|
db_->CompactRange(compact_options, handles_[2], nullptr, nullptr));
|
|
|
|
TEST_SYNC_POINT("ColumnFamilyTest::AutoManual:2");
|
|
|
|
});
|
|
|
|
|
|
|
|
TEST_SYNC_POINT("ColumnFamilyTest::AutoManual:5");
|
|
|
|
threads.join();
|
|
|
|
|
|
|
|
// WAIT for compactions
|
|
|
|
WaitForCompaction();
|
|
|
|
|
|
|
|
// VERIFY compaction "one"
|
|
|
|
AssertFilesPerLevel("1", 1);
|
|
|
|
|
|
|
|
// VERIFY compaction "two"
|
|
|
|
AssertFilesPerLevel("0,1", 2);
|
|
|
|
CompactAll(2);
|
|
|
|
AssertFilesPerLevel("0,1", 2);
|
|
|
|
// Compare against saved keys
|
|
|
|
std::set<std::string>::iterator key_iter = keys_.begin();
|
|
|
|
while (key_iter != keys_.end()) {
|
|
|
|
ASSERT_NE("NOT_FOUND", Get(1, *key_iter));
|
|
|
|
key_iter++;
|
|
|
|
}
|
|
|
|
Close();
|
|
|
|
}
|
|
|
|
|
|
|
|
TEST_F(ColumnFamilyTest, ManualAndAutomaticCompactions) {
|
|
|
|
Open();
|
|
|
|
CreateColumnFamilies({"one", "two"});
|
|
|
|
ColumnFamilyOptions default_cf, one, two;
|
|
|
|
db_options_.max_open_files = 20; // only 10 files in file cache
|
|
|
|
db_options_.disableDataSync = true;
|
|
|
|
db_options_.max_background_compactions = 3;
|
2016-04-08 00:40:42 +00:00
|
|
|
db_options_.base_background_compactions = 3;
|
Running manual compactions in parallel with other automatic or manual compactions in restricted cases
Summary:
This diff provides a framework for doing manual
compactions in parallel with other compactions. We now have a deque of manual compactions. We also pass manual compactions as an argument from RunManualCompactions down to
BackgroundCompactions, so that RunManualCompactions can be reentrant.
Parallelism is controlled by the two routines
ConflictingManualCompaction to allow/disallow new parallel/manual
compactions based on already existing ManualCompactions. In this diff, by default manual compactions still have to run exclusive of other compactions. However, by setting the compaction option, exclusive_manual_compaction to false, it is possible to run other compactions in parallel with a manual compaction. However, we are still restricted to one manual compaction per column family at a time. All of these restrictions will be relaxed in future diffs.
I will be adding more tests later.
Test Plan: Rocksdb regression + new tests + valgrind
Reviewers: igor, anthony, IslamAbdelRahman, kradhakrishnan, yhchiang, sdong
Reviewed By: sdong
Subscribers: yoshinorim, dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D47973
2015-12-14 19:20:34 +00:00
|
|
|
|
|
|
|
default_cf.compaction_style = kCompactionStyleLevel;
|
|
|
|
default_cf.num_levels = 3;
|
|
|
|
default_cf.write_buffer_size = 64 << 10; // 64KB
|
|
|
|
default_cf.target_file_size_base = 30 << 10;
|
|
|
|
default_cf.source_compaction_factor = 100;
|
|
|
|
BlockBasedTableOptions table_options;
|
|
|
|
table_options.no_block_cache = true;
|
|
|
|
default_cf.table_factory.reset(NewBlockBasedTableFactory(table_options));
|
|
|
|
|
|
|
|
one.compaction_style = kCompactionStyleUniversal;
|
|
|
|
|
|
|
|
one.num_levels = 1;
|
|
|
|
// trigger compaction if there are >= 4 files
|
|
|
|
one.level0_file_num_compaction_trigger = 4;
|
|
|
|
one.write_buffer_size = 120000;
|
|
|
|
|
|
|
|
two.compaction_style = kCompactionStyleLevel;
|
|
|
|
two.num_levels = 4;
|
|
|
|
two.level0_file_num_compaction_trigger = 3;
|
|
|
|
two.write_buffer_size = 100000;
|
|
|
|
|
|
|
|
Reopen({default_cf, one, two});
|
|
|
|
|
|
|
|
// SETUP column family "one" -- universal style
|
|
|
|
for (int i = 0; i < one.level0_file_num_compaction_trigger - 2; ++i) {
|
|
|
|
PutRandomData(1, 10, 12000, true);
|
|
|
|
PutRandomData(1, 1, 10, true);
|
|
|
|
WaitForFlush(1);
|
|
|
|
AssertFilesPerLevel(ToString(i + 1), 1);
|
|
|
|
}
|
|
|
|
bool cf_1_1 = true;
|
|
|
|
bool cf_1_2 = true;
|
|
|
|
rocksdb::SyncPoint::GetInstance()->LoadDependency(
|
|
|
|
{{"ColumnFamilyTest::ManualAuto:4", "ColumnFamilyTest::ManualAuto:1"},
|
|
|
|
{"ColumnFamilyTest::ManualAuto:5", "ColumnFamilyTest::ManualAuto:2"},
|
|
|
|
{"ColumnFamilyTest::ManualAuto:2", "ColumnFamilyTest::ManualAuto:3"}});
|
|
|
|
rocksdb::SyncPoint::GetInstance()->SetCallBack(
|
|
|
|
"DBImpl::BackgroundCompaction:NonTrivial:AfterRun", [&](void* arg) {
|
|
|
|
if (cf_1_1) {
|
|
|
|
TEST_SYNC_POINT("ColumnFamilyTest::ManualAuto:4");
|
|
|
|
cf_1_1 = false;
|
|
|
|
TEST_SYNC_POINT("ColumnFamilyTest::ManualAuto:3");
|
|
|
|
} else if (cf_1_2) {
|
|
|
|
TEST_SYNC_POINT("ColumnFamilyTest::ManualAuto:2");
|
|
|
|
cf_1_2 = false;
|
|
|
|
}
|
|
|
|
});
|
|
|
|
|
|
|
|
rocksdb::SyncPoint::GetInstance()->EnableProcessing();
|
|
|
|
std::thread threads([&] {
|
|
|
|
CompactRangeOptions compact_options;
|
|
|
|
compact_options.exclusive_manual_compaction = false;
|
|
|
|
ASSERT_OK(
|
|
|
|
db_->CompactRange(compact_options, handles_[1], nullptr, nullptr));
|
|
|
|
});
|
|
|
|
|
|
|
|
TEST_SYNC_POINT("ColumnFamilyTest::ManualAuto:1");
|
|
|
|
|
|
|
|
// SETUP column family "two" -- level style with 4 levels
|
|
|
|
for (int i = 0; i < two.level0_file_num_compaction_trigger; ++i) {
|
|
|
|
PutRandomData(2, 10, 12000);
|
|
|
|
PutRandomData(2, 1, 10);
|
|
|
|
WaitForFlush(2);
|
|
|
|
AssertFilesPerLevel(ToString(i + 1), 2);
|
|
|
|
}
|
|
|
|
TEST_SYNC_POINT("ColumnFamilyTest::ManualAuto:5");
|
|
|
|
threads.join();
|
|
|
|
|
|
|
|
// WAIT for compactions
|
|
|
|
WaitForCompaction();
|
|
|
|
|
|
|
|
// VERIFY compaction "one"
|
|
|
|
AssertFilesPerLevel("1", 1);
|
|
|
|
|
|
|
|
// VERIFY compaction "two"
|
|
|
|
AssertFilesPerLevel("0,1", 2);
|
|
|
|
CompactAll(2);
|
|
|
|
AssertFilesPerLevel("0,1", 2);
|
|
|
|
// Compare against saved keys
|
|
|
|
std::set<std::string>::iterator key_iter = keys_.begin();
|
|
|
|
while (key_iter != keys_.end()) {
|
|
|
|
ASSERT_NE("NOT_FOUND", Get(1, *key_iter));
|
|
|
|
key_iter++;
|
|
|
|
}
|
|
|
|
Close();
|
|
|
|
}
|
|
|
|
|
2015-12-18 00:59:00 +00:00
|
|
|
TEST_F(ColumnFamilyTest, SameCFManualManualCompactions) {
|
|
|
|
Open();
|
|
|
|
CreateColumnFamilies({"one"});
|
|
|
|
ColumnFamilyOptions default_cf, one;
|
|
|
|
db_options_.max_open_files = 20; // only 10 files in file cache
|
|
|
|
db_options_.disableDataSync = true;
|
|
|
|
db_options_.max_background_compactions = 3;
|
2016-04-08 00:40:42 +00:00
|
|
|
db_options_.base_background_compactions = 3;
|
2015-12-18 00:59:00 +00:00
|
|
|
|
|
|
|
default_cf.compaction_style = kCompactionStyleLevel;
|
|
|
|
default_cf.num_levels = 3;
|
|
|
|
default_cf.write_buffer_size = 64 << 10; // 64KB
|
|
|
|
default_cf.target_file_size_base = 30 << 10;
|
|
|
|
default_cf.source_compaction_factor = 100;
|
|
|
|
BlockBasedTableOptions table_options;
|
|
|
|
table_options.no_block_cache = true;
|
|
|
|
default_cf.table_factory.reset(NewBlockBasedTableFactory(table_options));
|
|
|
|
|
|
|
|
one.compaction_style = kCompactionStyleUniversal;
|
|
|
|
|
|
|
|
one.num_levels = 1;
|
|
|
|
// trigger compaction if there are >= 4 files
|
|
|
|
one.level0_file_num_compaction_trigger = 4;
|
|
|
|
one.write_buffer_size = 120000;
|
|
|
|
|
|
|
|
Reopen({default_cf, one});
|
|
|
|
|
|
|
|
// SETUP column family "one" -- universal style
|
|
|
|
for (int i = 0; i < one.level0_file_num_compaction_trigger - 2; ++i) {
|
|
|
|
PutRandomData(1, 10, 12000, true);
|
|
|
|
PutRandomData(1, 1, 10, true);
|
|
|
|
WaitForFlush(1);
|
|
|
|
AssertFilesPerLevel(ToString(i + 1), 1);
|
|
|
|
}
|
|
|
|
bool cf_1_1 = true;
|
|
|
|
bool cf_1_2 = true;
|
|
|
|
rocksdb::SyncPoint::GetInstance()->LoadDependency(
|
|
|
|
{{"ColumnFamilyTest::ManualManual:4", "ColumnFamilyTest::ManualManual:2"},
|
|
|
|
{"ColumnFamilyTest::ManualManual:4", "ColumnFamilyTest::ManualManual:5"},
|
|
|
|
{"ColumnFamilyTest::ManualManual:1", "ColumnFamilyTest::ManualManual:2"},
|
|
|
|
{"ColumnFamilyTest::ManualManual:1",
|
|
|
|
"ColumnFamilyTest::ManualManual:3"}});
|
|
|
|
rocksdb::SyncPoint::GetInstance()->SetCallBack(
|
|
|
|
"DBImpl::BackgroundCompaction:NonTrivial:AfterRun", [&](void* arg) {
|
|
|
|
if (cf_1_1) {
|
|
|
|
TEST_SYNC_POINT("ColumnFamilyTest::ManualManual:4");
|
|
|
|
cf_1_1 = false;
|
|
|
|
TEST_SYNC_POINT("ColumnFamilyTest::ManualManual:3");
|
|
|
|
} else if (cf_1_2) {
|
|
|
|
TEST_SYNC_POINT("ColumnFamilyTest::ManualManual:2");
|
|
|
|
cf_1_2 = false;
|
|
|
|
}
|
|
|
|
});
|
|
|
|
|
|
|
|
rocksdb::SyncPoint::GetInstance()->EnableProcessing();
|
|
|
|
std::thread threads([&] {
|
|
|
|
CompactRangeOptions compact_options;
|
|
|
|
compact_options.exclusive_manual_compaction = true;
|
|
|
|
ASSERT_OK(
|
|
|
|
db_->CompactRange(compact_options, handles_[1], nullptr, nullptr));
|
|
|
|
});
|
|
|
|
|
|
|
|
TEST_SYNC_POINT("ColumnFamilyTest::ManualManual:5");
|
|
|
|
|
|
|
|
WaitForFlush(1);
|
|
|
|
|
|
|
|
// Add more L0 files and force another manual compaction
|
|
|
|
for (int i = 0; i < one.level0_file_num_compaction_trigger - 2; ++i) {
|
|
|
|
PutRandomData(1, 10, 12000, true);
|
|
|
|
PutRandomData(1, 1, 10, true);
|
|
|
|
WaitForFlush(1);
|
|
|
|
AssertFilesPerLevel(ToString(one.level0_file_num_compaction_trigger + i),
|
|
|
|
1);
|
|
|
|
}
|
|
|
|
|
|
|
|
std::thread threads1([&] {
|
|
|
|
CompactRangeOptions compact_options;
|
|
|
|
compact_options.exclusive_manual_compaction = false;
|
|
|
|
ASSERT_OK(
|
|
|
|
db_->CompactRange(compact_options, handles_[1], nullptr, nullptr));
|
|
|
|
});
|
|
|
|
|
|
|
|
TEST_SYNC_POINT("ColumnFamilyTest::ManualManual:1");
|
|
|
|
|
|
|
|
threads.join();
|
|
|
|
threads1.join();
|
|
|
|
WaitForCompaction();
|
|
|
|
// VERIFY compaction "one"
|
|
|
|
ASSERT_LE(NumTableFilesAtLevel(0, 1), 2);
|
|
|
|
|
|
|
|
// Compare against saved keys
|
|
|
|
std::set<std::string>::iterator key_iter = keys_.begin();
|
|
|
|
while (key_iter != keys_.end()) {
|
|
|
|
ASSERT_NE("NOT_FOUND", Get(1, *key_iter));
|
|
|
|
key_iter++;
|
|
|
|
}
|
|
|
|
Close();
|
|
|
|
}
|
|
|
|
|
Running manual compactions in parallel with other automatic or manual compactions in restricted cases
Summary:
This diff provides a framework for doing manual
compactions in parallel with other compactions. We now have a deque of manual compactions. We also pass manual compactions as an argument from RunManualCompactions down to
BackgroundCompactions, so that RunManualCompactions can be reentrant.
Parallelism is controlled by the two routines
ConflictingManualCompaction to allow/disallow new parallel/manual
compactions based on already existing ManualCompactions. In this diff, by default manual compactions still have to run exclusive of other compactions. However, by setting the compaction option, exclusive_manual_compaction to false, it is possible to run other compactions in parallel with a manual compaction. However, we are still restricted to one manual compaction per column family at a time. All of these restrictions will be relaxed in future diffs.
I will be adding more tests later.
Test Plan: Rocksdb regression + new tests + valgrind
Reviewers: igor, anthony, IslamAbdelRahman, kradhakrishnan, yhchiang, sdong
Reviewed By: sdong
Subscribers: yoshinorim, dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D47973
2015-12-14 19:20:34 +00:00
|
|
|
TEST_F(ColumnFamilyTest, SameCFManualAutomaticCompactions) {
|
|
|
|
Open();
|
|
|
|
CreateColumnFamilies({"one"});
|
|
|
|
ColumnFamilyOptions default_cf, one;
|
|
|
|
db_options_.max_open_files = 20; // only 10 files in file cache
|
|
|
|
db_options_.disableDataSync = true;
|
|
|
|
db_options_.max_background_compactions = 3;
|
2016-04-08 00:40:42 +00:00
|
|
|
db_options_.base_background_compactions = 3;
|
Running manual compactions in parallel with other automatic or manual compactions in restricted cases
Summary:
This diff provides a framework for doing manual
compactions in parallel with other compactions. We now have a deque of manual compactions. We also pass manual compactions as an argument from RunManualCompactions down to
BackgroundCompactions, so that RunManualCompactions can be reentrant.
Parallelism is controlled by the two routines
ConflictingManualCompaction to allow/disallow new parallel/manual
compactions based on already existing ManualCompactions. In this diff, by default manual compactions still have to run exclusive of other compactions. However, by setting the compaction option, exclusive_manual_compaction to false, it is possible to run other compactions in parallel with a manual compaction. However, we are still restricted to one manual compaction per column family at a time. All of these restrictions will be relaxed in future diffs.
I will be adding more tests later.
Test Plan: Rocksdb regression + new tests + valgrind
Reviewers: igor, anthony, IslamAbdelRahman, kradhakrishnan, yhchiang, sdong
Reviewed By: sdong
Subscribers: yoshinorim, dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D47973
2015-12-14 19:20:34 +00:00
|
|
|
|
|
|
|
default_cf.compaction_style = kCompactionStyleLevel;
|
|
|
|
default_cf.num_levels = 3;
|
|
|
|
default_cf.write_buffer_size = 64 << 10; // 64KB
|
|
|
|
default_cf.target_file_size_base = 30 << 10;
|
|
|
|
default_cf.source_compaction_factor = 100;
|
|
|
|
BlockBasedTableOptions table_options;
|
|
|
|
table_options.no_block_cache = true;
|
|
|
|
default_cf.table_factory.reset(NewBlockBasedTableFactory(table_options));
|
|
|
|
|
|
|
|
one.compaction_style = kCompactionStyleUniversal;
|
|
|
|
|
|
|
|
one.num_levels = 1;
|
|
|
|
// trigger compaction if there are >= 4 files
|
|
|
|
one.level0_file_num_compaction_trigger = 4;
|
|
|
|
one.write_buffer_size = 120000;
|
|
|
|
|
|
|
|
Reopen({default_cf, one});
|
|
|
|
|
|
|
|
// SETUP column family "one" -- universal style
|
|
|
|
for (int i = 0; i < one.level0_file_num_compaction_trigger - 2; ++i) {
|
|
|
|
PutRandomData(1, 10, 12000, true);
|
|
|
|
PutRandomData(1, 1, 10, true);
|
|
|
|
WaitForFlush(1);
|
|
|
|
AssertFilesPerLevel(ToString(i + 1), 1);
|
|
|
|
}
|
|
|
|
bool cf_1_1 = true;
|
|
|
|
bool cf_1_2 = true;
|
|
|
|
rocksdb::SyncPoint::GetInstance()->LoadDependency(
|
|
|
|
{{"ColumnFamilyTest::ManualAuto:4", "ColumnFamilyTest::ManualAuto:2"},
|
|
|
|
{"ColumnFamilyTest::ManualAuto:4", "ColumnFamilyTest::ManualAuto:5"},
|
|
|
|
{"ColumnFamilyTest::ManualAuto:1", "ColumnFamilyTest::ManualAuto:2"},
|
|
|
|
{"ColumnFamilyTest::ManualAuto:1", "ColumnFamilyTest::ManualAuto:3"}});
|
|
|
|
rocksdb::SyncPoint::GetInstance()->SetCallBack(
|
|
|
|
"DBImpl::BackgroundCompaction:NonTrivial:AfterRun", [&](void* arg) {
|
|
|
|
if (cf_1_1) {
|
|
|
|
TEST_SYNC_POINT("ColumnFamilyTest::ManualAuto:4");
|
|
|
|
cf_1_1 = false;
|
|
|
|
TEST_SYNC_POINT("ColumnFamilyTest::ManualAuto:3");
|
|
|
|
} else if (cf_1_2) {
|
|
|
|
TEST_SYNC_POINT("ColumnFamilyTest::ManualAuto:2");
|
|
|
|
cf_1_2 = false;
|
|
|
|
}
|
|
|
|
});
|
|
|
|
|
|
|
|
rocksdb::SyncPoint::GetInstance()->EnableProcessing();
|
|
|
|
std::thread threads([&] {
|
|
|
|
CompactRangeOptions compact_options;
|
|
|
|
compact_options.exclusive_manual_compaction = false;
|
|
|
|
ASSERT_OK(
|
|
|
|
db_->CompactRange(compact_options, handles_[1], nullptr, nullptr));
|
|
|
|
});
|
|
|
|
|
|
|
|
TEST_SYNC_POINT("ColumnFamilyTest::ManualAuto:5");
|
|
|
|
|
|
|
|
WaitForFlush(1);
|
|
|
|
|
|
|
|
// Add more L0 files and force automatic compaction
|
|
|
|
for (int i = 0; i < one.level0_file_num_compaction_trigger; ++i) {
|
|
|
|
PutRandomData(1, 10, 12000, true);
|
|
|
|
PutRandomData(1, 1, 10, true);
|
|
|
|
WaitForFlush(1);
|
|
|
|
AssertFilesPerLevel(ToString(one.level0_file_num_compaction_trigger + i),
|
|
|
|
1);
|
|
|
|
}
|
|
|
|
|
|
|
|
TEST_SYNC_POINT("ColumnFamilyTest::ManualAuto:1");
|
|
|
|
|
|
|
|
threads.join();
|
|
|
|
WaitForCompaction();
|
|
|
|
// VERIFY compaction "one"
|
|
|
|
ASSERT_LE(NumTableFilesAtLevel(0, 1), 2);
|
|
|
|
|
|
|
|
// Compare against saved keys
|
|
|
|
std::set<std::string>::iterator key_iter = keys_.begin();
|
|
|
|
while (key_iter != keys_.end()) {
|
|
|
|
ASSERT_NE("NOT_FOUND", Get(1, *key_iter));
|
|
|
|
key_iter++;
|
|
|
|
}
|
|
|
|
Close();
|
|
|
|
}
|
|
|
|
|
|
|
|
TEST_F(ColumnFamilyTest, SameCFManualAutomaticCompactionsLevel) {
|
|
|
|
Open();
|
|
|
|
CreateColumnFamilies({"one"});
|
|
|
|
ColumnFamilyOptions default_cf, one;
|
|
|
|
db_options_.max_open_files = 20; // only 10 files in file cache
|
|
|
|
db_options_.disableDataSync = true;
|
|
|
|
db_options_.max_background_compactions = 3;
|
2016-04-08 00:40:42 +00:00
|
|
|
db_options_.base_background_compactions = 3;
|
Running manual compactions in parallel with other automatic or manual compactions in restricted cases
Summary:
This diff provides a framework for doing manual
compactions in parallel with other compactions. We now have a deque of manual compactions. We also pass manual compactions as an argument from RunManualCompactions down to
BackgroundCompactions, so that RunManualCompactions can be reentrant.
Parallelism is controlled by the two routines
ConflictingManualCompaction to allow/disallow new parallel/manual
compactions based on already existing ManualCompactions. In this diff, by default manual compactions still have to run exclusive of other compactions. However, by setting the compaction option, exclusive_manual_compaction to false, it is possible to run other compactions in parallel with a manual compaction. However, we are still restricted to one manual compaction per column family at a time. All of these restrictions will be relaxed in future diffs.
I will be adding more tests later.
Test Plan: Rocksdb regression + new tests + valgrind
Reviewers: igor, anthony, IslamAbdelRahman, kradhakrishnan, yhchiang, sdong
Reviewed By: sdong
Subscribers: yoshinorim, dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D47973
2015-12-14 19:20:34 +00:00
|
|
|
|
|
|
|
default_cf.compaction_style = kCompactionStyleLevel;
|
|
|
|
default_cf.num_levels = 3;
|
|
|
|
default_cf.write_buffer_size = 64 << 10; // 64KB
|
|
|
|
default_cf.target_file_size_base = 30 << 10;
|
|
|
|
default_cf.source_compaction_factor = 100;
|
|
|
|
BlockBasedTableOptions table_options;
|
|
|
|
table_options.no_block_cache = true;
|
|
|
|
default_cf.table_factory.reset(NewBlockBasedTableFactory(table_options));
|
|
|
|
|
|
|
|
one.compaction_style = kCompactionStyleLevel;
|
|
|
|
|
|
|
|
one.num_levels = 1;
|
|
|
|
// trigger compaction if there are >= 4 files
|
|
|
|
one.level0_file_num_compaction_trigger = 4;
|
|
|
|
one.write_buffer_size = 120000;
|
|
|
|
|
|
|
|
Reopen({default_cf, one});
|
|
|
|
|
|
|
|
// SETUP column family "one" -- level style
|
|
|
|
for (int i = 0; i < one.level0_file_num_compaction_trigger - 2; ++i) {
|
|
|
|
PutRandomData(1, 10, 12000, true);
|
|
|
|
PutRandomData(1, 1, 10, true);
|
|
|
|
WaitForFlush(1);
|
|
|
|
AssertFilesPerLevel(ToString(i + 1), 1);
|
|
|
|
}
|
|
|
|
bool cf_1_1 = true;
|
|
|
|
bool cf_1_2 = true;
|
|
|
|
rocksdb::SyncPoint::GetInstance()->LoadDependency(
|
|
|
|
{{"ColumnFamilyTest::ManualAuto:4", "ColumnFamilyTest::ManualAuto:2"},
|
|
|
|
{"ColumnFamilyTest::ManualAuto:4", "ColumnFamilyTest::ManualAuto:5"},
|
|
|
|
{"ColumnFamilyTest::ManualAuto:3", "ColumnFamilyTest::ManualAuto:2"},
|
|
|
|
{"LevelCompactionPicker::PickCompactionBySize:0",
|
|
|
|
"ColumnFamilyTest::ManualAuto:3"},
|
|
|
|
{"ColumnFamilyTest::ManualAuto:1", "ColumnFamilyTest::ManualAuto:3"}});
|
|
|
|
rocksdb::SyncPoint::GetInstance()->SetCallBack(
|
|
|
|
"DBImpl::BackgroundCompaction:NonTrivial:AfterRun", [&](void* arg) {
|
|
|
|
if (cf_1_1) {
|
|
|
|
TEST_SYNC_POINT("ColumnFamilyTest::ManualAuto:4");
|
|
|
|
cf_1_1 = false;
|
|
|
|
TEST_SYNC_POINT("ColumnFamilyTest::ManualAuto:3");
|
|
|
|
} else if (cf_1_2) {
|
|
|
|
TEST_SYNC_POINT("ColumnFamilyTest::ManualAuto:2");
|
|
|
|
cf_1_2 = false;
|
|
|
|
}
|
|
|
|
});
|
|
|
|
|
|
|
|
rocksdb::SyncPoint::GetInstance()->EnableProcessing();
|
|
|
|
std::thread threads([&] {
|
|
|
|
CompactRangeOptions compact_options;
|
|
|
|
compact_options.exclusive_manual_compaction = false;
|
|
|
|
ASSERT_OK(
|
|
|
|
db_->CompactRange(compact_options, handles_[1], nullptr, nullptr));
|
|
|
|
});
|
|
|
|
|
|
|
|
TEST_SYNC_POINT("ColumnFamilyTest::ManualAuto:5");
|
|
|
|
|
|
|
|
// Add more L0 files and force automatic compaction
|
|
|
|
for (int i = 0; i < one.level0_file_num_compaction_trigger; ++i) {
|
|
|
|
PutRandomData(1, 10, 12000, true);
|
|
|
|
PutRandomData(1, 1, 10, true);
|
|
|
|
WaitForFlush(1);
|
|
|
|
AssertFilesPerLevel(ToString(one.level0_file_num_compaction_trigger + i),
|
|
|
|
1);
|
|
|
|
}
|
|
|
|
|
|
|
|
TEST_SYNC_POINT("ColumnFamilyTest::ManualAuto:1");
|
|
|
|
|
|
|
|
threads.join();
|
|
|
|
WaitForCompaction();
|
|
|
|
// VERIFY compaction "one"
|
|
|
|
AssertFilesPerLevel("0,1", 1);
|
|
|
|
|
|
|
|
// Compare against saved keys
|
|
|
|
std::set<std::string>::iterator key_iter = keys_.begin();
|
|
|
|
while (key_iter != keys_.end()) {
|
|
|
|
ASSERT_NE("NOT_FOUND", Get(1, *key_iter));
|
|
|
|
key_iter++;
|
|
|
|
}
|
|
|
|
Close();
|
|
|
|
}
|
|
|
|
|
|
|
|
// This test checks for automatic getting a conflict if there is a
|
|
|
|
// manual which has not yet been scheduled.
|
|
|
|
// The manual compaction waits in NotScheduled
|
|
|
|
// We generate more files and then trigger an automatic compaction
|
|
|
|
// This will wait because there is an unscheduled manual compaction.
|
|
|
|
// Once the conflict is hit, the manual compaction starts and ends
|
|
|
|
// Then another automatic will start and end.
|
|
|
|
TEST_F(ColumnFamilyTest, SameCFManualAutomaticConflict) {
|
|
|
|
Open();
|
|
|
|
CreateColumnFamilies({"one"});
|
|
|
|
ColumnFamilyOptions default_cf, one;
|
|
|
|
db_options_.max_open_files = 20; // only 10 files in file cache
|
|
|
|
db_options_.disableDataSync = true;
|
|
|
|
db_options_.max_background_compactions = 3;
|
2016-04-08 00:40:42 +00:00
|
|
|
db_options_.base_background_compactions = 3;
|
Running manual compactions in parallel with other automatic or manual compactions in restricted cases
Summary:
This diff provides a framework for doing manual
compactions in parallel with other compactions. We now have a deque of manual compactions. We also pass manual compactions as an argument from RunManualCompactions down to
BackgroundCompactions, so that RunManualCompactions can be reentrant.
Parallelism is controlled by the two routines
ConflictingManualCompaction to allow/disallow new parallel/manual
compactions based on already existing ManualCompactions. In this diff, by default manual compactions still have to run exclusive of other compactions. However, by setting the compaction option, exclusive_manual_compaction to false, it is possible to run other compactions in parallel with a manual compaction. However, we are still restricted to one manual compaction per column family at a time. All of these restrictions will be relaxed in future diffs.
I will be adding more tests later.
Test Plan: Rocksdb regression + new tests + valgrind
Reviewers: igor, anthony, IslamAbdelRahman, kradhakrishnan, yhchiang, sdong
Reviewed By: sdong
Subscribers: yoshinorim, dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D47973
2015-12-14 19:20:34 +00:00
|
|
|
|
|
|
|
default_cf.compaction_style = kCompactionStyleLevel;
|
|
|
|
default_cf.num_levels = 3;
|
|
|
|
default_cf.write_buffer_size = 64 << 10; // 64KB
|
|
|
|
default_cf.target_file_size_base = 30 << 10;
|
|
|
|
default_cf.source_compaction_factor = 100;
|
|
|
|
BlockBasedTableOptions table_options;
|
|
|
|
table_options.no_block_cache = true;
|
|
|
|
default_cf.table_factory.reset(NewBlockBasedTableFactory(table_options));
|
|
|
|
|
|
|
|
one.compaction_style = kCompactionStyleUniversal;
|
|
|
|
|
|
|
|
one.num_levels = 1;
|
|
|
|
// trigger compaction if there are >= 4 files
|
|
|
|
one.level0_file_num_compaction_trigger = 4;
|
|
|
|
one.write_buffer_size = 120000;
|
|
|
|
|
|
|
|
Reopen({default_cf, one});
|
|
|
|
|
|
|
|
// SETUP column family "one" -- universal style
|
|
|
|
for (int i = 0; i < one.level0_file_num_compaction_trigger - 2; ++i) {
|
|
|
|
PutRandomData(1, 10, 12000, true);
|
|
|
|
PutRandomData(1, 1, 10, true);
|
|
|
|
WaitForFlush(1);
|
|
|
|
AssertFilesPerLevel(ToString(i + 1), 1);
|
|
|
|
}
|
|
|
|
bool cf_1_1 = true;
|
|
|
|
bool cf_1_2 = true;
|
|
|
|
rocksdb::SyncPoint::GetInstance()->LoadDependency(
|
|
|
|
{{"DBImpl::BackgroundCompaction()::Conflict",
|
|
|
|
"ColumnFamilyTest::ManualAutoCon:7"},
|
|
|
|
{"ColumnFamilyTest::ManualAutoCon:9",
|
|
|
|
"ColumnFamilyTest::ManualAutoCon:8"},
|
|
|
|
{"ColumnFamilyTest::ManualAutoCon:2",
|
|
|
|
"ColumnFamilyTest::ManualAutoCon:6"},
|
|
|
|
{"ColumnFamilyTest::ManualAutoCon:4",
|
|
|
|
"ColumnFamilyTest::ManualAutoCon:5"},
|
|
|
|
{"ColumnFamilyTest::ManualAutoCon:1",
|
|
|
|
"ColumnFamilyTest::ManualAutoCon:2"},
|
|
|
|
{"ColumnFamilyTest::ManualAutoCon:1",
|
|
|
|
"ColumnFamilyTest::ManualAutoCon:3"}});
|
|
|
|
rocksdb::SyncPoint::GetInstance()->SetCallBack(
|
|
|
|
"DBImpl::BackgroundCompaction:NonTrivial:AfterRun", [&](void* arg) {
|
|
|
|
if (cf_1_1) {
|
|
|
|
TEST_SYNC_POINT("ColumnFamilyTest::ManualAutoCon:4");
|
|
|
|
cf_1_1 = false;
|
|
|
|
TEST_SYNC_POINT("ColumnFamilyTest::ManualAutoCon:3");
|
|
|
|
} else if (cf_1_2) {
|
|
|
|
cf_1_2 = false;
|
|
|
|
TEST_SYNC_POINT("ColumnFamilyTest::ManualAutoCon:2");
|
|
|
|
}
|
|
|
|
});
|
|
|
|
rocksdb::SyncPoint::GetInstance()->SetCallBack(
|
|
|
|
"DBImpl::RunManualCompaction:NotScheduled", [&](void* arg) {
|
|
|
|
InstrumentedMutex* mutex = static_cast<InstrumentedMutex*>(arg);
|
|
|
|
mutex->Unlock();
|
|
|
|
TEST_SYNC_POINT("ColumnFamilyTest::ManualAutoCon:9");
|
|
|
|
TEST_SYNC_POINT("ColumnFamilyTest::ManualAutoCon:7");
|
|
|
|
mutex->Lock();
|
|
|
|
});
|
|
|
|
|
|
|
|
rocksdb::SyncPoint::GetInstance()->EnableProcessing();
|
|
|
|
std::thread threads([&] {
|
|
|
|
CompactRangeOptions compact_options;
|
|
|
|
compact_options.exclusive_manual_compaction = false;
|
|
|
|
ASSERT_OK(
|
|
|
|
db_->CompactRange(compact_options, handles_[1], nullptr, nullptr));
|
|
|
|
TEST_SYNC_POINT("ColumnFamilyTest::ManualAutoCon:6");
|
|
|
|
});
|
|
|
|
|
|
|
|
TEST_SYNC_POINT("ColumnFamilyTest::ManualAutoCon:8");
|
|
|
|
WaitForFlush(1);
|
|
|
|
|
|
|
|
// Add more L0 files and force automatic compaction
|
|
|
|
for (int i = 0; i < one.level0_file_num_compaction_trigger; ++i) {
|
|
|
|
PutRandomData(1, 10, 12000, true);
|
|
|
|
PutRandomData(1, 1, 10, true);
|
|
|
|
WaitForFlush(1);
|
|
|
|
AssertFilesPerLevel(ToString(one.level0_file_num_compaction_trigger + i),
|
|
|
|
1);
|
|
|
|
}
|
|
|
|
|
|
|
|
TEST_SYNC_POINT("ColumnFamilyTest::ManualAutoCon:5");
|
|
|
|
// Add more L0 files and force automatic compaction
|
|
|
|
for (int i = 0; i < one.level0_file_num_compaction_trigger; ++i) {
|
|
|
|
PutRandomData(1, 10, 12000, true);
|
|
|
|
PutRandomData(1, 1, 10, true);
|
|
|
|
WaitForFlush(1);
|
|
|
|
}
|
|
|
|
TEST_SYNC_POINT("ColumnFamilyTest::ManualAutoCon:1");
|
|
|
|
|
|
|
|
threads.join();
|
|
|
|
WaitForCompaction();
|
|
|
|
// VERIFY compaction "one"
|
|
|
|
ASSERT_LE(NumTableFilesAtLevel(0, 1), 3);
|
|
|
|
|
|
|
|
// Compare against saved keys
|
|
|
|
std::set<std::string>::iterator key_iter = keys_.begin();
|
|
|
|
while (key_iter != keys_.end()) {
|
|
|
|
ASSERT_NE("NOT_FOUND", Get(1, *key_iter));
|
|
|
|
key_iter++;
|
|
|
|
}
|
|
|
|
|
|
|
|
Close();
|
|
|
|
}
|
|
|
|
|
|
|
|
// In this test, we generate enough files to trigger automatic compactions.
|
|
|
|
// The automatic compaction waits in NonTrivial:AfterRun
|
|
|
|
// We generate more files and then trigger an automatic compaction
|
|
|
|
// This will wait because the automatic compaction has files it needs.
|
|
|
|
// Once the conflict is hit, the automatic compaction starts and ends
|
|
|
|
// Then the manual will run and end.
|
|
|
|
TEST_F(ColumnFamilyTest, SameCFAutomaticManualCompactions) {
|
|
|
|
Open();
|
|
|
|
CreateColumnFamilies({"one"});
|
|
|
|
ColumnFamilyOptions default_cf, one;
|
|
|
|
db_options_.max_open_files = 20; // only 10 files in file cache
|
|
|
|
db_options_.disableDataSync = true;
|
|
|
|
db_options_.max_background_compactions = 3;
|
2016-04-08 00:40:42 +00:00
|
|
|
db_options_.base_background_compactions = 3;
|
Running manual compactions in parallel with other automatic or manual compactions in restricted cases
Summary:
This diff provides a framework for doing manual
compactions in parallel with other compactions. We now have a deque of manual compactions. We also pass manual compactions as an argument from RunManualCompactions down to
BackgroundCompactions, so that RunManualCompactions can be reentrant.
Parallelism is controlled by the two routines
ConflictingManualCompaction to allow/disallow new parallel/manual
compactions based on already existing ManualCompactions. In this diff, by default manual compactions still have to run exclusive of other compactions. However, by setting the compaction option, exclusive_manual_compaction to false, it is possible to run other compactions in parallel with a manual compaction. However, we are still restricted to one manual compaction per column family at a time. All of these restrictions will be relaxed in future diffs.
I will be adding more tests later.
Test Plan: Rocksdb regression + new tests + valgrind
Reviewers: igor, anthony, IslamAbdelRahman, kradhakrishnan, yhchiang, sdong
Reviewed By: sdong
Subscribers: yoshinorim, dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D47973
2015-12-14 19:20:34 +00:00
|
|
|
|
|
|
|
default_cf.compaction_style = kCompactionStyleLevel;
|
|
|
|
default_cf.num_levels = 3;
|
|
|
|
default_cf.write_buffer_size = 64 << 10; // 64KB
|
|
|
|
default_cf.target_file_size_base = 30 << 10;
|
|
|
|
default_cf.source_compaction_factor = 100;
|
|
|
|
BlockBasedTableOptions table_options;
|
|
|
|
table_options.no_block_cache = true;
|
|
|
|
default_cf.table_factory.reset(NewBlockBasedTableFactory(table_options));
|
|
|
|
|
|
|
|
one.compaction_style = kCompactionStyleUniversal;
|
|
|
|
|
|
|
|
one.num_levels = 1;
|
|
|
|
// trigger compaction if there are >= 4 files
|
|
|
|
one.level0_file_num_compaction_trigger = 4;
|
|
|
|
one.write_buffer_size = 120000;
|
|
|
|
|
|
|
|
Reopen({default_cf, one});
|
|
|
|
|
|
|
|
bool cf_1_1 = true;
|
|
|
|
bool cf_1_2 = true;
|
|
|
|
rocksdb::SyncPoint::GetInstance()->LoadDependency(
|
|
|
|
{{"ColumnFamilyTest::AutoManual:4", "ColumnFamilyTest::AutoManual:2"},
|
|
|
|
{"ColumnFamilyTest::AutoManual:4", "ColumnFamilyTest::AutoManual:5"},
|
|
|
|
{"CompactionPicker::CompactRange:Conflict",
|
|
|
|
"ColumnFamilyTest::AutoManual:3"}});
|
|
|
|
rocksdb::SyncPoint::GetInstance()->SetCallBack(
|
|
|
|
"DBImpl::BackgroundCompaction:NonTrivial:AfterRun", [&](void* arg) {
|
|
|
|
if (cf_1_1) {
|
|
|
|
TEST_SYNC_POINT("ColumnFamilyTest::AutoManual:4");
|
|
|
|
cf_1_1 = false;
|
|
|
|
TEST_SYNC_POINT("ColumnFamilyTest::AutoManual:3");
|
|
|
|
} else if (cf_1_2) {
|
|
|
|
TEST_SYNC_POINT("ColumnFamilyTest::AutoManual:2");
|
|
|
|
cf_1_2 = false;
|
|
|
|
}
|
|
|
|
});
|
|
|
|
|
|
|
|
rocksdb::SyncPoint::GetInstance()->EnableProcessing();
|
|
|
|
|
|
|
|
// SETUP column family "one" -- universal style
|
|
|
|
for (int i = 0; i < one.level0_file_num_compaction_trigger; ++i) {
|
|
|
|
PutRandomData(1, 10, 12000, true);
|
|
|
|
PutRandomData(1, 1, 10, true);
|
|
|
|
WaitForFlush(1);
|
|
|
|
AssertFilesPerLevel(ToString(i + 1), 1);
|
|
|
|
}
|
|
|
|
|
|
|
|
TEST_SYNC_POINT("ColumnFamilyTest::AutoManual:5");
|
|
|
|
|
|
|
|
// Add another L0 file and force automatic compaction
|
|
|
|
for (int i = 0; i < one.level0_file_num_compaction_trigger - 2; ++i) {
|
|
|
|
PutRandomData(1, 10, 12000, true);
|
|
|
|
PutRandomData(1, 1, 10, true);
|
|
|
|
WaitForFlush(1);
|
|
|
|
}
|
|
|
|
|
|
|
|
CompactRangeOptions compact_options;
|
|
|
|
compact_options.exclusive_manual_compaction = false;
|
|
|
|
ASSERT_OK(db_->CompactRange(compact_options, handles_[1], nullptr, nullptr));
|
|
|
|
|
|
|
|
TEST_SYNC_POINT("ColumnFamilyTest::AutoManual:1");
|
|
|
|
|
|
|
|
WaitForCompaction();
|
|
|
|
// VERIFY compaction "one"
|
|
|
|
AssertFilesPerLevel("1", 1);
|
|
|
|
// Compare against saved keys
|
|
|
|
std::set<std::string>::iterator key_iter = keys_.begin();
|
|
|
|
while (key_iter != keys_.end()) {
|
|
|
|
ASSERT_NE("NOT_FOUND", Get(1, *key_iter));
|
|
|
|
key_iter++;
|
|
|
|
}
|
|
|
|
|
|
|
|
Close();
|
|
|
|
}
|
|
|
|
#endif // !ROCKSDB_LITE
|
|
|
|
|
2015-06-29 21:39:01 +00:00
|
|
|
#ifndef ROCKSDB_LITE // Tailing interator not supported
|
2014-03-08 00:12:34 +00:00
|
|
|
namespace {
|
|
|
|
std::string IterStatus(Iterator* iter) {
|
|
|
|
std::string result;
|
|
|
|
if (iter->Valid()) {
|
|
|
|
result = iter->key().ToString() + "->" + iter->value().ToString();
|
|
|
|
} else {
|
|
|
|
result = "(invalid)";
|
|
|
|
}
|
|
|
|
return result;
|
|
|
|
}
|
2014-03-11 21:52:17 +00:00
|
|
|
} // anonymous namespace
|
2014-03-08 00:12:34 +00:00
|
|
|
|
2015-03-17 21:08:00 +00:00
|
|
|
TEST_F(ColumnFamilyTest, NewIteratorsTest) {
|
2014-03-08 00:12:34 +00:00
|
|
|
// iter == 0 -- no tailing
|
|
|
|
// iter == 2 -- tailing
|
|
|
|
for (int iter = 0; iter < 2; ++iter) {
|
|
|
|
Open();
|
|
|
|
CreateColumnFamiliesAndReopen({"one", "two"});
|
|
|
|
ASSERT_OK(Put(0, "a", "b"));
|
|
|
|
ASSERT_OK(Put(1, "b", "a"));
|
|
|
|
ASSERT_OK(Put(2, "c", "m"));
|
|
|
|
ASSERT_OK(Put(2, "v", "t"));
|
|
|
|
std::vector<Iterator*> iterators;
|
|
|
|
ReadOptions options;
|
|
|
|
options.tailing = (iter == 1);
|
|
|
|
ASSERT_OK(db_->NewIterators(options, handles_, &iterators));
|
|
|
|
|
|
|
|
for (auto it : iterators) {
|
|
|
|
it->SeekToFirst();
|
|
|
|
}
|
|
|
|
ASSERT_EQ(IterStatus(iterators[0]), "a->b");
|
|
|
|
ASSERT_EQ(IterStatus(iterators[1]), "b->a");
|
|
|
|
ASSERT_EQ(IterStatus(iterators[2]), "c->m");
|
|
|
|
|
|
|
|
ASSERT_OK(Put(1, "x", "x"));
|
|
|
|
|
|
|
|
for (auto it : iterators) {
|
|
|
|
it->Next();
|
|
|
|
}
|
|
|
|
|
|
|
|
ASSERT_EQ(IterStatus(iterators[0]), "(invalid)");
|
|
|
|
if (iter == 0) {
|
|
|
|
// no tailing
|
|
|
|
ASSERT_EQ(IterStatus(iterators[1]), "(invalid)");
|
|
|
|
} else {
|
|
|
|
// tailing
|
|
|
|
ASSERT_EQ(IterStatus(iterators[1]), "x->x");
|
|
|
|
}
|
|
|
|
ASSERT_EQ(IterStatus(iterators[2]), "v->t");
|
|
|
|
|
|
|
|
for (auto it : iterators) {
|
|
|
|
delete it;
|
|
|
|
}
|
|
|
|
Destroy();
|
|
|
|
}
|
|
|
|
}
|
2015-06-29 21:39:01 +00:00
|
|
|
#endif // !ROCKSDB_LITE
|
2014-03-08 00:12:34 +00:00
|
|
|
|
2015-06-29 21:39:01 +00:00
|
|
|
#ifndef ROCKSDB_LITE // ReadOnlyDB is not supported
|
2015-03-17 21:08:00 +00:00
|
|
|
TEST_F(ColumnFamilyTest, ReadOnlyDBTest) {
|
2014-04-09 16:56:17 +00:00
|
|
|
Open();
|
|
|
|
CreateColumnFamiliesAndReopen({"one", "two", "three", "four"});
|
2014-07-23 20:52:11 +00:00
|
|
|
ASSERT_OK(Put(0, "a", "b"));
|
2014-04-09 16:56:17 +00:00
|
|
|
ASSERT_OK(Put(1, "foo", "bla"));
|
|
|
|
ASSERT_OK(Put(2, "foo", "blabla"));
|
|
|
|
ASSERT_OK(Put(3, "foo", "blablabla"));
|
|
|
|
ASSERT_OK(Put(4, "foo", "blablablabla"));
|
|
|
|
|
|
|
|
DropColumnFamilies({2});
|
|
|
|
Close();
|
|
|
|
// open only a subset of column families
|
|
|
|
AssertOpenReadOnly({"default", "one", "four"});
|
|
|
|
ASSERT_EQ("NOT_FOUND", Get(0, "foo"));
|
|
|
|
ASSERT_EQ("bla", Get(1, "foo"));
|
|
|
|
ASSERT_EQ("blablablabla", Get(2, "foo"));
|
|
|
|
|
2014-07-23 20:52:11 +00:00
|
|
|
|
|
|
|
// test newiterators
|
|
|
|
{
|
|
|
|
std::vector<Iterator*> iterators;
|
|
|
|
ASSERT_OK(db_->NewIterators(ReadOptions(), handles_, &iterators));
|
|
|
|
for (auto it : iterators) {
|
|
|
|
it->SeekToFirst();
|
|
|
|
}
|
|
|
|
ASSERT_EQ(IterStatus(iterators[0]), "a->b");
|
|
|
|
ASSERT_EQ(IterStatus(iterators[1]), "foo->bla");
|
|
|
|
ASSERT_EQ(IterStatus(iterators[2]), "foo->blablablabla");
|
|
|
|
for (auto it : iterators) {
|
|
|
|
it->Next();
|
|
|
|
}
|
|
|
|
ASSERT_EQ(IterStatus(iterators[0]), "(invalid)");
|
|
|
|
ASSERT_EQ(IterStatus(iterators[1]), "(invalid)");
|
|
|
|
ASSERT_EQ(IterStatus(iterators[2]), "(invalid)");
|
|
|
|
|
|
|
|
for (auto it : iterators) {
|
|
|
|
delete it;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2014-04-09 16:56:17 +00:00
|
|
|
Close();
|
|
|
|
// can't open dropped column family
|
|
|
|
Status s = OpenReadOnly({"default", "one", "two"});
|
|
|
|
ASSERT_TRUE(!s.ok());
|
|
|
|
|
|
|
|
// Can't open without specifying default column family
|
|
|
|
s = OpenReadOnly({"one", "four"});
|
|
|
|
ASSERT_TRUE(!s.ok());
|
|
|
|
}
|
2015-06-29 21:39:01 +00:00
|
|
|
#endif // !ROCKSDB_LITE
|
2014-04-09 16:56:17 +00:00
|
|
|
|
2015-03-17 21:08:00 +00:00
|
|
|
TEST_F(ColumnFamilyTest, DontRollEmptyLogs) {
|
2014-04-15 16:57:25 +00:00
|
|
|
Open();
|
|
|
|
CreateColumnFamiliesAndReopen({"one", "two", "three", "four"});
|
|
|
|
|
2014-04-29 19:33:57 +00:00
|
|
|
for (size_t i = 0; i < handles_.size(); ++i) {
|
2014-11-11 21:47:22 +00:00
|
|
|
PutRandomData(static_cast<int>(i), 10, 100);
|
2014-04-15 16:57:25 +00:00
|
|
|
}
|
|
|
|
int num_writable_file_start = env_->GetNumberOfNewWritableFileCalls();
|
|
|
|
// this will trigger the flushes
|
2014-11-11 21:47:22 +00:00
|
|
|
for (int i = 0; i <= 4; ++i) {
|
2014-08-12 05:10:32 +00:00
|
|
|
ASSERT_OK(Flush(i));
|
|
|
|
}
|
2014-04-15 16:57:25 +00:00
|
|
|
|
|
|
|
for (int i = 0; i < 4; ++i) {
|
2015-06-29 21:39:01 +00:00
|
|
|
WaitForFlush(i);
|
2014-04-15 16:57:25 +00:00
|
|
|
}
|
|
|
|
int total_new_writable_files =
|
|
|
|
env_->GetNumberOfNewWritableFileCalls() - num_writable_file_start;
|
2014-04-29 19:47:48 +00:00
|
|
|
ASSERT_EQ(static_cast<size_t>(total_new_writable_files), handles_.size() + 1);
|
2014-04-16 00:12:18 +00:00
|
|
|
Close();
|
2014-04-15 16:57:25 +00:00
|
|
|
}
|
|
|
|
|
2015-03-17 21:08:00 +00:00
|
|
|
TEST_F(ColumnFamilyTest, FlushStaleColumnFamilies) {
|
2014-04-30 18:33:40 +00:00
|
|
|
Open();
|
|
|
|
CreateColumnFamilies({"one", "two"});
|
|
|
|
ColumnFamilyOptions default_cf, one, two;
|
|
|
|
default_cf.write_buffer_size = 100000; // small write buffer size
|
2015-08-26 21:19:31 +00:00
|
|
|
default_cf.arena_block_size = 4096;
|
2014-04-30 18:33:40 +00:00
|
|
|
default_cf.disable_auto_compactions = true;
|
|
|
|
one.disable_auto_compactions = true;
|
|
|
|
two.disable_auto_compactions = true;
|
|
|
|
db_options_.max_total_wal_size = 210000;
|
|
|
|
|
|
|
|
Reopen({default_cf, one, two});
|
|
|
|
|
|
|
|
PutRandomData(2, 1, 10); // 10 bytes
|
|
|
|
for (int i = 0; i < 2; ++i) {
|
|
|
|
PutRandomData(0, 100, 1000); // flush
|
|
|
|
WaitForFlush(0);
|
2015-06-29 21:39:01 +00:00
|
|
|
|
|
|
|
AssertCountLiveFiles(i + 1);
|
2014-04-30 18:33:40 +00:00
|
|
|
}
|
|
|
|
// third flush. now, CF [two] should be detected as stale and flushed
|
|
|
|
// column family 1 should not be flushed since it's empty
|
|
|
|
PutRandomData(0, 100, 1000); // flush
|
|
|
|
WaitForFlush(0);
|
|
|
|
WaitForFlush(2);
|
|
|
|
// 3 files for default column families, 1 file for column family [two], zero
|
|
|
|
// files for column family [one], because it's empty
|
2015-06-29 21:39:01 +00:00
|
|
|
AssertCountLiveFiles(4);
|
2016-06-28 00:42:14 +00:00
|
|
|
|
|
|
|
Flush(0);
|
|
|
|
ASSERT_EQ(0, dbfull()->TEST_total_log_size());
|
2014-04-30 18:33:40 +00:00
|
|
|
Close();
|
|
|
|
}
|
|
|
|
|
2015-03-17 21:08:00 +00:00
|
|
|
TEST_F(ColumnFamilyTest, CreateMissingColumnFamilies) {
|
2014-06-07 01:04:56 +00:00
|
|
|
Status s = TryOpen({"one", "two"});
|
|
|
|
ASSERT_TRUE(!s.ok());
|
|
|
|
db_options_.create_missing_column_families = true;
|
|
|
|
s = TryOpen({"default", "one", "two"});
|
|
|
|
ASSERT_TRUE(s.ok());
|
|
|
|
Close();
|
|
|
|
}
|
|
|
|
|
2015-03-17 21:08:00 +00:00
|
|
|
TEST_F(ColumnFamilyTest, SanitizeOptions) {
|
options.level_compaction_dynamic_level_bytes to allow RocksDB to pick size bases of levels dynamically.
Summary:
When having fixed max_bytes_for_level_base, the ratio of size of largest level and the second one can range from 0 to the multiplier. This makes LSM tree frequently irregular and unpredictable. It can also cause poor space amplification in some cases.
In this improvement (proposed by Igor Kabiljo), we introduce a parameter option.level_compaction_use_dynamic_max_bytes. When turning it on, RocksDB is free to pick a level base in the range of (options.max_bytes_for_level_base/options.max_bytes_for_level_multiplier, options.max_bytes_for_level_base] so that real level ratios are close to options.max_bytes_for_level_multiplier.
Test Plan: New unit tests and pass tests suites including valgrind.
Reviewers: MarkCallaghan, rven, yhchiang, igor, ikabiljo
Reviewed By: ikabiljo
Subscribers: yoshinorim, ikabiljo, dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D31437
2015-02-05 19:44:17 +00:00
|
|
|
DBOptions db_options;
|
2015-05-22 18:35:40 +00:00
|
|
|
for (int s = kCompactionStyleLevel; s <= kCompactionStyleUniversal; ++s) {
|
|
|
|
for (int l = 0; l <= 2; l++) {
|
|
|
|
for (int i = 1; i <= 3; i++) {
|
|
|
|
for (int j = 1; j <= 3; j++) {
|
|
|
|
for (int k = 1; k <= 3; k++) {
|
|
|
|
ColumnFamilyOptions original;
|
|
|
|
original.compaction_style = static_cast<CompactionStyle>(s);
|
|
|
|
original.num_levels = l;
|
|
|
|
original.level0_stop_writes_trigger = i;
|
|
|
|
original.level0_slowdown_writes_trigger = j;
|
|
|
|
original.level0_file_num_compaction_trigger = k;
|
2015-09-01 21:43:23 +00:00
|
|
|
original.write_buffer_size =
|
|
|
|
l * 4 * 1024 * 1024 + i * 1024 * 1024 + j * 1024 + k;
|
|
|
|
|
2015-05-22 18:35:40 +00:00
|
|
|
ColumnFamilyOptions result =
|
|
|
|
SanitizeOptions(db_options, nullptr, original);
|
|
|
|
ASSERT_TRUE(result.level0_stop_writes_trigger >=
|
|
|
|
result.level0_slowdown_writes_trigger);
|
|
|
|
ASSERT_TRUE(result.level0_slowdown_writes_trigger >=
|
|
|
|
result.level0_file_num_compaction_trigger);
|
|
|
|
ASSERT_TRUE(result.level0_file_num_compaction_trigger ==
|
|
|
|
original.level0_file_num_compaction_trigger);
|
|
|
|
if (s == kCompactionStyleLevel) {
|
|
|
|
ASSERT_GE(result.num_levels, 2);
|
|
|
|
} else {
|
|
|
|
ASSERT_GE(result.num_levels, 1);
|
|
|
|
if (original.num_levels >= 1) {
|
|
|
|
ASSERT_EQ(result.num_levels, original.num_levels);
|
|
|
|
}
|
|
|
|
}
|
2015-09-01 21:43:23 +00:00
|
|
|
|
|
|
|
// Make sure Sanitize options sets arena_block_size to 1/8 of
|
|
|
|
// the write_buffer_size, rounded up to a multiple of 4k.
|
|
|
|
size_t expected_arena_block_size =
|
|
|
|
l * 4 * 1024 * 1024 / 8 + i * 1024 * 1024 / 8;
|
|
|
|
if (j + k != 0) {
|
|
|
|
// not a multiple of 4k, round up 4k
|
|
|
|
expected_arena_block_size += 4 * 1024;
|
|
|
|
}
|
|
|
|
ASSERT_EQ(expected_arena_block_size, result.arena_block_size);
|
2015-05-22 18:35:40 +00:00
|
|
|
}
|
|
|
|
}
|
2015-02-24 00:08:27 +00:00
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2015-03-20 00:04:29 +00:00
|
|
|
TEST_F(ColumnFamilyTest, ReadDroppedColumnFamily) {
|
|
|
|
// iter 0 -- drop CF, don't reopen
|
|
|
|
// iter 1 -- delete CF, reopen
|
|
|
|
for (int iter = 0; iter < 2; ++iter) {
|
|
|
|
db_options_.create_missing_column_families = true;
|
|
|
|
db_options_.max_open_files = 20;
|
|
|
|
// delete obsolete files always
|
|
|
|
db_options_.delete_obsolete_files_period_micros = 0;
|
|
|
|
Open({"default", "one", "two"});
|
|
|
|
ColumnFamilyOptions options;
|
|
|
|
options.level0_file_num_compaction_trigger = 100;
|
|
|
|
options.level0_slowdown_writes_trigger = 200;
|
|
|
|
options.level0_stop_writes_trigger = 200;
|
|
|
|
options.write_buffer_size = 100000; // small write buffer size
|
|
|
|
Reopen({options, options, options});
|
|
|
|
|
|
|
|
// 1MB should create ~10 files for each CF
|
|
|
|
int kKeysNum = 10000;
|
|
|
|
PutRandomData(0, kKeysNum, 100);
|
|
|
|
PutRandomData(1, kKeysNum, 100);
|
|
|
|
PutRandomData(2, kKeysNum, 100);
|
|
|
|
|
2016-02-23 22:54:05 +00:00
|
|
|
{
|
|
|
|
std::unique_ptr<Iterator> iterator(
|
|
|
|
db_->NewIterator(ReadOptions(), handles_[2]));
|
|
|
|
iterator->SeekToFirst();
|
|
|
|
|
|
|
|
if (iter == 0) {
|
|
|
|
// Drop CF two
|
|
|
|
ASSERT_OK(db_->DropColumnFamily(handles_[2]));
|
|
|
|
} else {
|
|
|
|
// delete CF two
|
2016-07-14 00:59:25 +00:00
|
|
|
db_->DestroyColumnFamilyHandle(handles_[2]);
|
2016-02-23 22:54:05 +00:00
|
|
|
handles_[2] = nullptr;
|
|
|
|
}
|
|
|
|
// Make sure iterator created can still be used.
|
|
|
|
int count = 0;
|
|
|
|
for (; iterator->Valid(); iterator->Next()) {
|
|
|
|
ASSERT_OK(iterator->status());
|
|
|
|
++count;
|
|
|
|
}
|
|
|
|
ASSERT_OK(iterator->status());
|
|
|
|
ASSERT_EQ(count, kKeysNum);
|
2015-03-20 00:04:29 +00:00
|
|
|
}
|
|
|
|
|
|
|
|
// Add bunch more data to other CFs
|
|
|
|
PutRandomData(0, kKeysNum, 100);
|
|
|
|
PutRandomData(1, kKeysNum, 100);
|
|
|
|
|
|
|
|
if (iter == 1) {
|
|
|
|
Reopen();
|
|
|
|
}
|
|
|
|
|
|
|
|
// Since we didn't delete CF handle, RocksDB's contract guarantees that
|
|
|
|
// we're still able to read dropped CF
|
|
|
|
for (int i = 0; i < 3; ++i) {
|
|
|
|
std::unique_ptr<Iterator> iterator(
|
|
|
|
db_->NewIterator(ReadOptions(), handles_[i]));
|
|
|
|
int count = 0;
|
|
|
|
for (iterator->SeekToFirst(); iterator->Valid(); iterator->Next()) {
|
|
|
|
ASSERT_OK(iterator->status());
|
|
|
|
++count;
|
|
|
|
}
|
2015-09-10 21:17:12 +00:00
|
|
|
ASSERT_OK(iterator->status());
|
2015-03-20 00:04:29 +00:00
|
|
|
ASSERT_EQ(count, kKeysNum * ((i == 2) ? 1 : 2));
|
|
|
|
}
|
|
|
|
|
|
|
|
Close();
|
|
|
|
Destroy();
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
LogAndApply() should fail if the column family has been dropped
Summary:
This patch finally fixes the ColumnFamilyTest.ReadDroppedColumnFamily test. The test has been failing very sporadically and it was hard to repro. However, I managed to write a new tests that reproes the failure deterministically.
Here's what happens:
1. We start the flush for the column family
2. We check if the column family was dropped here: https://github.com/facebook/rocksdb/blob/a3fc49bfddcdb1ff29409aacd06c04df56c7a1d7/db/flush_job.cc#L149
3. This check goes through, ends up in InstallMemtableFlushResults() and it goes into LogAndApply()
4. At about this time, we start dropping the column family. Dropping the column family process gets to LogAndApply() at about the same time as LogAndApply() from flush process
5. Drop column family goes through LogAndApply() first, marking the column family as dropped.
6. Flush process gets woken up and gets a chance to write to the MANIFEST. However, this is where it gets stuck: https://github.com/facebook/rocksdb/blob/a3fc49bfddcdb1ff29409aacd06c04df56c7a1d7/db/version_set.cc#L1975
7. We see that the column family was dropped, so there is no need to write to the MANIFEST. We return OK.
8. Flush gets OK back from LogAndApply() and it deletes the memtable, thinking that the data is now safely persisted to sst file.
The fix is pretty simple. Instead of OK, we return ShutdownInProgress. This is not really true, but we have been using this status code to also mean "this operation was canceled because the column family has been dropped".
The fix is only one LOC. All other code is related to tests. I added a new test that reproes the failure. I also moved SleepingBackgroundTask to util/testutil.h (because I needed it in column_family_test for my new test). There's plenty of other places where we reimplement SleepingBackgroundTask, but I'll address that in a separate commit.
Test Plan:
1. new test
2. make check
3. Make sure the ColumnFamilyTest.ReadDroppedColumnFamily doesn't fail on Travis: https://travis-ci.org/facebook/rocksdb/jobs/79952386
Reviewers: yhchiang, anthony, IslamAbdelRahman, kradhakrishnan, rven, sdong
Reviewed By: sdong
Subscribers: dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D46773
2015-09-15 18:28:44 +00:00
|
|
|
TEST_F(ColumnFamilyTest, FlushAndDropRaceCondition) {
|
|
|
|
db_options_.create_missing_column_families = true;
|
|
|
|
Open({"default", "one"});
|
|
|
|
ColumnFamilyOptions options;
|
|
|
|
options.level0_file_num_compaction_trigger = 100;
|
|
|
|
options.level0_slowdown_writes_trigger = 200;
|
|
|
|
options.level0_stop_writes_trigger = 200;
|
|
|
|
options.max_write_buffer_number = 20;
|
|
|
|
options.write_buffer_size = 100000; // small write buffer size
|
|
|
|
Reopen({options, options});
|
|
|
|
|
|
|
|
rocksdb::SyncPoint::GetInstance()->LoadDependency(
|
2016-01-26 17:12:20 +00:00
|
|
|
{{"VersionSet::LogAndApply::ColumnFamilyDrop:0",
|
|
|
|
"FlushJob::WriteLevel0Table"},
|
|
|
|
{"VersionSet::LogAndApply::ColumnFamilyDrop:1",
|
LogAndApply() should fail if the column family has been dropped
Summary:
This patch finally fixes the ColumnFamilyTest.ReadDroppedColumnFamily test. The test has been failing very sporadically and it was hard to repro. However, I managed to write a new tests that reproes the failure deterministically.
Here's what happens:
1. We start the flush for the column family
2. We check if the column family was dropped here: https://github.com/facebook/rocksdb/blob/a3fc49bfddcdb1ff29409aacd06c04df56c7a1d7/db/flush_job.cc#L149
3. This check goes through, ends up in InstallMemtableFlushResults() and it goes into LogAndApply()
4. At about this time, we start dropping the column family. Dropping the column family process gets to LogAndApply() at about the same time as LogAndApply() from flush process
5. Drop column family goes through LogAndApply() first, marking the column family as dropped.
6. Flush process gets woken up and gets a chance to write to the MANIFEST. However, this is where it gets stuck: https://github.com/facebook/rocksdb/blob/a3fc49bfddcdb1ff29409aacd06c04df56c7a1d7/db/version_set.cc#L1975
7. We see that the column family was dropped, so there is no need to write to the MANIFEST. We return OK.
8. Flush gets OK back from LogAndApply() and it deletes the memtable, thinking that the data is now safely persisted to sst file.
The fix is pretty simple. Instead of OK, we return ShutdownInProgress. This is not really true, but we have been using this status code to also mean "this operation was canceled because the column family has been dropped".
The fix is only one LOC. All other code is related to tests. I added a new test that reproes the failure. I also moved SleepingBackgroundTask to util/testutil.h (because I needed it in column_family_test for my new test). There's plenty of other places where we reimplement SleepingBackgroundTask, but I'll address that in a separate commit.
Test Plan:
1. new test
2. make check
3. Make sure the ColumnFamilyTest.ReadDroppedColumnFamily doesn't fail on Travis: https://travis-ci.org/facebook/rocksdb/jobs/79952386
Reviewers: yhchiang, anthony, IslamAbdelRahman, kradhakrishnan, rven, sdong
Reviewed By: sdong
Subscribers: dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D46773
2015-09-15 18:28:44 +00:00
|
|
|
"FlushJob::InstallResults"},
|
|
|
|
{"FlushJob::InstallResults",
|
2015-12-30 13:49:06 +00:00
|
|
|
"VersionSet::LogAndApply::ColumnFamilyDrop:2"}});
|
LogAndApply() should fail if the column family has been dropped
Summary:
This patch finally fixes the ColumnFamilyTest.ReadDroppedColumnFamily test. The test has been failing very sporadically and it was hard to repro. However, I managed to write a new tests that reproes the failure deterministically.
Here's what happens:
1. We start the flush for the column family
2. We check if the column family was dropped here: https://github.com/facebook/rocksdb/blob/a3fc49bfddcdb1ff29409aacd06c04df56c7a1d7/db/flush_job.cc#L149
3. This check goes through, ends up in InstallMemtableFlushResults() and it goes into LogAndApply()
4. At about this time, we start dropping the column family. Dropping the column family process gets to LogAndApply() at about the same time as LogAndApply() from flush process
5. Drop column family goes through LogAndApply() first, marking the column family as dropped.
6. Flush process gets woken up and gets a chance to write to the MANIFEST. However, this is where it gets stuck: https://github.com/facebook/rocksdb/blob/a3fc49bfddcdb1ff29409aacd06c04df56c7a1d7/db/version_set.cc#L1975
7. We see that the column family was dropped, so there is no need to write to the MANIFEST. We return OK.
8. Flush gets OK back from LogAndApply() and it deletes the memtable, thinking that the data is now safely persisted to sst file.
The fix is pretty simple. Instead of OK, we return ShutdownInProgress. This is not really true, but we have been using this status code to also mean "this operation was canceled because the column family has been dropped".
The fix is only one LOC. All other code is related to tests. I added a new test that reproes the failure. I also moved SleepingBackgroundTask to util/testutil.h (because I needed it in column_family_test for my new test). There's plenty of other places where we reimplement SleepingBackgroundTask, but I'll address that in a separate commit.
Test Plan:
1. new test
2. make check
3. Make sure the ColumnFamilyTest.ReadDroppedColumnFamily doesn't fail on Travis: https://travis-ci.org/facebook/rocksdb/jobs/79952386
Reviewers: yhchiang, anthony, IslamAbdelRahman, kradhakrishnan, rven, sdong
Reviewed By: sdong
Subscribers: dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D46773
2015-09-15 18:28:44 +00:00
|
|
|
|
|
|
|
rocksdb::SyncPoint::GetInstance()->EnableProcessing();
|
|
|
|
test::SleepingBackgroundTask sleeping_task;
|
|
|
|
|
|
|
|
env_->Schedule(&test::SleepingBackgroundTask::DoSleepTask, &sleeping_task,
|
|
|
|
Env::Priority::HIGH);
|
|
|
|
|
|
|
|
// 1MB should create ~10 files for each CF
|
|
|
|
int kKeysNum = 10000;
|
|
|
|
PutRandomData(1, kKeysNum, 100);
|
|
|
|
|
|
|
|
std::vector<std::thread> threads;
|
|
|
|
threads.emplace_back([&] { ASSERT_OK(db_->DropColumnFamily(handles_[1])); });
|
|
|
|
|
|
|
|
sleeping_task.WakeUp();
|
|
|
|
sleeping_task.WaitUntilDone();
|
|
|
|
sleeping_task.Reset();
|
|
|
|
// now we sleep again. this is just so we're certain that flush job finished
|
|
|
|
env_->Schedule(&test::SleepingBackgroundTask::DoSleepTask, &sleeping_task,
|
|
|
|
Env::Priority::HIGH);
|
|
|
|
sleeping_task.WakeUp();
|
|
|
|
sleeping_task.WaitUntilDone();
|
|
|
|
|
|
|
|
{
|
|
|
|
// Since we didn't delete CF handle, RocksDB's contract guarantees that
|
|
|
|
// we're still able to read dropped CF
|
|
|
|
std::unique_ptr<Iterator> iterator(
|
|
|
|
db_->NewIterator(ReadOptions(), handles_[1]));
|
|
|
|
int count = 0;
|
|
|
|
for (iterator->SeekToFirst(); iterator->Valid(); iterator->Next()) {
|
|
|
|
ASSERT_OK(iterator->status());
|
|
|
|
++count;
|
|
|
|
}
|
|
|
|
ASSERT_OK(iterator->status());
|
|
|
|
ASSERT_EQ(count, kKeysNum);
|
|
|
|
}
|
|
|
|
for (auto& t : threads) {
|
|
|
|
t.join();
|
|
|
|
}
|
|
|
|
|
|
|
|
Close();
|
|
|
|
Destroy();
|
|
|
|
}
|
|
|
|
|
2015-12-09 01:01:02 +00:00
|
|
|
#ifndef ROCKSDB_LITE
|
|
|
|
// skipped as persisting options is not supported in ROCKSDB_LITE
|
|
|
|
namespace {
|
|
|
|
std::atomic<int> test_stage(0);
|
|
|
|
const int kMainThreadStartPersistingOptionsFile = 1;
|
|
|
|
const int kChildThreadFinishDroppingColumnFamily = 2;
|
|
|
|
const int kChildThreadWaitingMainThreadPersistOptions = 3;
|
|
|
|
void DropSingleColumnFamily(ColumnFamilyTest* cf_test, int cf_id,
|
2015-12-10 19:53:53 +00:00
|
|
|
std::vector<Comparator*>* comparators) {
|
2015-12-09 01:01:02 +00:00
|
|
|
while (test_stage < kMainThreadStartPersistingOptionsFile) {
|
|
|
|
Env::Default()->SleepForMicroseconds(100);
|
|
|
|
}
|
|
|
|
cf_test->DropColumnFamilies({cf_id});
|
2015-12-10 19:53:53 +00:00
|
|
|
if ((*comparators)[cf_id]) {
|
|
|
|
delete (*comparators)[cf_id];
|
|
|
|
(*comparators)[cf_id] = nullptr;
|
|
|
|
}
|
2015-12-09 01:01:02 +00:00
|
|
|
test_stage = kChildThreadFinishDroppingColumnFamily;
|
|
|
|
}
|
|
|
|
} // namespace
|
|
|
|
|
|
|
|
TEST_F(ColumnFamilyTest, CreateAndDropRace) {
|
|
|
|
const int kCfCount = 5;
|
|
|
|
std::vector<ColumnFamilyOptions> cf_opts;
|
|
|
|
std::vector<Comparator*> comparators;
|
|
|
|
for (int i = 0; i < kCfCount; ++i) {
|
|
|
|
cf_opts.emplace_back();
|
|
|
|
comparators.push_back(new test::SimpleSuffixReverseComparator());
|
|
|
|
cf_opts.back().comparator = comparators.back();
|
|
|
|
}
|
|
|
|
db_options_.create_if_missing = true;
|
|
|
|
db_options_.create_missing_column_families = true;
|
|
|
|
|
|
|
|
auto main_thread_id = std::this_thread::get_id();
|
|
|
|
|
|
|
|
rocksdb::SyncPoint::GetInstance()->SetCallBack("PersistRocksDBOptions:start",
|
|
|
|
[&](void* arg) {
|
|
|
|
auto current_thread_id = std::this_thread::get_id();
|
|
|
|
// If it's the main thread hitting this sync-point, then it
|
|
|
|
// will be blocked until some other thread update the test_stage.
|
|
|
|
if (main_thread_id == current_thread_id) {
|
|
|
|
test_stage = kMainThreadStartPersistingOptionsFile;
|
|
|
|
while (test_stage < kChildThreadFinishDroppingColumnFamily) {
|
|
|
|
Env::Default()->SleepForMicroseconds(100);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
});
|
|
|
|
|
|
|
|
rocksdb::SyncPoint::GetInstance()->SetCallBack(
|
|
|
|
"WriteThread::EnterUnbatched:Wait", [&](void* arg) {
|
|
|
|
// This means a thread doing DropColumnFamily() is waiting for
|
|
|
|
// other thread to finish persisting options.
|
|
|
|
// In such case, we update the test_stage to unblock the main thread.
|
|
|
|
test_stage = kChildThreadWaitingMainThreadPersistOptions;
|
|
|
|
|
|
|
|
// Note that based on the test setting, this must not be the
|
|
|
|
// main thread.
|
|
|
|
ASSERT_NE(main_thread_id, std::this_thread::get_id());
|
|
|
|
});
|
|
|
|
|
|
|
|
// Create a database with four column families
|
|
|
|
Open({"default", "one", "two", "three"},
|
|
|
|
{cf_opts[0], cf_opts[1], cf_opts[2], cf_opts[3]});
|
|
|
|
|
|
|
|
rocksdb::SyncPoint::GetInstance()->EnableProcessing();
|
|
|
|
|
|
|
|
// Start a thread that will drop the first column family
|
|
|
|
// and its comparator
|
2015-12-10 19:53:53 +00:00
|
|
|
std::thread drop_cf_thread(DropSingleColumnFamily, this, 1, &comparators);
|
2015-12-09 01:01:02 +00:00
|
|
|
|
|
|
|
DropColumnFamilies({2});
|
|
|
|
|
|
|
|
drop_cf_thread.join();
|
|
|
|
Close();
|
|
|
|
Destroy();
|
2015-12-10 19:53:53 +00:00
|
|
|
for (auto* comparator : comparators) {
|
|
|
|
if (comparator) {
|
|
|
|
delete comparator;
|
|
|
|
}
|
|
|
|
}
|
2015-12-09 01:01:02 +00:00
|
|
|
}
|
|
|
|
#endif // !ROCKSDB_LITE
|
|
|
|
|
When slowdown is triggered, reduce the write rate
Summary: It's usually hard for users to set a value of options.delayed_write_rate. With this diff, after slowdown condition triggers, we greedily reduce write rate if estimated pending compaction bytes increase. If estimated compaction pending bytes drop, we increase the write rate.
Test Plan:
Add a unit test
Test with db_bench setting:
TEST_TMPDIR=/dev/shm/ ./db_bench --benchmarks=fillrandom -num=10000000 --soft_pending_compaction_bytes_limit=1000000000 --hard_pending_compaction_bytes_limit=3000000000 --delayed_write_rate=100000000
and make sure without the commit, write stop will happen, but with the commit, it will not happen.
Reviewers: igor, anthony, rven, yhchiang, kradhakrishnan, IslamAbdelRahman
Reviewed By: IslamAbdelRahman
Subscribers: leveldb, dhruba
Differential Revision: https://reviews.facebook.net/D52131
2015-12-18 01:07:44 +00:00
|
|
|
TEST_F(ColumnFamilyTest, WriteStallSingleColumnFamily) {
|
|
|
|
const uint64_t kBaseRate = 810000u;
|
|
|
|
db_options_.delayed_write_rate = kBaseRate;
|
Add options.base_background_compactions as a number of compaction threads for low compaction debt
Summary:
If options.base_background_compactions is given, we try to schedule number of compactions not existing this number, only when L0 files increase to certain number, or pending compaction bytes more than certain threshold, we schedule compactions based on options.max_background_compactions.
The watermarks are calculated based on slowdown thresholds.
Test Plan:
Add new test cases in column_family_test.
Adding more unit tests.
Reviewers: IslamAbdelRahman, yhchiang, kradhakrishnan, rven, anthony
Reviewed By: anthony
Subscribers: leveldb, dhruba, yoshinorim
Differential Revision: https://reviews.facebook.net/D53409
2016-01-28 19:56:16 +00:00
|
|
|
db_options_.base_background_compactions = 2;
|
|
|
|
db_options_.max_background_compactions = 6;
|
|
|
|
|
When slowdown is triggered, reduce the write rate
Summary: It's usually hard for users to set a value of options.delayed_write_rate. With this diff, after slowdown condition triggers, we greedily reduce write rate if estimated pending compaction bytes increase. If estimated compaction pending bytes drop, we increase the write rate.
Test Plan:
Add a unit test
Test with db_bench setting:
TEST_TMPDIR=/dev/shm/ ./db_bench --benchmarks=fillrandom -num=10000000 --soft_pending_compaction_bytes_limit=1000000000 --hard_pending_compaction_bytes_limit=3000000000 --delayed_write_rate=100000000
and make sure without the commit, write stop will happen, but with the commit, it will not happen.
Reviewers: igor, anthony, rven, yhchiang, kradhakrishnan, IslamAbdelRahman
Reviewed By: IslamAbdelRahman
Subscribers: leveldb, dhruba
Differential Revision: https://reviews.facebook.net/D52131
2015-12-18 01:07:44 +00:00
|
|
|
Open({"default"});
|
|
|
|
ColumnFamilyData* cfd =
|
|
|
|
static_cast<ColumnFamilyHandleImpl*>(db_->DefaultColumnFamily())->cfd();
|
|
|
|
|
|
|
|
VersionStorageInfo* vstorage = cfd->current()->storage_info();
|
|
|
|
|
|
|
|
MutableCFOptions mutable_cf_options(
|
|
|
|
Options(db_options_, column_family_options_),
|
|
|
|
ImmutableCFOptions(Options(db_options_, column_family_options_)));
|
|
|
|
|
|
|
|
mutable_cf_options.level0_slowdown_writes_trigger = 20;
|
|
|
|
mutable_cf_options.level0_stop_writes_trigger = 10000;
|
|
|
|
mutable_cf_options.soft_pending_compaction_bytes_limit = 200;
|
|
|
|
mutable_cf_options.hard_pending_compaction_bytes_limit = 2000;
|
2016-08-03 04:55:26 +00:00
|
|
|
mutable_cf_options.disable_auto_compactions = false;
|
When slowdown is triggered, reduce the write rate
Summary: It's usually hard for users to set a value of options.delayed_write_rate. With this diff, after slowdown condition triggers, we greedily reduce write rate if estimated pending compaction bytes increase. If estimated compaction pending bytes drop, we increase the write rate.
Test Plan:
Add a unit test
Test with db_bench setting:
TEST_TMPDIR=/dev/shm/ ./db_bench --benchmarks=fillrandom -num=10000000 --soft_pending_compaction_bytes_limit=1000000000 --hard_pending_compaction_bytes_limit=3000000000 --delayed_write_rate=100000000
and make sure without the commit, write stop will happen, but with the commit, it will not happen.
Reviewers: igor, anthony, rven, yhchiang, kradhakrishnan, IslamAbdelRahman
Reviewed By: IslamAbdelRahman
Subscribers: leveldb, dhruba
Differential Revision: https://reviews.facebook.net/D52131
2015-12-18 01:07:44 +00:00
|
|
|
|
|
|
|
vstorage->TEST_set_estimated_compaction_needed_bytes(50);
|
|
|
|
cfd->RecalculateWriteStallConditions(mutable_cf_options);
|
|
|
|
ASSERT_TRUE(!dbfull()->TEST_write_controler().IsStopped());
|
|
|
|
ASSERT_TRUE(!dbfull()->TEST_write_controler().NeedsDelay());
|
|
|
|
|
|
|
|
vstorage->TEST_set_estimated_compaction_needed_bytes(201);
|
|
|
|
cfd->RecalculateWriteStallConditions(mutable_cf_options);
|
|
|
|
ASSERT_TRUE(!dbfull()->TEST_write_controler().IsStopped());
|
|
|
|
ASSERT_TRUE(dbfull()->TEST_write_controler().NeedsDelay());
|
|
|
|
ASSERT_EQ(kBaseRate, dbfull()->TEST_write_controler().delayed_write_rate());
|
Add options.base_background_compactions as a number of compaction threads for low compaction debt
Summary:
If options.base_background_compactions is given, we try to schedule number of compactions not existing this number, only when L0 files increase to certain number, or pending compaction bytes more than certain threshold, we schedule compactions based on options.max_background_compactions.
The watermarks are calculated based on slowdown thresholds.
Test Plan:
Add new test cases in column_family_test.
Adding more unit tests.
Reviewers: IslamAbdelRahman, yhchiang, kradhakrishnan, rven, anthony
Reviewed By: anthony
Subscribers: leveldb, dhruba, yoshinorim
Differential Revision: https://reviews.facebook.net/D53409
2016-01-28 19:56:16 +00:00
|
|
|
ASSERT_EQ(6, dbfull()->BGCompactionsAllowed());
|
When slowdown is triggered, reduce the write rate
Summary: It's usually hard for users to set a value of options.delayed_write_rate. With this diff, after slowdown condition triggers, we greedily reduce write rate if estimated pending compaction bytes increase. If estimated compaction pending bytes drop, we increase the write rate.
Test Plan:
Add a unit test
Test with db_bench setting:
TEST_TMPDIR=/dev/shm/ ./db_bench --benchmarks=fillrandom -num=10000000 --soft_pending_compaction_bytes_limit=1000000000 --hard_pending_compaction_bytes_limit=3000000000 --delayed_write_rate=100000000
and make sure without the commit, write stop will happen, but with the commit, it will not happen.
Reviewers: igor, anthony, rven, yhchiang, kradhakrishnan, IslamAbdelRahman
Reviewed By: IslamAbdelRahman
Subscribers: leveldb, dhruba
Differential Revision: https://reviews.facebook.net/D52131
2015-12-18 01:07:44 +00:00
|
|
|
|
|
|
|
vstorage->TEST_set_estimated_compaction_needed_bytes(400);
|
|
|
|
cfd->RecalculateWriteStallConditions(mutable_cf_options);
|
|
|
|
ASSERT_TRUE(!dbfull()->TEST_write_controler().IsStopped());
|
|
|
|
ASSERT_TRUE(dbfull()->TEST_write_controler().NeedsDelay());
|
|
|
|
ASSERT_EQ(kBaseRate / 1.2,
|
|
|
|
dbfull()->TEST_write_controler().delayed_write_rate());
|
Add options.base_background_compactions as a number of compaction threads for low compaction debt
Summary:
If options.base_background_compactions is given, we try to schedule number of compactions not existing this number, only when L0 files increase to certain number, or pending compaction bytes more than certain threshold, we schedule compactions based on options.max_background_compactions.
The watermarks are calculated based on slowdown thresholds.
Test Plan:
Add new test cases in column_family_test.
Adding more unit tests.
Reviewers: IslamAbdelRahman, yhchiang, kradhakrishnan, rven, anthony
Reviewed By: anthony
Subscribers: leveldb, dhruba, yoshinorim
Differential Revision: https://reviews.facebook.net/D53409
2016-01-28 19:56:16 +00:00
|
|
|
ASSERT_EQ(6, dbfull()->BGCompactionsAllowed());
|
When slowdown is triggered, reduce the write rate
Summary: It's usually hard for users to set a value of options.delayed_write_rate. With this diff, after slowdown condition triggers, we greedily reduce write rate if estimated pending compaction bytes increase. If estimated compaction pending bytes drop, we increase the write rate.
Test Plan:
Add a unit test
Test with db_bench setting:
TEST_TMPDIR=/dev/shm/ ./db_bench --benchmarks=fillrandom -num=10000000 --soft_pending_compaction_bytes_limit=1000000000 --hard_pending_compaction_bytes_limit=3000000000 --delayed_write_rate=100000000
and make sure without the commit, write stop will happen, but with the commit, it will not happen.
Reviewers: igor, anthony, rven, yhchiang, kradhakrishnan, IslamAbdelRahman
Reviewed By: IslamAbdelRahman
Subscribers: leveldb, dhruba
Differential Revision: https://reviews.facebook.net/D52131
2015-12-18 01:07:44 +00:00
|
|
|
|
|
|
|
vstorage->TEST_set_estimated_compaction_needed_bytes(500);
|
|
|
|
cfd->RecalculateWriteStallConditions(mutable_cf_options);
|
|
|
|
ASSERT_TRUE(!dbfull()->TEST_write_controler().IsStopped());
|
|
|
|
ASSERT_TRUE(dbfull()->TEST_write_controler().NeedsDelay());
|
|
|
|
ASSERT_EQ(kBaseRate / 1.2 / 1.2,
|
|
|
|
dbfull()->TEST_write_controler().delayed_write_rate());
|
|
|
|
|
|
|
|
vstorage->TEST_set_estimated_compaction_needed_bytes(450);
|
|
|
|
cfd->RecalculateWriteStallConditions(mutable_cf_options);
|
|
|
|
ASSERT_TRUE(!dbfull()->TEST_write_controler().IsStopped());
|
|
|
|
ASSERT_TRUE(dbfull()->TEST_write_controler().NeedsDelay());
|
|
|
|
ASSERT_EQ(kBaseRate / 1.2,
|
|
|
|
dbfull()->TEST_write_controler().delayed_write_rate());
|
|
|
|
|
|
|
|
vstorage->TEST_set_estimated_compaction_needed_bytes(205);
|
|
|
|
cfd->RecalculateWriteStallConditions(mutable_cf_options);
|
|
|
|
ASSERT_TRUE(!dbfull()->TEST_write_controler().IsStopped());
|
|
|
|
ASSERT_TRUE(dbfull()->TEST_write_controler().NeedsDelay());
|
|
|
|
ASSERT_EQ(kBaseRate, dbfull()->TEST_write_controler().delayed_write_rate());
|
|
|
|
|
|
|
|
vstorage->TEST_set_estimated_compaction_needed_bytes(202);
|
|
|
|
cfd->RecalculateWriteStallConditions(mutable_cf_options);
|
|
|
|
ASSERT_TRUE(!dbfull()->TEST_write_controler().IsStopped());
|
|
|
|
ASSERT_TRUE(dbfull()->TEST_write_controler().NeedsDelay());
|
|
|
|
ASSERT_EQ(kBaseRate, dbfull()->TEST_write_controler().delayed_write_rate());
|
|
|
|
|
|
|
|
vstorage->TEST_set_estimated_compaction_needed_bytes(201);
|
|
|
|
cfd->RecalculateWriteStallConditions(mutable_cf_options);
|
|
|
|
ASSERT_TRUE(!dbfull()->TEST_write_controler().IsStopped());
|
|
|
|
ASSERT_TRUE(dbfull()->TEST_write_controler().NeedsDelay());
|
|
|
|
ASSERT_EQ(kBaseRate, dbfull()->TEST_write_controler().delayed_write_rate());
|
|
|
|
|
|
|
|
vstorage->TEST_set_estimated_compaction_needed_bytes(198);
|
|
|
|
cfd->RecalculateWriteStallConditions(mutable_cf_options);
|
|
|
|
ASSERT_TRUE(!dbfull()->TEST_write_controler().IsStopped());
|
|
|
|
ASSERT_TRUE(!dbfull()->TEST_write_controler().NeedsDelay());
|
|
|
|
|
|
|
|
vstorage->TEST_set_estimated_compaction_needed_bytes(399);
|
|
|
|
cfd->RecalculateWriteStallConditions(mutable_cf_options);
|
|
|
|
ASSERT_TRUE(!dbfull()->TEST_write_controler().IsStopped());
|
|
|
|
ASSERT_TRUE(dbfull()->TEST_write_controler().NeedsDelay());
|
|
|
|
ASSERT_EQ(kBaseRate, dbfull()->TEST_write_controler().delayed_write_rate());
|
|
|
|
|
|
|
|
vstorage->TEST_set_estimated_compaction_needed_bytes(599);
|
|
|
|
cfd->RecalculateWriteStallConditions(mutable_cf_options);
|
|
|
|
ASSERT_TRUE(!dbfull()->TEST_write_controler().IsStopped());
|
|
|
|
ASSERT_TRUE(dbfull()->TEST_write_controler().NeedsDelay());
|
|
|
|
ASSERT_EQ(kBaseRate / 1.2,
|
|
|
|
dbfull()->TEST_write_controler().delayed_write_rate());
|
|
|
|
|
|
|
|
vstorage->TEST_set_estimated_compaction_needed_bytes(2001);
|
|
|
|
cfd->RecalculateWriteStallConditions(mutable_cf_options);
|
|
|
|
ASSERT_TRUE(dbfull()->TEST_write_controler().IsStopped());
|
|
|
|
ASSERT_TRUE(!dbfull()->TEST_write_controler().NeedsDelay());
|
Add options.base_background_compactions as a number of compaction threads for low compaction debt
Summary:
If options.base_background_compactions is given, we try to schedule number of compactions not existing this number, only when L0 files increase to certain number, or pending compaction bytes more than certain threshold, we schedule compactions based on options.max_background_compactions.
The watermarks are calculated based on slowdown thresholds.
Test Plan:
Add new test cases in column_family_test.
Adding more unit tests.
Reviewers: IslamAbdelRahman, yhchiang, kradhakrishnan, rven, anthony
Reviewed By: anthony
Subscribers: leveldb, dhruba, yoshinorim
Differential Revision: https://reviews.facebook.net/D53409
2016-01-28 19:56:16 +00:00
|
|
|
ASSERT_EQ(6, dbfull()->BGCompactionsAllowed());
|
When slowdown is triggered, reduce the write rate
Summary: It's usually hard for users to set a value of options.delayed_write_rate. With this diff, after slowdown condition triggers, we greedily reduce write rate if estimated pending compaction bytes increase. If estimated compaction pending bytes drop, we increase the write rate.
Test Plan:
Add a unit test
Test with db_bench setting:
TEST_TMPDIR=/dev/shm/ ./db_bench --benchmarks=fillrandom -num=10000000 --soft_pending_compaction_bytes_limit=1000000000 --hard_pending_compaction_bytes_limit=3000000000 --delayed_write_rate=100000000
and make sure without the commit, write stop will happen, but with the commit, it will not happen.
Reviewers: igor, anthony, rven, yhchiang, kradhakrishnan, IslamAbdelRahman
Reviewed By: IslamAbdelRahman
Subscribers: leveldb, dhruba
Differential Revision: https://reviews.facebook.net/D52131
2015-12-18 01:07:44 +00:00
|
|
|
|
|
|
|
vstorage->TEST_set_estimated_compaction_needed_bytes(3001);
|
|
|
|
cfd->RecalculateWriteStallConditions(mutable_cf_options);
|
|
|
|
ASSERT_TRUE(dbfull()->TEST_write_controler().IsStopped());
|
|
|
|
ASSERT_TRUE(!dbfull()->TEST_write_controler().NeedsDelay());
|
|
|
|
|
|
|
|
vstorage->TEST_set_estimated_compaction_needed_bytes(390);
|
|
|
|
cfd->RecalculateWriteStallConditions(mutable_cf_options);
|
|
|
|
ASSERT_TRUE(!dbfull()->TEST_write_controler().IsStopped());
|
|
|
|
ASSERT_TRUE(dbfull()->TEST_write_controler().NeedsDelay());
|
|
|
|
ASSERT_EQ(kBaseRate / 1.2,
|
|
|
|
dbfull()->TEST_write_controler().delayed_write_rate());
|
|
|
|
|
|
|
|
vstorage->TEST_set_estimated_compaction_needed_bytes(100);
|
|
|
|
cfd->RecalculateWriteStallConditions(mutable_cf_options);
|
|
|
|
ASSERT_TRUE(!dbfull()->TEST_write_controler().IsStopped());
|
|
|
|
ASSERT_TRUE(!dbfull()->TEST_write_controler().NeedsDelay());
|
|
|
|
|
|
|
|
vstorage->set_l0_delay_trigger_count(100);
|
|
|
|
cfd->RecalculateWriteStallConditions(mutable_cf_options);
|
|
|
|
ASSERT_TRUE(!dbfull()->TEST_write_controler().IsStopped());
|
|
|
|
ASSERT_TRUE(dbfull()->TEST_write_controler().NeedsDelay());
|
|
|
|
ASSERT_EQ(kBaseRate / 1.2,
|
|
|
|
dbfull()->TEST_write_controler().delayed_write_rate());
|
Add options.base_background_compactions as a number of compaction threads for low compaction debt
Summary:
If options.base_background_compactions is given, we try to schedule number of compactions not existing this number, only when L0 files increase to certain number, or pending compaction bytes more than certain threshold, we schedule compactions based on options.max_background_compactions.
The watermarks are calculated based on slowdown thresholds.
Test Plan:
Add new test cases in column_family_test.
Adding more unit tests.
Reviewers: IslamAbdelRahman, yhchiang, kradhakrishnan, rven, anthony
Reviewed By: anthony
Subscribers: leveldb, dhruba, yoshinorim
Differential Revision: https://reviews.facebook.net/D53409
2016-01-28 19:56:16 +00:00
|
|
|
ASSERT_EQ(6, dbfull()->BGCompactionsAllowed());
|
When slowdown is triggered, reduce the write rate
Summary: It's usually hard for users to set a value of options.delayed_write_rate. With this diff, after slowdown condition triggers, we greedily reduce write rate if estimated pending compaction bytes increase. If estimated compaction pending bytes drop, we increase the write rate.
Test Plan:
Add a unit test
Test with db_bench setting:
TEST_TMPDIR=/dev/shm/ ./db_bench --benchmarks=fillrandom -num=10000000 --soft_pending_compaction_bytes_limit=1000000000 --hard_pending_compaction_bytes_limit=3000000000 --delayed_write_rate=100000000
and make sure without the commit, write stop will happen, but with the commit, it will not happen.
Reviewers: igor, anthony, rven, yhchiang, kradhakrishnan, IslamAbdelRahman
Reviewed By: IslamAbdelRahman
Subscribers: leveldb, dhruba
Differential Revision: https://reviews.facebook.net/D52131
2015-12-18 01:07:44 +00:00
|
|
|
|
|
|
|
vstorage->set_l0_delay_trigger_count(101);
|
|
|
|
cfd->RecalculateWriteStallConditions(mutable_cf_options);
|
|
|
|
ASSERT_TRUE(!dbfull()->TEST_write_controler().IsStopped());
|
|
|
|
ASSERT_TRUE(dbfull()->TEST_write_controler().NeedsDelay());
|
|
|
|
ASSERT_EQ(kBaseRate / 1.2 / 1.2,
|
|
|
|
dbfull()->TEST_write_controler().delayed_write_rate());
|
|
|
|
|
|
|
|
vstorage->set_l0_delay_trigger_count(0);
|
|
|
|
vstorage->TEST_set_estimated_compaction_needed_bytes(300);
|
|
|
|
cfd->RecalculateWriteStallConditions(mutable_cf_options);
|
|
|
|
ASSERT_TRUE(!dbfull()->TEST_write_controler().IsStopped());
|
|
|
|
ASSERT_TRUE(dbfull()->TEST_write_controler().NeedsDelay());
|
|
|
|
ASSERT_EQ(kBaseRate / 1.2 / 1.2 / 1.2,
|
|
|
|
dbfull()->TEST_write_controler().delayed_write_rate());
|
|
|
|
|
|
|
|
vstorage->set_l0_delay_trigger_count(101);
|
|
|
|
cfd->RecalculateWriteStallConditions(mutable_cf_options);
|
|
|
|
ASSERT_TRUE(!dbfull()->TEST_write_controler().IsStopped());
|
|
|
|
ASSERT_TRUE(dbfull()->TEST_write_controler().NeedsDelay());
|
|
|
|
ASSERT_EQ(kBaseRate / 1.2 / 1.2 / 1.2 / 1.2,
|
|
|
|
dbfull()->TEST_write_controler().delayed_write_rate());
|
|
|
|
|
|
|
|
vstorage->TEST_set_estimated_compaction_needed_bytes(200);
|
|
|
|
cfd->RecalculateWriteStallConditions(mutable_cf_options);
|
|
|
|
ASSERT_TRUE(!dbfull()->TEST_write_controler().IsStopped());
|
|
|
|
ASSERT_TRUE(dbfull()->TEST_write_controler().NeedsDelay());
|
|
|
|
ASSERT_EQ(kBaseRate / 1.2 / 1.2 / 1.2,
|
|
|
|
dbfull()->TEST_write_controler().delayed_write_rate());
|
|
|
|
|
|
|
|
vstorage->set_l0_delay_trigger_count(0);
|
|
|
|
vstorage->TEST_set_estimated_compaction_needed_bytes(0);
|
|
|
|
cfd->RecalculateWriteStallConditions(mutable_cf_options);
|
|
|
|
ASSERT_TRUE(!dbfull()->TEST_write_controler().IsStopped());
|
|
|
|
ASSERT_TRUE(!dbfull()->TEST_write_controler().NeedsDelay());
|
|
|
|
|
|
|
|
mutable_cf_options.disable_auto_compactions = true;
|
|
|
|
dbfull()->TEST_write_controler().set_delayed_write_rate(kBaseRate);
|
|
|
|
cfd->RecalculateWriteStallConditions(mutable_cf_options);
|
|
|
|
ASSERT_TRUE(!dbfull()->TEST_write_controler().IsStopped());
|
|
|
|
ASSERT_TRUE(!dbfull()->TEST_write_controler().NeedsDelay());
|
|
|
|
|
|
|
|
vstorage->set_l0_delay_trigger_count(50);
|
|
|
|
cfd->RecalculateWriteStallConditions(mutable_cf_options);
|
|
|
|
ASSERT_TRUE(!dbfull()->TEST_write_controler().IsStopped());
|
2016-08-03 04:55:26 +00:00
|
|
|
ASSERT_TRUE(!dbfull()->TEST_write_controler().NeedsDelay());
|
When slowdown is triggered, reduce the write rate
Summary: It's usually hard for users to set a value of options.delayed_write_rate. With this diff, after slowdown condition triggers, we greedily reduce write rate if estimated pending compaction bytes increase. If estimated compaction pending bytes drop, we increase the write rate.
Test Plan:
Add a unit test
Test with db_bench setting:
TEST_TMPDIR=/dev/shm/ ./db_bench --benchmarks=fillrandom -num=10000000 --soft_pending_compaction_bytes_limit=1000000000 --hard_pending_compaction_bytes_limit=3000000000 --delayed_write_rate=100000000
and make sure without the commit, write stop will happen, but with the commit, it will not happen.
Reviewers: igor, anthony, rven, yhchiang, kradhakrishnan, IslamAbdelRahman
Reviewed By: IslamAbdelRahman
Subscribers: leveldb, dhruba
Differential Revision: https://reviews.facebook.net/D52131
2015-12-18 01:07:44 +00:00
|
|
|
ASSERT_EQ(kBaseRate, dbfull()->TEST_write_controler().delayed_write_rate());
|
|
|
|
|
|
|
|
vstorage->set_l0_delay_trigger_count(60);
|
|
|
|
vstorage->TEST_set_estimated_compaction_needed_bytes(300);
|
|
|
|
cfd->RecalculateWriteStallConditions(mutable_cf_options);
|
|
|
|
ASSERT_TRUE(!dbfull()->TEST_write_controler().IsStopped());
|
2016-08-03 04:55:26 +00:00
|
|
|
ASSERT_TRUE(!dbfull()->TEST_write_controler().NeedsDelay());
|
When slowdown is triggered, reduce the write rate
Summary: It's usually hard for users to set a value of options.delayed_write_rate. With this diff, after slowdown condition triggers, we greedily reduce write rate if estimated pending compaction bytes increase. If estimated compaction pending bytes drop, we increase the write rate.
Test Plan:
Add a unit test
Test with db_bench setting:
TEST_TMPDIR=/dev/shm/ ./db_bench --benchmarks=fillrandom -num=10000000 --soft_pending_compaction_bytes_limit=1000000000 --hard_pending_compaction_bytes_limit=3000000000 --delayed_write_rate=100000000
and make sure without the commit, write stop will happen, but with the commit, it will not happen.
Reviewers: igor, anthony, rven, yhchiang, kradhakrishnan, IslamAbdelRahman
Reviewed By: IslamAbdelRahman
Subscribers: leveldb, dhruba
Differential Revision: https://reviews.facebook.net/D52131
2015-12-18 01:07:44 +00:00
|
|
|
ASSERT_EQ(kBaseRate, dbfull()->TEST_write_controler().delayed_write_rate());
|
|
|
|
|
2016-08-03 04:55:26 +00:00
|
|
|
mutable_cf_options.disable_auto_compactions = false;
|
When slowdown is triggered, reduce the write rate
Summary: It's usually hard for users to set a value of options.delayed_write_rate. With this diff, after slowdown condition triggers, we greedily reduce write rate if estimated pending compaction bytes increase. If estimated compaction pending bytes drop, we increase the write rate.
Test Plan:
Add a unit test
Test with db_bench setting:
TEST_TMPDIR=/dev/shm/ ./db_bench --benchmarks=fillrandom -num=10000000 --soft_pending_compaction_bytes_limit=1000000000 --hard_pending_compaction_bytes_limit=3000000000 --delayed_write_rate=100000000
and make sure without the commit, write stop will happen, but with the commit, it will not happen.
Reviewers: igor, anthony, rven, yhchiang, kradhakrishnan, IslamAbdelRahman
Reviewed By: IslamAbdelRahman
Subscribers: leveldb, dhruba
Differential Revision: https://reviews.facebook.net/D52131
2015-12-18 01:07:44 +00:00
|
|
|
vstorage->set_l0_delay_trigger_count(70);
|
|
|
|
vstorage->TEST_set_estimated_compaction_needed_bytes(500);
|
|
|
|
cfd->RecalculateWriteStallConditions(mutable_cf_options);
|
|
|
|
ASSERT_TRUE(!dbfull()->TEST_write_controler().IsStopped());
|
|
|
|
ASSERT_TRUE(dbfull()->TEST_write_controler().NeedsDelay());
|
|
|
|
ASSERT_EQ(kBaseRate, dbfull()->TEST_write_controler().delayed_write_rate());
|
|
|
|
|
|
|
|
vstorage->set_l0_delay_trigger_count(71);
|
|
|
|
vstorage->TEST_set_estimated_compaction_needed_bytes(501);
|
|
|
|
cfd->RecalculateWriteStallConditions(mutable_cf_options);
|
|
|
|
ASSERT_TRUE(!dbfull()->TEST_write_controler().IsStopped());
|
|
|
|
ASSERT_TRUE(dbfull()->TEST_write_controler().NeedsDelay());
|
|
|
|
ASSERT_EQ(kBaseRate / 1.2,
|
|
|
|
dbfull()->TEST_write_controler().delayed_write_rate());
|
|
|
|
}
|
|
|
|
|
Add options.base_background_compactions as a number of compaction threads for low compaction debt
Summary:
If options.base_background_compactions is given, we try to schedule number of compactions not existing this number, only when L0 files increase to certain number, or pending compaction bytes more than certain threshold, we schedule compactions based on options.max_background_compactions.
The watermarks are calculated based on slowdown thresholds.
Test Plan:
Add new test cases in column_family_test.
Adding more unit tests.
Reviewers: IslamAbdelRahman, yhchiang, kradhakrishnan, rven, anthony
Reviewed By: anthony
Subscribers: leveldb, dhruba, yoshinorim
Differential Revision: https://reviews.facebook.net/D53409
2016-01-28 19:56:16 +00:00
|
|
|
TEST_F(ColumnFamilyTest, CompactionSpeedupSingleColumnFamily) {
|
|
|
|
db_options_.base_background_compactions = 2;
|
|
|
|
db_options_.max_background_compactions = 6;
|
|
|
|
Open({"default"});
|
|
|
|
ColumnFamilyData* cfd =
|
|
|
|
static_cast<ColumnFamilyHandleImpl*>(db_->DefaultColumnFamily())->cfd();
|
|
|
|
|
|
|
|
VersionStorageInfo* vstorage = cfd->current()->storage_info();
|
|
|
|
|
|
|
|
MutableCFOptions mutable_cf_options(
|
|
|
|
Options(db_options_, column_family_options_),
|
|
|
|
ImmutableCFOptions(Options(db_options_, column_family_options_)));
|
|
|
|
|
|
|
|
// Speed up threshold = min(4 * 2, 4 + (36 - 4)/4) = 8
|
|
|
|
mutable_cf_options.level0_file_num_compaction_trigger = 4;
|
|
|
|
mutable_cf_options.level0_slowdown_writes_trigger = 36;
|
|
|
|
mutable_cf_options.level0_stop_writes_trigger = 50;
|
|
|
|
// Speedup threshold = 200 / 4 = 50
|
|
|
|
mutable_cf_options.soft_pending_compaction_bytes_limit = 200;
|
|
|
|
mutable_cf_options.hard_pending_compaction_bytes_limit = 2000;
|
|
|
|
|
|
|
|
vstorage->TEST_set_estimated_compaction_needed_bytes(40);
|
|
|
|
cfd->RecalculateWriteStallConditions(mutable_cf_options);
|
|
|
|
ASSERT_EQ(2, dbfull()->BGCompactionsAllowed());
|
|
|
|
|
|
|
|
vstorage->TEST_set_estimated_compaction_needed_bytes(50);
|
|
|
|
cfd->RecalculateWriteStallConditions(mutable_cf_options);
|
|
|
|
ASSERT_EQ(6, dbfull()->BGCompactionsAllowed());
|
|
|
|
|
|
|
|
vstorage->TEST_set_estimated_compaction_needed_bytes(300);
|
|
|
|
cfd->RecalculateWriteStallConditions(mutable_cf_options);
|
|
|
|
ASSERT_EQ(6, dbfull()->BGCompactionsAllowed());
|
|
|
|
|
|
|
|
vstorage->TEST_set_estimated_compaction_needed_bytes(45);
|
|
|
|
cfd->RecalculateWriteStallConditions(mutable_cf_options);
|
|
|
|
ASSERT_EQ(2, dbfull()->BGCompactionsAllowed());
|
|
|
|
|
|
|
|
vstorage->set_l0_delay_trigger_count(7);
|
|
|
|
cfd->RecalculateWriteStallConditions(mutable_cf_options);
|
|
|
|
ASSERT_EQ(2, dbfull()->BGCompactionsAllowed());
|
|
|
|
|
|
|
|
vstorage->set_l0_delay_trigger_count(9);
|
|
|
|
cfd->RecalculateWriteStallConditions(mutable_cf_options);
|
|
|
|
ASSERT_EQ(6, dbfull()->BGCompactionsAllowed());
|
|
|
|
|
|
|
|
vstorage->set_l0_delay_trigger_count(6);
|
|
|
|
cfd->RecalculateWriteStallConditions(mutable_cf_options);
|
|
|
|
ASSERT_EQ(2, dbfull()->BGCompactionsAllowed());
|
|
|
|
|
|
|
|
// Speed up threshold = min(4 * 2, 4 + (12 - 4)/4) = 6
|
|
|
|
mutable_cf_options.level0_file_num_compaction_trigger = 4;
|
|
|
|
mutable_cf_options.level0_slowdown_writes_trigger = 16;
|
|
|
|
mutable_cf_options.level0_stop_writes_trigger = 30;
|
|
|
|
|
|
|
|
vstorage->set_l0_delay_trigger_count(5);
|
|
|
|
cfd->RecalculateWriteStallConditions(mutable_cf_options);
|
|
|
|
ASSERT_EQ(2, dbfull()->BGCompactionsAllowed());
|
|
|
|
|
|
|
|
vstorage->set_l0_delay_trigger_count(7);
|
|
|
|
cfd->RecalculateWriteStallConditions(mutable_cf_options);
|
|
|
|
ASSERT_EQ(6, dbfull()->BGCompactionsAllowed());
|
|
|
|
|
|
|
|
vstorage->set_l0_delay_trigger_count(3);
|
|
|
|
cfd->RecalculateWriteStallConditions(mutable_cf_options);
|
|
|
|
ASSERT_EQ(2, dbfull()->BGCompactionsAllowed());
|
|
|
|
}
|
|
|
|
|
When slowdown is triggered, reduce the write rate
Summary: It's usually hard for users to set a value of options.delayed_write_rate. With this diff, after slowdown condition triggers, we greedily reduce write rate if estimated pending compaction bytes increase. If estimated compaction pending bytes drop, we increase the write rate.
Test Plan:
Add a unit test
Test with db_bench setting:
TEST_TMPDIR=/dev/shm/ ./db_bench --benchmarks=fillrandom -num=10000000 --soft_pending_compaction_bytes_limit=1000000000 --hard_pending_compaction_bytes_limit=3000000000 --delayed_write_rate=100000000
and make sure without the commit, write stop will happen, but with the commit, it will not happen.
Reviewers: igor, anthony, rven, yhchiang, kradhakrishnan, IslamAbdelRahman
Reviewed By: IslamAbdelRahman
Subscribers: leveldb, dhruba
Differential Revision: https://reviews.facebook.net/D52131
2015-12-18 01:07:44 +00:00
|
|
|
TEST_F(ColumnFamilyTest, WriteStallTwoColumnFamilies) {
|
|
|
|
const uint64_t kBaseRate = 810000u;
|
|
|
|
db_options_.delayed_write_rate = kBaseRate;
|
|
|
|
Open();
|
|
|
|
CreateColumnFamilies({"one"});
|
|
|
|
ColumnFamilyData* cfd =
|
|
|
|
static_cast<ColumnFamilyHandleImpl*>(db_->DefaultColumnFamily())->cfd();
|
|
|
|
VersionStorageInfo* vstorage = cfd->current()->storage_info();
|
|
|
|
|
|
|
|
ColumnFamilyData* cfd1 =
|
|
|
|
static_cast<ColumnFamilyHandleImpl*>(handles_[1])->cfd();
|
|
|
|
VersionStorageInfo* vstorage1 = cfd1->current()->storage_info();
|
|
|
|
|
|
|
|
MutableCFOptions mutable_cf_options(
|
|
|
|
Options(db_options_, column_family_options_),
|
|
|
|
ImmutableCFOptions(Options(db_options_, column_family_options_)));
|
|
|
|
mutable_cf_options.level0_slowdown_writes_trigger = 20;
|
|
|
|
mutable_cf_options.level0_stop_writes_trigger = 10000;
|
|
|
|
mutable_cf_options.soft_pending_compaction_bytes_limit = 200;
|
|
|
|
mutable_cf_options.hard_pending_compaction_bytes_limit = 2000;
|
|
|
|
|
|
|
|
MutableCFOptions mutable_cf_options1 = mutable_cf_options;
|
|
|
|
mutable_cf_options1.soft_pending_compaction_bytes_limit = 500;
|
|
|
|
|
|
|
|
vstorage->TEST_set_estimated_compaction_needed_bytes(50);
|
|
|
|
cfd->RecalculateWriteStallConditions(mutable_cf_options);
|
|
|
|
ASSERT_TRUE(!dbfull()->TEST_write_controler().IsStopped());
|
|
|
|
ASSERT_TRUE(!dbfull()->TEST_write_controler().NeedsDelay());
|
|
|
|
|
|
|
|
vstorage1->TEST_set_estimated_compaction_needed_bytes(201);
|
|
|
|
cfd->RecalculateWriteStallConditions(mutable_cf_options);
|
|
|
|
ASSERT_TRUE(!dbfull()->TEST_write_controler().IsStopped());
|
|
|
|
ASSERT_TRUE(!dbfull()->TEST_write_controler().NeedsDelay());
|
|
|
|
|
|
|
|
vstorage1->TEST_set_estimated_compaction_needed_bytes(600);
|
|
|
|
cfd1->RecalculateWriteStallConditions(mutable_cf_options);
|
|
|
|
ASSERT_TRUE(!dbfull()->TEST_write_controler().IsStopped());
|
|
|
|
ASSERT_TRUE(dbfull()->TEST_write_controler().NeedsDelay());
|
|
|
|
ASSERT_EQ(kBaseRate, dbfull()->TEST_write_controler().delayed_write_rate());
|
|
|
|
|
|
|
|
vstorage->TEST_set_estimated_compaction_needed_bytes(70);
|
|
|
|
cfd->RecalculateWriteStallConditions(mutable_cf_options);
|
|
|
|
ASSERT_TRUE(!dbfull()->TEST_write_controler().IsStopped());
|
|
|
|
ASSERT_TRUE(dbfull()->TEST_write_controler().NeedsDelay());
|
|
|
|
ASSERT_EQ(kBaseRate, dbfull()->TEST_write_controler().delayed_write_rate());
|
|
|
|
|
|
|
|
vstorage1->TEST_set_estimated_compaction_needed_bytes(800);
|
|
|
|
cfd1->RecalculateWriteStallConditions(mutable_cf_options);
|
|
|
|
ASSERT_TRUE(!dbfull()->TEST_write_controler().IsStopped());
|
|
|
|
ASSERT_TRUE(dbfull()->TEST_write_controler().NeedsDelay());
|
|
|
|
ASSERT_EQ(kBaseRate / 1.2,
|
|
|
|
dbfull()->TEST_write_controler().delayed_write_rate());
|
|
|
|
|
|
|
|
vstorage->TEST_set_estimated_compaction_needed_bytes(300);
|
|
|
|
cfd->RecalculateWriteStallConditions(mutable_cf_options);
|
|
|
|
ASSERT_TRUE(!dbfull()->TEST_write_controler().IsStopped());
|
|
|
|
ASSERT_TRUE(dbfull()->TEST_write_controler().NeedsDelay());
|
|
|
|
ASSERT_EQ(kBaseRate / 1.2 / 1.2,
|
|
|
|
dbfull()->TEST_write_controler().delayed_write_rate());
|
|
|
|
|
|
|
|
vstorage1->TEST_set_estimated_compaction_needed_bytes(700);
|
|
|
|
cfd1->RecalculateWriteStallConditions(mutable_cf_options);
|
|
|
|
ASSERT_TRUE(!dbfull()->TEST_write_controler().IsStopped());
|
|
|
|
ASSERT_TRUE(dbfull()->TEST_write_controler().NeedsDelay());
|
|
|
|
ASSERT_EQ(kBaseRate / 1.2,
|
|
|
|
dbfull()->TEST_write_controler().delayed_write_rate());
|
|
|
|
|
|
|
|
vstorage->TEST_set_estimated_compaction_needed_bytes(500);
|
|
|
|
cfd->RecalculateWriteStallConditions(mutable_cf_options);
|
|
|
|
ASSERT_TRUE(!dbfull()->TEST_write_controler().IsStopped());
|
|
|
|
ASSERT_TRUE(dbfull()->TEST_write_controler().NeedsDelay());
|
|
|
|
ASSERT_EQ(kBaseRate / 1.2 / 1.2,
|
|
|
|
dbfull()->TEST_write_controler().delayed_write_rate());
|
|
|
|
|
|
|
|
vstorage1->TEST_set_estimated_compaction_needed_bytes(600);
|
|
|
|
cfd1->RecalculateWriteStallConditions(mutable_cf_options);
|
|
|
|
ASSERT_TRUE(!dbfull()->TEST_write_controler().IsStopped());
|
|
|
|
ASSERT_TRUE(dbfull()->TEST_write_controler().NeedsDelay());
|
|
|
|
ASSERT_EQ(kBaseRate / 1.2,
|
|
|
|
dbfull()->TEST_write_controler().delayed_write_rate());
|
|
|
|
}
|
Add options.base_background_compactions as a number of compaction threads for low compaction debt
Summary:
If options.base_background_compactions is given, we try to schedule number of compactions not existing this number, only when L0 files increase to certain number, or pending compaction bytes more than certain threshold, we schedule compactions based on options.max_background_compactions.
The watermarks are calculated based on slowdown thresholds.
Test Plan:
Add new test cases in column_family_test.
Adding more unit tests.
Reviewers: IslamAbdelRahman, yhchiang, kradhakrishnan, rven, anthony
Reviewed By: anthony
Subscribers: leveldb, dhruba, yoshinorim
Differential Revision: https://reviews.facebook.net/D53409
2016-01-28 19:56:16 +00:00
|
|
|
|
|
|
|
TEST_F(ColumnFamilyTest, CompactionSpeedupTwoColumnFamilies) {
|
|
|
|
db_options_.base_background_compactions = 2;
|
|
|
|
db_options_.max_background_compactions = 6;
|
|
|
|
column_family_options_.soft_pending_compaction_bytes_limit = 200;
|
|
|
|
column_family_options_.hard_pending_compaction_bytes_limit = 2000;
|
|
|
|
Open();
|
|
|
|
CreateColumnFamilies({"one"});
|
|
|
|
ColumnFamilyData* cfd =
|
|
|
|
static_cast<ColumnFamilyHandleImpl*>(db_->DefaultColumnFamily())->cfd();
|
|
|
|
VersionStorageInfo* vstorage = cfd->current()->storage_info();
|
|
|
|
|
|
|
|
ColumnFamilyData* cfd1 =
|
|
|
|
static_cast<ColumnFamilyHandleImpl*>(handles_[1])->cfd();
|
|
|
|
VersionStorageInfo* vstorage1 = cfd1->current()->storage_info();
|
|
|
|
|
|
|
|
MutableCFOptions mutable_cf_options(
|
|
|
|
Options(db_options_, column_family_options_),
|
|
|
|
ImmutableCFOptions(Options(db_options_, column_family_options_)));
|
|
|
|
// Speed up threshold = min(4 * 2, 4 + (36 - 4)/4) = 8
|
|
|
|
mutable_cf_options.level0_file_num_compaction_trigger = 4;
|
|
|
|
mutable_cf_options.level0_slowdown_writes_trigger = 36;
|
|
|
|
mutable_cf_options.level0_stop_writes_trigger = 30;
|
|
|
|
// Speedup threshold = 200 / 4 = 50
|
|
|
|
mutable_cf_options.soft_pending_compaction_bytes_limit = 200;
|
|
|
|
mutable_cf_options.hard_pending_compaction_bytes_limit = 2000;
|
|
|
|
|
|
|
|
MutableCFOptions mutable_cf_options1 = mutable_cf_options;
|
|
|
|
mutable_cf_options1.level0_slowdown_writes_trigger = 16;
|
|
|
|
|
|
|
|
vstorage->TEST_set_estimated_compaction_needed_bytes(40);
|
|
|
|
cfd->RecalculateWriteStallConditions(mutable_cf_options);
|
|
|
|
ASSERT_EQ(2, dbfull()->BGCompactionsAllowed());
|
|
|
|
|
|
|
|
vstorage->TEST_set_estimated_compaction_needed_bytes(60);
|
|
|
|
cfd1->RecalculateWriteStallConditions(mutable_cf_options);
|
|
|
|
ASSERT_EQ(2, dbfull()->BGCompactionsAllowed());
|
|
|
|
cfd->RecalculateWriteStallConditions(mutable_cf_options);
|
|
|
|
ASSERT_EQ(6, dbfull()->BGCompactionsAllowed());
|
|
|
|
|
|
|
|
vstorage1->TEST_set_estimated_compaction_needed_bytes(30);
|
|
|
|
cfd1->RecalculateWriteStallConditions(mutable_cf_options);
|
|
|
|
ASSERT_EQ(6, dbfull()->BGCompactionsAllowed());
|
|
|
|
|
|
|
|
vstorage1->TEST_set_estimated_compaction_needed_bytes(70);
|
|
|
|
cfd1->RecalculateWriteStallConditions(mutable_cf_options);
|
|
|
|
ASSERT_EQ(6, dbfull()->BGCompactionsAllowed());
|
|
|
|
|
|
|
|
vstorage->TEST_set_estimated_compaction_needed_bytes(20);
|
|
|
|
cfd->RecalculateWriteStallConditions(mutable_cf_options);
|
|
|
|
ASSERT_EQ(6, dbfull()->BGCompactionsAllowed());
|
|
|
|
|
|
|
|
vstorage1->TEST_set_estimated_compaction_needed_bytes(3);
|
|
|
|
cfd1->RecalculateWriteStallConditions(mutable_cf_options);
|
|
|
|
ASSERT_EQ(2, dbfull()->BGCompactionsAllowed());
|
|
|
|
|
|
|
|
vstorage->set_l0_delay_trigger_count(9);
|
|
|
|
cfd->RecalculateWriteStallConditions(mutable_cf_options);
|
|
|
|
ASSERT_EQ(6, dbfull()->BGCompactionsAllowed());
|
|
|
|
|
|
|
|
vstorage1->set_l0_delay_trigger_count(2);
|
|
|
|
cfd1->RecalculateWriteStallConditions(mutable_cf_options);
|
|
|
|
ASSERT_EQ(6, dbfull()->BGCompactionsAllowed());
|
|
|
|
|
|
|
|
vstorage->set_l0_delay_trigger_count(0);
|
|
|
|
cfd->RecalculateWriteStallConditions(mutable_cf_options);
|
|
|
|
ASSERT_EQ(2, dbfull()->BGCompactionsAllowed());
|
|
|
|
}
|
2016-02-23 01:29:28 +00:00
|
|
|
|
2016-08-04 00:42:06 +00:00
|
|
|
#ifndef ROCKSDB_LITE
|
|
|
|
TEST_F(ColumnFamilyTest, FlushCloseWALFiles) {
|
|
|
|
SpecialEnv env(Env::Default());
|
|
|
|
db_options_.env = &env;
|
|
|
|
db_options_.max_background_flushes = 1;
|
|
|
|
column_family_options_.memtable_factory.reset(new SpecialSkipListFactory(2));
|
|
|
|
Open();
|
|
|
|
CreateColumnFamilies({"one"});
|
|
|
|
ASSERT_OK(Put(1, "fodor", "mirko"));
|
|
|
|
ASSERT_OK(Put(0, "fodor", "mirko"));
|
|
|
|
ASSERT_OK(Put(1, "fodor", "mirko"));
|
|
|
|
|
|
|
|
// Block flush jobs from running
|
|
|
|
test::SleepingBackgroundTask sleeping_task;
|
|
|
|
env_->Schedule(&test::SleepingBackgroundTask::DoSleepTask, &sleeping_task,
|
|
|
|
Env::Priority::HIGH);
|
|
|
|
|
|
|
|
WriteOptions wo;
|
|
|
|
wo.sync = true;
|
|
|
|
ASSERT_OK(db_->Put(wo, handles_[1], "fodor", "mirko"));
|
|
|
|
|
|
|
|
ASSERT_EQ(2, env.num_open_wal_file_.load());
|
|
|
|
|
|
|
|
sleeping_task.WakeUp();
|
|
|
|
sleeping_task.WaitUntilDone();
|
|
|
|
WaitForFlush(1);
|
|
|
|
ASSERT_EQ(1, env.num_open_wal_file_.load());
|
|
|
|
|
|
|
|
Reopen();
|
|
|
|
ASSERT_EQ("mirko", Get(0, "fodor"));
|
|
|
|
ASSERT_EQ("mirko", Get(1, "fodor"));
|
|
|
|
db_options_.env = env_;
|
|
|
|
Close();
|
|
|
|
}
|
|
|
|
#endif // !ROCKSDB_LITE
|
|
|
|
|
|
|
|
TEST_F(ColumnFamilyTest, IteratorCloseWALFile1) {
|
|
|
|
SpecialEnv env(Env::Default());
|
|
|
|
db_options_.env = &env;
|
|
|
|
db_options_.max_background_flushes = 1;
|
|
|
|
column_family_options_.memtable_factory.reset(new SpecialSkipListFactory(2));
|
|
|
|
Open();
|
|
|
|
CreateColumnFamilies({"one"});
|
|
|
|
ASSERT_OK(Put(1, "fodor", "mirko"));
|
|
|
|
// Create an iterator holding the current super version.
|
|
|
|
Iterator* it = db_->NewIterator(ReadOptions(), handles_[1]);
|
|
|
|
// A flush will make `it` hold the last reference of its super version.
|
|
|
|
Flush(1);
|
|
|
|
|
|
|
|
ASSERT_OK(Put(1, "fodor", "mirko"));
|
|
|
|
ASSERT_OK(Put(0, "fodor", "mirko"));
|
|
|
|
ASSERT_OK(Put(1, "fodor", "mirko"));
|
|
|
|
|
|
|
|
// Flush jobs will close previous WAL files after finishing. By
|
|
|
|
// block flush jobs from running, we trigger a condition where
|
|
|
|
// the iterator destructor should close the WAL files.
|
|
|
|
test::SleepingBackgroundTask sleeping_task;
|
|
|
|
env_->Schedule(&test::SleepingBackgroundTask::DoSleepTask, &sleeping_task,
|
|
|
|
Env::Priority::HIGH);
|
|
|
|
|
|
|
|
WriteOptions wo;
|
|
|
|
wo.sync = true;
|
|
|
|
ASSERT_OK(db_->Put(wo, handles_[1], "fodor", "mirko"));
|
|
|
|
|
|
|
|
ASSERT_EQ(2, env.num_open_wal_file_.load());
|
|
|
|
// Deleting the iterator will clear its super version, triggering
|
|
|
|
// closing all files
|
|
|
|
delete it;
|
|
|
|
ASSERT_EQ(1, env.num_open_wal_file_.load());
|
|
|
|
|
|
|
|
sleeping_task.WakeUp();
|
|
|
|
sleeping_task.WaitUntilDone();
|
|
|
|
WaitForFlush(1);
|
|
|
|
|
|
|
|
Reopen();
|
|
|
|
ASSERT_EQ("mirko", Get(0, "fodor"));
|
|
|
|
ASSERT_EQ("mirko", Get(1, "fodor"));
|
|
|
|
db_options_.env = env_;
|
|
|
|
Close();
|
|
|
|
}
|
|
|
|
|
|
|
|
TEST_F(ColumnFamilyTest, IteratorCloseWALFile2) {
|
|
|
|
SpecialEnv env(Env::Default());
|
|
|
|
// Allow both of flush and purge job to schedule.
|
|
|
|
env.SetBackgroundThreads(2, Env::HIGH);
|
|
|
|
db_options_.env = &env;
|
|
|
|
db_options_.max_background_flushes = 1;
|
|
|
|
column_family_options_.memtable_factory.reset(new SpecialSkipListFactory(2));
|
|
|
|
Open();
|
|
|
|
CreateColumnFamilies({"one"});
|
|
|
|
ASSERT_OK(Put(1, "fodor", "mirko"));
|
|
|
|
// Create an iterator holding the current super version.
|
|
|
|
ReadOptions ro;
|
|
|
|
ro.background_purge_on_iterator_cleanup = true;
|
|
|
|
Iterator* it = db_->NewIterator(ro, handles_[1]);
|
|
|
|
// A flush will make `it` hold the last reference of its super version.
|
|
|
|
Flush(1);
|
|
|
|
|
|
|
|
ASSERT_OK(Put(1, "fodor", "mirko"));
|
|
|
|
ASSERT_OK(Put(0, "fodor", "mirko"));
|
|
|
|
ASSERT_OK(Put(1, "fodor", "mirko"));
|
|
|
|
|
|
|
|
rocksdb::SyncPoint::GetInstance()->LoadDependency({
|
|
|
|
{"ColumnFamilyTest::IteratorCloseWALFile2:0",
|
|
|
|
"DBImpl::BGWorkPurge:start"},
|
|
|
|
{"ColumnFamilyTest::IteratorCloseWALFile2:2",
|
|
|
|
"DBImpl::BackgroundCallFlush:start"},
|
|
|
|
{"DBImpl::BGWorkPurge:end", "ColumnFamilyTest::IteratorCloseWALFile2:1"},
|
|
|
|
});
|
|
|
|
rocksdb::SyncPoint::GetInstance()->EnableProcessing();
|
|
|
|
|
|
|
|
WriteOptions wo;
|
|
|
|
wo.sync = true;
|
|
|
|
ASSERT_OK(db_->Put(wo, handles_[1], "fodor", "mirko"));
|
|
|
|
|
|
|
|
ASSERT_EQ(2, env.num_open_wal_file_.load());
|
|
|
|
// Deleting the iterator will clear its super version, triggering
|
|
|
|
// closing all files
|
|
|
|
delete it;
|
|
|
|
ASSERT_EQ(2, env.num_open_wal_file_.load());
|
|
|
|
|
|
|
|
TEST_SYNC_POINT("ColumnFamilyTest::IteratorCloseWALFile2:0");
|
|
|
|
TEST_SYNC_POINT("ColumnFamilyTest::IteratorCloseWALFile2:1");
|
|
|
|
ASSERT_EQ(1, env.num_open_wal_file_.load());
|
|
|
|
TEST_SYNC_POINT("ColumnFamilyTest::IteratorCloseWALFile2:2");
|
|
|
|
WaitForFlush(1);
|
|
|
|
ASSERT_EQ(1, env.num_open_wal_file_.load());
|
|
|
|
rocksdb::SyncPoint::GetInstance()->DisableProcessing();
|
|
|
|
|
|
|
|
Reopen();
|
|
|
|
ASSERT_EQ("mirko", Get(0, "fodor"));
|
|
|
|
ASSERT_EQ("mirko", Get(1, "fodor"));
|
|
|
|
db_options_.env = env_;
|
|
|
|
Close();
|
|
|
|
}
|
|
|
|
|
|
|
|
#ifndef ROCKSDB_LITE // TEST functions are not supported in lite
|
|
|
|
TEST_F(ColumnFamilyTest, ForwardIteratorCloseWALFile) {
|
|
|
|
SpecialEnv env(Env::Default());
|
|
|
|
// Allow both of flush and purge job to schedule.
|
|
|
|
env.SetBackgroundThreads(2, Env::HIGH);
|
|
|
|
db_options_.env = &env;
|
|
|
|
db_options_.max_background_flushes = 1;
|
|
|
|
column_family_options_.memtable_factory.reset(new SpecialSkipListFactory(3));
|
|
|
|
column_family_options_.level0_file_num_compaction_trigger = 2;
|
|
|
|
Open();
|
|
|
|
CreateColumnFamilies({"one"});
|
|
|
|
ASSERT_OK(Put(1, "fodor", "mirko"));
|
|
|
|
ASSERT_OK(Put(1, "fodar2", "mirko"));
|
|
|
|
Flush(1);
|
|
|
|
|
|
|
|
// Create an iterator holding the current super version, as well as
|
|
|
|
// the SST file just flushed.
|
|
|
|
ReadOptions ro;
|
|
|
|
ro.tailing = true;
|
|
|
|
ro.background_purge_on_iterator_cleanup = true;
|
|
|
|
Iterator* it = db_->NewIterator(ro, handles_[1]);
|
|
|
|
// A flush will make `it` hold the last reference of its super version.
|
|
|
|
|
|
|
|
ASSERT_OK(Put(1, "fodor", "mirko"));
|
|
|
|
ASSERT_OK(Put(1, "fodar2", "mirko"));
|
|
|
|
Flush(1);
|
|
|
|
|
|
|
|
WaitForCompaction();
|
|
|
|
|
|
|
|
ASSERT_OK(Put(1, "fodor", "mirko"));
|
|
|
|
ASSERT_OK(Put(1, "fodor", "mirko"));
|
|
|
|
ASSERT_OK(Put(0, "fodor", "mirko"));
|
|
|
|
ASSERT_OK(Put(1, "fodor", "mirko"));
|
|
|
|
|
|
|
|
rocksdb::SyncPoint::GetInstance()->LoadDependency({
|
|
|
|
{"ColumnFamilyTest::IteratorCloseWALFile2:0",
|
|
|
|
"DBImpl::BGWorkPurge:start"},
|
|
|
|
{"ColumnFamilyTest::IteratorCloseWALFile2:2",
|
|
|
|
"DBImpl::BackgroundCallFlush:start"},
|
|
|
|
{"DBImpl::BGWorkPurge:end", "ColumnFamilyTest::IteratorCloseWALFile2:1"},
|
|
|
|
});
|
|
|
|
rocksdb::SyncPoint::GetInstance()->EnableProcessing();
|
|
|
|
|
|
|
|
WriteOptions wo;
|
|
|
|
wo.sync = true;
|
|
|
|
ASSERT_OK(db_->Put(wo, handles_[1], "fodor", "mirko"));
|
|
|
|
|
|
|
|
env.delete_count_.store(0);
|
|
|
|
ASSERT_EQ(2, env.num_open_wal_file_.load());
|
|
|
|
// Deleting the iterator will clear its super version, triggering
|
|
|
|
// closing all files
|
|
|
|
it->Seek("");
|
|
|
|
ASSERT_EQ(2, env.num_open_wal_file_.load());
|
|
|
|
ASSERT_EQ(0, env.delete_count_.load());
|
|
|
|
|
|
|
|
TEST_SYNC_POINT("ColumnFamilyTest::IteratorCloseWALFile2:0");
|
|
|
|
TEST_SYNC_POINT("ColumnFamilyTest::IteratorCloseWALFile2:1");
|
|
|
|
ASSERT_EQ(1, env.num_open_wal_file_.load());
|
|
|
|
ASSERT_EQ(1, env.delete_count_.load());
|
|
|
|
TEST_SYNC_POINT("ColumnFamilyTest::IteratorCloseWALFile2:2");
|
|
|
|
WaitForFlush(1);
|
|
|
|
ASSERT_EQ(1, env.num_open_wal_file_.load());
|
|
|
|
ASSERT_EQ(1, env.delete_count_.load());
|
|
|
|
|
|
|
|
delete it;
|
|
|
|
rocksdb::SyncPoint::GetInstance()->DisableProcessing();
|
|
|
|
|
|
|
|
Reopen();
|
|
|
|
ASSERT_EQ("mirko", Get(0, "fodor"));
|
|
|
|
ASSERT_EQ("mirko", Get(1, "fodor"));
|
|
|
|
db_options_.env = env_;
|
|
|
|
Close();
|
|
|
|
}
|
|
|
|
#endif // !ROCKSDB_LITE
|
|
|
|
|
2016-04-15 00:33:53 +00:00
|
|
|
// Disable on windows because SyncWAL requires env->IsSyncThreadSafe()
|
|
|
|
// to return true which is not so in unbuffered mode.
|
|
|
|
#ifndef OS_WIN
|
2016-02-23 01:29:28 +00:00
|
|
|
TEST_F(ColumnFamilyTest, LogSyncConflictFlush) {
|
|
|
|
Open();
|
|
|
|
CreateColumnFamiliesAndReopen({"one", "two"});
|
|
|
|
|
|
|
|
Put(0, "", "");
|
|
|
|
Put(1, "foo", "bar");
|
|
|
|
|
|
|
|
rocksdb::SyncPoint::GetInstance()->LoadDependency(
|
|
|
|
{{"DBImpl::SyncWAL:BeforeMarkLogsSynced:1",
|
|
|
|
"ColumnFamilyTest::LogSyncConflictFlush:1"},
|
|
|
|
{"ColumnFamilyTest::LogSyncConflictFlush:2",
|
|
|
|
"DBImpl::SyncWAL:BeforeMarkLogsSynced:2"}});
|
|
|
|
|
|
|
|
rocksdb::SyncPoint::GetInstance()->EnableProcessing();
|
|
|
|
|
|
|
|
std::thread thread([&] { db_->SyncWAL(); });
|
|
|
|
|
|
|
|
TEST_SYNC_POINT("ColumnFamilyTest::LogSyncConflictFlush:1");
|
|
|
|
Flush(1);
|
|
|
|
Put(1, "foo", "bar");
|
|
|
|
Flush(1);
|
|
|
|
|
|
|
|
TEST_SYNC_POINT("ColumnFamilyTest::LogSyncConflictFlush:2");
|
|
|
|
|
|
|
|
thread.join();
|
|
|
|
|
|
|
|
rocksdb::SyncPoint::GetInstance()->DisableProcessing();
|
|
|
|
Close();
|
|
|
|
}
|
2016-04-15 00:33:53 +00:00
|
|
|
#endif
|
2016-06-06 21:40:36 +00:00
|
|
|
|
|
|
|
// this test is placed here, because the infrastructure for Column Family
|
|
|
|
// test is being used to ensure a roll of wal files.
|
|
|
|
// Basic idea is to test that WAL truncation is being detected and not
|
|
|
|
// ignored
|
|
|
|
TEST_F(ColumnFamilyTest, DISABLED_LogTruncationTest) {
|
|
|
|
Open();
|
|
|
|
CreateColumnFamiliesAndReopen({"one", "two"});
|
|
|
|
|
|
|
|
Build(0, 100);
|
|
|
|
|
|
|
|
// Flush the 0th column family to force a roll of the wal log
|
|
|
|
Flush(0);
|
|
|
|
|
|
|
|
// Add some more entries
|
|
|
|
Build(100, 100);
|
|
|
|
|
|
|
|
std::vector<std::string> filenames;
|
|
|
|
ASSERT_OK(env_->GetChildren(dbname_, &filenames));
|
|
|
|
|
|
|
|
// collect wal files
|
|
|
|
std::vector<std::string> logfs;
|
|
|
|
for (size_t i = 0; i < filenames.size(); i++) {
|
|
|
|
uint64_t number;
|
|
|
|
FileType type;
|
|
|
|
if (!(ParseFileName(filenames[i], &number, &type))) continue;
|
|
|
|
|
|
|
|
if (type != kLogFile) continue;
|
|
|
|
|
|
|
|
logfs.push_back(filenames[i]);
|
|
|
|
}
|
|
|
|
|
|
|
|
std::sort(logfs.begin(), logfs.end());
|
|
|
|
ASSERT_GE(logfs.size(), 2);
|
|
|
|
|
|
|
|
// Take the last but one file, and truncate it
|
|
|
|
std::string fpath = dbname_ + "/" + logfs[logfs.size() - 2];
|
|
|
|
std::vector<std::string> names_save = names_;
|
|
|
|
|
|
|
|
uint64_t fsize;
|
|
|
|
ASSERT_OK(env_->GetFileSize(fpath, &fsize));
|
|
|
|
ASSERT_GT(fsize, 0);
|
|
|
|
|
|
|
|
Close();
|
|
|
|
|
|
|
|
std::string backup_logs = dbname_ + "/backup_logs";
|
|
|
|
std::string t_fpath = backup_logs + "/" + logfs[logfs.size() - 2];
|
|
|
|
|
|
|
|
ASSERT_OK(env_->CreateDirIfMissing(backup_logs));
|
|
|
|
// Not sure how easy it is to make this data driven.
|
|
|
|
// need to read back the WAL file and truncate last 10
|
|
|
|
// entries
|
|
|
|
CopyFile(fpath, t_fpath, fsize - 9180);
|
|
|
|
|
|
|
|
ASSERT_OK(env_->DeleteFile(fpath));
|
|
|
|
ASSERT_OK(env_->RenameFile(t_fpath, fpath));
|
|
|
|
|
|
|
|
db_options_.wal_recovery_mode = WALRecoveryMode::kPointInTimeRecovery;
|
|
|
|
|
|
|
|
OpenReadOnly(names_save);
|
|
|
|
|
|
|
|
CheckMissed();
|
|
|
|
|
|
|
|
Close();
|
|
|
|
|
|
|
|
Open(names_save);
|
|
|
|
|
|
|
|
CheckMissed();
|
|
|
|
|
|
|
|
Close();
|
|
|
|
|
|
|
|
// cleanup
|
|
|
|
env_->DeleteDir(backup_logs);
|
|
|
|
}
|
2014-01-02 17:08:12 +00:00
|
|
|
} // namespace rocksdb
|
|
|
|
|
|
|
|
int main(int argc, char** argv) {
|
2015-03-17 21:08:00 +00:00
|
|
|
::testing::InitGoogleTest(&argc, argv);
|
|
|
|
return RUN_ALL_TESTS();
|
2014-01-02 17:08:12 +00:00
|
|
|
}
|