You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
rocksdb/db/listener_test.cc

1596 lines
52 KiB

// Copyright (c) 2011-present, Facebook, Inc. All rights reserved.
// This source code is licensed under both the GPLv2 (found in the
// COPYING file in the root directory) and Apache 2.0 License
// (found in the LICENSE.Apache file in the root directory).
#include "db/blob/blob_index.h"
#include "db/db_impl/db_impl.h"
#include "db/db_test_util.h"
#include "db/dbformat.h"
#include "db/version_set.h"
#include "db/write_batch_internal.h"
#include "file/filename.h"
#include "monitoring/statistics.h"
#include "rocksdb/cache.h"
#include "rocksdb/compaction_filter.h"
#include "rocksdb/db.h"
#include "rocksdb/env.h"
#include "rocksdb/filter_policy.h"
#include "rocksdb/options.h"
#include "rocksdb/perf_context.h"
#include "rocksdb/slice.h"
#include "rocksdb/slice_transform.h"
#include "rocksdb/table.h"
#include "rocksdb/table_properties.h"
#include "test_util/sync_point.h"
#include "test_util/testharness.h"
#include "test_util/testutil.h"
#include "util/hash.h"
#include "util/mutexlock.h"
#include "util/rate_limiter.h"
#include "util/string_util.h"
#include "utilities/merge_operators.h"
#ifndef ROCKSDB_LITE
namespace ROCKSDB_NAMESPACE {
class EventListenerTest : public DBTestBase {
public:
EventListenerTest() : DBTestBase("listener_test", /*env_do_fsync=*/true) {}
static std::string BlobStr(uint64_t blob_file_number, uint64_t offset,
uint64_t size) {
std::string blob_index;
BlobIndex::EncodeBlob(&blob_index, blob_file_number, offset, size,
kNoCompression);
return blob_index;
}
const size_t k110KB = 110 << 10;
};
struct TestPropertiesCollector
: public ROCKSDB_NAMESPACE::TablePropertiesCollector {
ROCKSDB_NAMESPACE::Status AddUserKey(
const ROCKSDB_NAMESPACE::Slice& /*key*/,
const ROCKSDB_NAMESPACE::Slice& /*value*/,
ROCKSDB_NAMESPACE::EntryType /*type*/,
ROCKSDB_NAMESPACE::SequenceNumber /*seq*/,
uint64_t /*file_size*/) override {
return Status::OK();
}
ROCKSDB_NAMESPACE::Status Finish(
ROCKSDB_NAMESPACE::UserCollectedProperties* properties) override {
properties->insert({"0", "1"});
return Status::OK();
}
const char* Name() const override { return "TestTablePropertiesCollector"; }
ROCKSDB_NAMESPACE::UserCollectedProperties GetReadableProperties()
const override {
ROCKSDB_NAMESPACE::UserCollectedProperties ret;
ret["2"] = "3";
return ret;
}
};
class TestPropertiesCollectorFactory : public TablePropertiesCollectorFactory {
public:
TablePropertiesCollector* CreateTablePropertiesCollector(
TablePropertiesCollectorFactory::Context /*context*/) override {
return new TestPropertiesCollector;
}
const char* Name() const override { return "TestTablePropertiesCollector"; }
};
class TestCompactionListener : public EventListener {
public:
explicit TestCompactionListener(EventListenerTest* test) : test_(test) {}
void OnCompactionCompleted(DB *db, const CompactionJobInfo& ci) override {
std::lock_guard<std::mutex> lock(mutex_);
compacted_dbs_.push_back(db);
ASSERT_GT(ci.input_files.size(), 0U);
ASSERT_EQ(ci.input_files.size(), ci.input_file_infos.size());
for (size_t i = 0; i < ci.input_file_infos.size(); ++i) {
ASSERT_EQ(ci.input_file_infos[i].level, ci.base_input_level);
ASSERT_EQ(ci.input_file_infos[i].file_number,
TableFileNameToNumber(ci.input_files[i]));
}
ASSERT_GT(ci.output_files.size(), 0U);
ASSERT_EQ(ci.output_files.size(), ci.output_file_infos.size());
ASSERT_TRUE(test_);
ASSERT_EQ(test_->db_, db);
std::vector<std::vector<FileMetaData>> files_by_level;
test_->dbfull()->TEST_GetFilesMetaData(test_->handles_[ci.cf_id],
&files_by_level);
ASSERT_GT(files_by_level.size(), ci.output_level);
for (size_t i = 0; i < ci.output_file_infos.size(); ++i) {
ASSERT_EQ(ci.output_file_infos[i].level, ci.output_level);
ASSERT_EQ(ci.output_file_infos[i].file_number,
TableFileNameToNumber(ci.output_files[i]));
auto it = std::find_if(
files_by_level[ci.output_level].begin(),
files_by_level[ci.output_level].end(), [&](const FileMetaData& meta) {
return meta.fd.GetNumber() == ci.output_file_infos[i].file_number;
});
ASSERT_NE(it, files_by_level[ci.output_level].end());
ASSERT_EQ(ci.output_file_infos[i].oldest_blob_file_number,
it->oldest_blob_file_number);
}
ASSERT_EQ(db->GetEnv()->GetThreadID(), ci.thread_id);
ASSERT_GT(ci.thread_id, 0U);
for (auto fl : {ci.input_files, ci.output_files}) {
for (auto fn : fl) {
auto it = ci.table_properties.find(fn);
ASSERT_NE(it, ci.table_properties.end());
auto tp = it->second;
ASSERT_TRUE(tp != nullptr);
ASSERT_EQ(tp->user_collected_properties.find("0")->second, "1");
}
}
}
EventListenerTest* test_;
std::vector<DB*> compacted_dbs_;
std::mutex mutex_;
};
rocksdb: switch to gtest Summary: Our existing test notation is very similar to what is used in gtest. It makes it easy to adopt what is different. In this diff I modify existing [[ https://code.google.com/p/googletest/wiki/Primer#Test_Fixtures:_Using_the_Same_Data_Configuration_for_Multiple_Te | test fixture ]] classes to inherit from `testing::Test`. Also for unit tests that use fixture class, `TEST` is replaced with `TEST_F` as required in gtest. There are several custom `main` functions in our existing tests. To make this transition easier, I modify all `main` functions to fallow gtest notation. But eventually we can remove them and use implementation of `main` that gtest provides. ```lang=bash % cat ~/transform #!/bin/sh files=$(git ls-files '*test\.cc') for file in $files do if grep -q "rocksdb::test::RunAllTests()" $file then if grep -Eq '^class \w+Test {' $file then perl -pi -e 's/^(class \w+Test) {/${1}: public testing::Test {/g' $file perl -pi -e 's/^(TEST)/${1}_F/g' $file fi perl -pi -e 's/(int main.*\{)/${1}::testing::InitGoogleTest(&argc, argv);/g' $file perl -pi -e 's/rocksdb::test::RunAllTests/RUN_ALL_TESTS/g' $file fi done % sh ~/transform % make format ``` Second iteration of this diff contains only scripted changes. Third iteration contains manual changes to fix last errors and make it compilable. Test Plan: Build and notice no errors. ```lang=bash % USE_CLANG=1 make check -j55 ``` Tests are still testing. Reviewers: meyering, sdong, rven, igor Reviewed By: igor Subscribers: dhruba, leveldb Differential Revision: https://reviews.facebook.net/D35157
10 years ago
TEST_F(EventListenerTest, OnSingleDBCompactionTest) {
const int kTestKeySize = 16;
const int kTestValueSize = 984;
const int kEntrySize = kTestKeySize + kTestValueSize;
const int kEntriesPerBuffer = 100;
const int kNumL0Files = 4;
Options options;
Encryption at rest support Summary: This PR adds support for encrypting data stored by RocksDB when written to disk. It adds an `EncryptedEnv` override of the `Env` class with matching overrides for sequential&random access files. The encryption itself is done through a configurable `EncryptionProvider`. This class creates is asked to create `BlockAccessCipherStream` for a file. This is where the actual encryption/decryption is being done. Currently there is a Counter mode implementation of `BlockAccessCipherStream` with a `ROT13` block cipher (NOTE the `ROT13` is for demo purposes only!!). The Counter operation mode uses an initial counter & random initialization vector (IV). Both are created randomly for each file and stored in a 4K (default size) block that is prefixed to that file. The `EncryptedEnv` implementation is such that clients of the `Env` class do not see this prefix (nor data, nor in filesize). The largest part of the prefix block is also encrypted, and there is room left for implementation specific settings/values/keys in there. To test the encryption, the `DBTestBase` class has been extended to consider a new environment variable called `ENCRYPTED_ENV`. If set, the test will setup a encrypted instance of the `Env` class to use for all tests. Typically you would run it like this: ``` ENCRYPTED_ENV=1 make check_some ``` There is also an added test that checks that some data inserted into the database is or is not "visible" on disk. With `ENCRYPTED_ENV` active it must not find plain text strings, with `ENCRYPTED_ENV` unset, it must find the plain text strings. Closes https://github.com/facebook/rocksdb/pull/2424 Differential Revision: D5322178 Pulled By: sdwilsh fbshipit-source-id: 253b0a9c2c498cc98f580df7f2623cbf7678a27f
7 years ago
options.env = CurrentOptions().env;
options.create_if_missing = true;
options.write_buffer_size = kEntrySize * kEntriesPerBuffer;
options.compaction_style = kCompactionStyleLevel;
options.target_file_size_base = options.write_buffer_size;
options.max_bytes_for_level_base = options.target_file_size_base * 2;
options.max_bytes_for_level_multiplier = 2;
options.compression = kNoCompression;
#ifdef ROCKSDB_USING_THREAD_STATUS
options.enable_thread_tracking = true;
#endif // ROCKSDB_USING_THREAD_STATUS
options.level0_file_num_compaction_trigger = kNumL0Files;
options.table_properties_collector_factories.push_back(
std::make_shared<TestPropertiesCollectorFactory>());
TestCompactionListener* listener = new TestCompactionListener(this);
options.listeners.emplace_back(listener);
std::vector<std::string> cf_names = {
"pikachu", "ilya", "muromec", "dobrynia",
"nikitich", "alyosha", "popovich"};
CreateAndReopenWithCF(cf_names, options);
ASSERT_OK(Put(1, "pikachu", std::string(90000, 'p')));
WriteBatch batch;
ASSERT_OK(WriteBatchInternal::PutBlobIndex(&batch, 1, "ditto",
BlobStr(123, 0, 1 << 10)));
ASSERT_OK(dbfull()->Write(WriteOptions(), &batch));
ASSERT_OK(Put(2, "ilya", std::string(90000, 'i')));
ASSERT_OK(Put(3, "muromec", std::string(90000, 'm')));
ASSERT_OK(Put(4, "dobrynia", std::string(90000, 'd')));
ASSERT_OK(Put(5, "nikitich", std::string(90000, 'n')));
ASSERT_OK(Put(6, "alyosha", std::string(90000, 'a')));
ASSERT_OK(Put(7, "popovich", std::string(90000, 'p')));
for (int i = 1; i < 8; ++i) {
ASSERT_OK(Flush(i));
ASSERT_OK(dbfull()->TEST_WaitForFlushMemTable());
ASSERT_OK(dbfull()->CompactRange(CompactRangeOptions(), handles_[i],
nullptr, nullptr));
ASSERT_OK(dbfull()->TEST_WaitForCompact());
}
ASSERT_EQ(listener->compacted_dbs_.size(), cf_names.size());
for (size_t i = 0; i < cf_names.size(); ++i) {
ASSERT_EQ(listener->compacted_dbs_[i], db_);
}
}
// This simple Listener can only handle one flush at a time.
class TestFlushListener : public EventListener {
public:
TestFlushListener(Env* env, EventListenerTest* test)
: slowdown_count(0), stop_count(0), db_closed(), env_(env), test_(test) {
db_closed = false;
}
virtual ~TestFlushListener() {
prev_fc_info_.status.PermitUncheckedError(); // Ignore the status
}
void OnTableFileCreated(
const TableFileCreationInfo& info) override {
// remember the info for later checking the FlushJobInfo.
prev_fc_info_ = info;
ASSERT_GT(info.db_name.size(), 0U);
ASSERT_GT(info.cf_name.size(), 0U);
ASSERT_GT(info.file_path.size(), 0U);
ASSERT_GT(info.job_id, 0);
ASSERT_GT(info.table_properties.data_size, 0U);
ASSERT_GT(info.table_properties.raw_key_size, 0U);
ASSERT_GT(info.table_properties.raw_value_size, 0U);
ASSERT_GT(info.table_properties.num_data_blocks, 0U);
ASSERT_GT(info.table_properties.num_entries, 0U);
ASSERT_EQ(info.file_checksum, kUnknownFileChecksum);
ASSERT_EQ(info.file_checksum_func_name, kUnknownFileChecksumFuncName);
#ifdef ROCKSDB_USING_THREAD_STATUS
// Verify the id of the current thread that created this table
// file matches the id of any active flush or compaction thread.
uint64_t thread_id = env_->GetThreadID();
std::vector<ThreadStatus> thread_list;
ASSERT_OK(env_->GetThreadList(&thread_list));
bool found_match = false;
for (auto thread_status : thread_list) {
if (thread_status.operation_type == ThreadStatus::OP_FLUSH ||
thread_status.operation_type == ThreadStatus::OP_COMPACTION) {
if (thread_id == thread_status.thread_id) {
found_match = true;
break;
}
}
}
ASSERT_TRUE(found_match);
#endif // ROCKSDB_USING_THREAD_STATUS
}
void OnFlushCompleted(
DB* db, const FlushJobInfo& info) override {
flushed_dbs_.push_back(db);
flushed_column_family_names_.push_back(info.cf_name);
if (info.triggered_writes_slowdown) {
slowdown_count++;
}
if (info.triggered_writes_stop) {
stop_count++;
}
// verify whether the previously created file matches the flushed file.
ASSERT_EQ(prev_fc_info_.db_name, db->GetName());
ASSERT_EQ(prev_fc_info_.cf_name, info.cf_name);
ASSERT_EQ(prev_fc_info_.job_id, info.job_id);
ASSERT_EQ(prev_fc_info_.file_path, info.file_path);
ASSERT_EQ(TableFileNameToNumber(info.file_path), info.file_number);
// Note: the following chunk relies on the notification pertaining to the
// database pointed to by DBTestBase::db_, and is thus bypassed when
// that assumption does not hold (see the test case MultiDBMultiListeners
// below).
ASSERT_TRUE(test_);
if (db == test_->db_) {
std::vector<std::vector<FileMetaData>> files_by_level;
More asserts in listener_test for debuggability (#9320) Summary: We ran into a flake I could not debug so instead added assertions in case it happens again. Command was: ``` TEST_TMPDIR=/dev/shm/rocksdb COMPILE_WITH_UBSAN=1 USE_CLANG=1 OPT=-g SKIP_FORMAT_BUCK_CHECKS=1 make J=80 -j80 ubsan_check ``` Failure output was: ``` [==========] Running 1 test from 1 test case. [----------] Global test environment set-up. [----------] 1 test from EventListenerTest [ RUN ] EventListenerTest.DisableBGCompaction UndefinedBehaviorSanitizer:DEADLYSIGNAL ==1558126==ERROR: UndefinedBehaviorSanitizer: SEGV on unknown address 0x000000000031 (pc 0x7fd9c04dda22 bp 0x7fd9bf8aa580 sp 0x7fd9bf8aa540 T1558147) ==1558126==The signal is caused by a READ memory access. ==1558126==Hint: address points to the zero page. #0 0x7fd9c04dda21 in __dynamic_cast /home/engshare/third-party2/libgcc/9.x/src/gcc-9.x/x86_64-facebook-linux/libstdc++-v3/libsupc++/../../.././libstdc++-v3/libsupc++/dyncast.cc:49:3 https://github.com/facebook/rocksdb/issues/1 0x510c53 in __ubsan::checkDynamicType(void*, void*, unsigned long) (/data/sandcastle/boxes/eden-trunk-hg-fbcode-fbsource/fbcode/internal_repo_rocksdb/repo/listener_test+0x510c53) https://github.com/facebook/rocksdb/issues/2 0x50fb32 in HandleDynamicTypeCacheMiss(__ubsan::DynamicTypeCacheMissData*, unsigned long, unsigned long, __ubsan::ReportOptions) (/data/sandcastle/boxes/eden-trunk-hg-fbcode-fbsource/fbcode/internal_repo_rocksdb/repo/listener_test+0x50fb32) https://github.com/facebook/rocksdb/issues/3 0x510230 in __ubsan_handle_dynamic_type_cache_miss_abort (/data/sandcastle/boxes/eden-trunk-hg-fbcode-fbsource/fbcode/internal_repo_rocksdb/repo/listener_test+0x510230) https://github.com/facebook/rocksdb/issues/4 0x63221a in rocksdb::ColumnFamilyHandleImpl* rocksdb::static_cast_with_check<rocksdb::ColumnFamilyHandleImpl, rocksdb::ColumnFamilyHandle>(rocksdb::ColumnFamilyHandle*) /data/sandcastle/boxes/trunk-hg-fbcode-fbsource/fbcode/internal_repo_rocksdb/repo/./util/cast_util.h:19:20 https://github.com/facebook/rocksdb/issues/5 0x71cafa in rocksdb::DBImpl::TEST_GetFilesMetaData(rocksdb::ColumnFamilyHandle*, std::vector<std::vector<rocksdb::FileMetaData, std::allocator<rocksdb::FileMetaData> >, std::allocator<std::vector<rocksdb::FileMetaData, std::allocator<rocksdb::FileMetaData> > > >*, std::vector<std::shared_ptr<rocksdb::BlobFileMetaData>, std::allocator<std::shared_ptr<rocksdb::BlobFileMetaData> > >*) /data/sandcastle/boxes/trunk-hg-fbcode-fbsource/fbcode/internal_repo_rocksdb/repo/db/db_impl/db_impl_debug.cc:63:14 https://github.com/facebook/rocksdb/issues/6 0x53f6b4 in rocksdb::TestFlushListener::OnFlushCompleted(rocksdb::DB*, rocksdb::FlushJobInfo const&) /data/sandcastle/boxes/trunk-hg-fbcode-fbsource/fbcode/internal_repo_rocksdb/repo/db/listener_test.cc:277:24 https://github.com/facebook/rocksdb/issues/7 0x6e2f7d in rocksdb::DBImpl::NotifyOnFlushCompleted(rocksdb::ColumnFamilyData*, rocksdb::MutableCFOptions const&, std::__cxx11::list<std::unique_ptr<rocksdb::FlushJobInfo, std::default_delete<rocksdb::FlushJobInfo> >, std::allocator<std::unique_ptr<rocksdb::FlushJobInfo, std::default_delete<rocksdb::FlushJobInfo> > > >*) /data/sandcastle/boxes/trunk-hg-fbcode-fbsource/fbcode/internal_repo_rocksdb/repo/db/db_impl/db_impl_compaction_flush.cc:863:19 https://github.com/facebook/rocksdb/issues/8 0x6e1074 in rocksdb::DBImpl::FlushMemTableToOutputFile(rocksdb::ColumnFamilyData*, rocksdb::MutableCFOptions const&, bool*, rocksdb::JobContext*, rocksdb::SuperVersionContext*, std::vector<unsigned long, std::allocator<unsigned long> >&, unsigned long, rocksdb::SnapshotChecker*, rocksdb::LogBuffer*, rocksdb::Env::Priority) /data/sandcastle/boxes/trunk-hg-fbcode-fbsource/fbcode/internal_repo_rocksdb/repo/db/db_impl/db_impl_compaction_flush.cc:314:5 https://github.com/facebook/rocksdb/issues/9 0x6e3412 in rocksdb::DBImpl::FlushMemTablesToOutputFiles(rocksdb::autovector<rocksdb::DBImpl::BGFlushArg, 8ul> const&, bool*, rocksdb::JobContext*, rocksdb::LogBuffer*, rocksdb::Env::Priority) /data/sandcastle/boxes/trunk-hg-fbcode-fbsource/fbcode/internal_repo_rocksdb/repo/db/db_impl/db_impl_compaction_flush.cc:359:14 https://github.com/facebook/rocksdb/issues/10 0x700df6 in rocksdb::DBImpl::BackgroundFlush(bool*, rocksdb::JobContext*, rocksdb::LogBuffer*, rocksdb::FlushReason*, rocksdb::Env::Priority) /data/sandcastle/boxes/trunk-hg-fbcode-fbsource/fbcode/internal_repo_rocksdb/repo/db/db_impl/db_impl_compaction_flush.cc:2703:14 https://github.com/facebook/rocksdb/issues/11 0x6fe1f0 in rocksdb::DBImpl::BackgroundCallFlush(rocksdb::Env::Priority) /data/sandcastle/boxes/trunk-hg-fbcode-fbsource/fbcode/internal_repo_rocksdb/repo/db/db_impl/db_impl_compaction_flush.cc:2742:16 https://github.com/facebook/rocksdb/issues/12 0x6fc732 in rocksdb::DBImpl::BGWorkFlush(void*) /data/sandcastle/boxes/trunk-hg-fbcode-fbsource/fbcode/internal_repo_rocksdb/repo/db/db_impl/db_impl_compaction_flush.cc:2569:44 https://github.com/facebook/rocksdb/issues/13 0xb3a820 in void std::_Bind<void (* (void*))(void*)>::operator()<void>() /mnt/gvfs/third-party2/libgcc/4959b39cfbe5965a37c861c4c327fa7c5c759b87/9.x/platform009/9202ce7/include/c++/9.x/functional:482:17 https://github.com/facebook/rocksdb/issues/14 0xb3a820 in std::_Function_handler<void (), std::_Bind<void (* (void*))(void*)> >::_M_invoke(std::_Any_data const&) /mnt/gvfs/third-party2/libgcc/4959b39cfbe5965a37c861c4c327fa7c5c759b87/9.x/platform009/9202ce7/include/c++/9.x/bits/std_function.h:300:2 https://github.com/facebook/rocksdb/issues/15 0xb347cc in rocksdb::ThreadPoolImpl::Impl::BGThread(unsigned long) /data/sandcastle/boxes/trunk-hg-fbcode-fbsource/fbcode/internal_repo_rocksdb/repo/util/threadpool_imp.cc:266:5 https://github.com/facebook/rocksdb/issues/16 0xb34a2f in rocksdb::ThreadPoolImpl::Impl::BGThreadWrapper(void*) /data/sandcastle/boxes/trunk-hg-fbcode-fbsource/fbcode/internal_repo_rocksdb/repo/util/threadpool_imp.cc:307:7 https://github.com/facebook/rocksdb/issues/17 0x7fd9c051a660 in execute_native_thread_routine /home/engshare/third-party2/libgcc/9.x/src/gcc-9.x/x86_64-facebook-linux/libstdc++-v3/src/c++11/../../../.././libstdc++-v3/src/c++11/thread.cc:80:18 https://github.com/facebook/rocksdb/issues/18 0x7fd9c041e20b in start_thread /home/engshare/third-party2/glibc/2.30/src/glibc-2.30/nptl/pthread_create.c:479:8 https://github.com/facebook/rocksdb/issues/19 0x7fd9c01dd16e in clone /home/engshare/third-party2/glibc/2.30/src/glibc-2.30/misc/../sysdeps/unix/sysv/linux/x86_64/clone.S:95 ``` Pull Request resolved: https://github.com/facebook/rocksdb/pull/9320 Reviewed By: jay-zhuang Differential Revision: D33242185 Pulled By: ajkr fbshipit-source-id: 741984b10a610e0509e0d4e54c42cdbac03f5285
3 years ago
ASSERT_LT(info.cf_id, test_->handles_.size());
ASSERT_GE(info.cf_id, 0u);
ASSERT_NE(test_->handles_[info.cf_id], nullptr);
test_->dbfull()->TEST_GetFilesMetaData(test_->handles_[info.cf_id],
&files_by_level);
ASSERT_FALSE(files_by_level.empty());
auto it = std::find_if(files_by_level[0].begin(), files_by_level[0].end(),
[&](const FileMetaData& meta) {
return meta.fd.GetNumber() == info.file_number;
});
ASSERT_NE(it, files_by_level[0].end());
ASSERT_EQ(info.oldest_blob_file_number, it->oldest_blob_file_number);
}
ASSERT_EQ(db->GetEnv()->GetThreadID(), info.thread_id);
ASSERT_GT(info.thread_id, 0U);
ASSERT_EQ(info.table_properties.user_collected_properties.find("0")->second,
"1");
}
std::vector<std::string> flushed_column_family_names_;
std::vector<DB*> flushed_dbs_;
int slowdown_count;
int stop_count;
bool db_closing;
std::atomic_bool db_closed;
TableFileCreationInfo prev_fc_info_;
protected:
Env* env_;
EventListenerTest* test_;
};
rocksdb: switch to gtest Summary: Our existing test notation is very similar to what is used in gtest. It makes it easy to adopt what is different. In this diff I modify existing [[ https://code.google.com/p/googletest/wiki/Primer#Test_Fixtures:_Using_the_Same_Data_Configuration_for_Multiple_Te | test fixture ]] classes to inherit from `testing::Test`. Also for unit tests that use fixture class, `TEST` is replaced with `TEST_F` as required in gtest. There are several custom `main` functions in our existing tests. To make this transition easier, I modify all `main` functions to fallow gtest notation. But eventually we can remove them and use implementation of `main` that gtest provides. ```lang=bash % cat ~/transform #!/bin/sh files=$(git ls-files '*test\.cc') for file in $files do if grep -q "rocksdb::test::RunAllTests()" $file then if grep -Eq '^class \w+Test {' $file then perl -pi -e 's/^(class \w+Test) {/${1}: public testing::Test {/g' $file perl -pi -e 's/^(TEST)/${1}_F/g' $file fi perl -pi -e 's/(int main.*\{)/${1}::testing::InitGoogleTest(&argc, argv);/g' $file perl -pi -e 's/rocksdb::test::RunAllTests/RUN_ALL_TESTS/g' $file fi done % sh ~/transform % make format ``` Second iteration of this diff contains only scripted changes. Third iteration contains manual changes to fix last errors and make it compilable. Test Plan: Build and notice no errors. ```lang=bash % USE_CLANG=1 make check -j55 ``` Tests are still testing. Reviewers: meyering, sdong, rven, igor Reviewed By: igor Subscribers: dhruba, leveldb Differential Revision: https://reviews.facebook.net/D35157
10 years ago
TEST_F(EventListenerTest, OnSingleDBFlushTest) {
Options options;
Encryption at rest support Summary: This PR adds support for encrypting data stored by RocksDB when written to disk. It adds an `EncryptedEnv` override of the `Env` class with matching overrides for sequential&random access files. The encryption itself is done through a configurable `EncryptionProvider`. This class creates is asked to create `BlockAccessCipherStream` for a file. This is where the actual encryption/decryption is being done. Currently there is a Counter mode implementation of `BlockAccessCipherStream` with a `ROT13` block cipher (NOTE the `ROT13` is for demo purposes only!!). The Counter operation mode uses an initial counter & random initialization vector (IV). Both are created randomly for each file and stored in a 4K (default size) block that is prefixed to that file. The `EncryptedEnv` implementation is such that clients of the `Env` class do not see this prefix (nor data, nor in filesize). The largest part of the prefix block is also encrypted, and there is room left for implementation specific settings/values/keys in there. To test the encryption, the `DBTestBase` class has been extended to consider a new environment variable called `ENCRYPTED_ENV`. If set, the test will setup a encrypted instance of the `Env` class to use for all tests. Typically you would run it like this: ``` ENCRYPTED_ENV=1 make check_some ``` There is also an added test that checks that some data inserted into the database is or is not "visible" on disk. With `ENCRYPTED_ENV` active it must not find plain text strings, with `ENCRYPTED_ENV` unset, it must find the plain text strings. Closes https://github.com/facebook/rocksdb/pull/2424 Differential Revision: D5322178 Pulled By: sdwilsh fbshipit-source-id: 253b0a9c2c498cc98f580df7f2623cbf7678a27f
7 years ago
options.env = CurrentOptions().env;
options.write_buffer_size = k110KB;
#ifdef ROCKSDB_USING_THREAD_STATUS
options.enable_thread_tracking = true;
#endif // ROCKSDB_USING_THREAD_STATUS
TestFlushListener* listener = new TestFlushListener(options.env, this);
options.listeners.emplace_back(listener);
std::vector<std::string> cf_names = {
"pikachu", "ilya", "muromec", "dobrynia",
"nikitich", "alyosha", "popovich"};
options.table_properties_collector_factories.push_back(
std::make_shared<TestPropertiesCollectorFactory>());
CreateAndReopenWithCF(cf_names, options);
ASSERT_OK(Put(1, "pikachu", std::string(90000, 'p')));
WriteBatch batch;
ASSERT_OK(WriteBatchInternal::PutBlobIndex(&batch, 1, "ditto",
BlobStr(456, 0, 1 << 10)));
ASSERT_OK(dbfull()->Write(WriteOptions(), &batch));
ASSERT_OK(Put(2, "ilya", std::string(90000, 'i')));
ASSERT_OK(Put(3, "muromec", std::string(90000, 'm')));
ASSERT_OK(Put(4, "dobrynia", std::string(90000, 'd')));
ASSERT_OK(Put(5, "nikitich", std::string(90000, 'n')));
ASSERT_OK(Put(6, "alyosha", std::string(90000, 'a')));
ASSERT_OK(Put(7, "popovich", std::string(90000, 'p')));
for (int i = 1; i < 8; ++i) {
ASSERT_OK(Flush(i));
ASSERT_OK(dbfull()->TEST_WaitForFlushMemTable());
Fix test race conditions with OnFlushCompleted() (#9617) Summary: We often see flaky tests due to `DB::Flush()` or `DBImpl::TEST_WaitForFlushMemTable()` not waiting until event listeners complete. For example, https://github.com/facebook/rocksdb/issues/9084, https://github.com/facebook/rocksdb/issues/9400, https://github.com/facebook/rocksdb/issues/9528, plus two new ones this week: "EventListenerTest.OnSingleDBFlushTest" and "DBFlushTest.FireOnFlushCompletedAfterCommittedResult". I ran a `make check` with the below race condition-coercing patch and fixed issues it found besides old BlobDB. ``` diff --git a/db/db_impl/db_impl_compaction_flush.cc b/db/db_impl/db_impl_compaction_flush.cc index 0e1864788..aaba68c4a 100644 --- a/db/db_impl/db_impl_compaction_flush.cc +++ b/db/db_impl/db_impl_compaction_flush.cc @@ -861,6 +861,8 @@ void DBImpl::NotifyOnFlushCompleted( mutable_cf_options.level0_stop_writes_trigger); // release lock while notifying events mutex_.Unlock(); + bg_cv_.SignalAll(); + sleep(1); { for (auto& info : *flush_jobs_info) { info->triggered_writes_slowdown = triggered_writes_slowdown; ``` The reason I did not fix old BlobDB issues is because it appears to have a fundamental (non-test) issue. In particular, it uses an EventListener to keep track of the files. OnFlushCompleted() could be delayed until even after a compaction involving that flushed file completes, causing the compaction to unexpectedly delete an untracked file. Pull Request resolved: https://github.com/facebook/rocksdb/pull/9617 Test Plan: `make check` including the race condition coercing patch Reviewed By: hx235 Differential Revision: D34384022 Pulled By: ajkr fbshipit-source-id: 2652ded39b415277c5d6a628414345223930514e
3 years ago
// Ensure background work is fully finished including listener callbacks
// before accessing listener state.
ASSERT_OK(dbfull()->TEST_WaitForBackgroundWork());
ASSERT_EQ(listener->flushed_dbs_.size(), i);
ASSERT_EQ(listener->flushed_column_family_names_.size(), i);
}
// make sure callback functions are called in the right order
for (size_t i = 0; i < cf_names.size(); ++i) {
ASSERT_EQ(listener->flushed_dbs_[i], db_);
ASSERT_EQ(listener->flushed_column_family_names_[i], cf_names[i]);
}
}
rocksdb: switch to gtest Summary: Our existing test notation is very similar to what is used in gtest. It makes it easy to adopt what is different. In this diff I modify existing [[ https://code.google.com/p/googletest/wiki/Primer#Test_Fixtures:_Using_the_Same_Data_Configuration_for_Multiple_Te | test fixture ]] classes to inherit from `testing::Test`. Also for unit tests that use fixture class, `TEST` is replaced with `TEST_F` as required in gtest. There are several custom `main` functions in our existing tests. To make this transition easier, I modify all `main` functions to fallow gtest notation. But eventually we can remove them and use implementation of `main` that gtest provides. ```lang=bash % cat ~/transform #!/bin/sh files=$(git ls-files '*test\.cc') for file in $files do if grep -q "rocksdb::test::RunAllTests()" $file then if grep -Eq '^class \w+Test {' $file then perl -pi -e 's/^(class \w+Test) {/${1}: public testing::Test {/g' $file perl -pi -e 's/^(TEST)/${1}_F/g' $file fi perl -pi -e 's/(int main.*\{)/${1}::testing::InitGoogleTest(&argc, argv);/g' $file perl -pi -e 's/rocksdb::test::RunAllTests/RUN_ALL_TESTS/g' $file fi done % sh ~/transform % make format ``` Second iteration of this diff contains only scripted changes. Third iteration contains manual changes to fix last errors and make it compilable. Test Plan: Build and notice no errors. ```lang=bash % USE_CLANG=1 make check -j55 ``` Tests are still testing. Reviewers: meyering, sdong, rven, igor Reviewed By: igor Subscribers: dhruba, leveldb Differential Revision: https://reviews.facebook.net/D35157
10 years ago
TEST_F(EventListenerTest, MultiCF) {
Options options;
Encryption at rest support Summary: This PR adds support for encrypting data stored by RocksDB when written to disk. It adds an `EncryptedEnv` override of the `Env` class with matching overrides for sequential&random access files. The encryption itself is done through a configurable `EncryptionProvider`. This class creates is asked to create `BlockAccessCipherStream` for a file. This is where the actual encryption/decryption is being done. Currently there is a Counter mode implementation of `BlockAccessCipherStream` with a `ROT13` block cipher (NOTE the `ROT13` is for demo purposes only!!). The Counter operation mode uses an initial counter & random initialization vector (IV). Both are created randomly for each file and stored in a 4K (default size) block that is prefixed to that file. The `EncryptedEnv` implementation is such that clients of the `Env` class do not see this prefix (nor data, nor in filesize). The largest part of the prefix block is also encrypted, and there is room left for implementation specific settings/values/keys in there. To test the encryption, the `DBTestBase` class has been extended to consider a new environment variable called `ENCRYPTED_ENV`. If set, the test will setup a encrypted instance of the `Env` class to use for all tests. Typically you would run it like this: ``` ENCRYPTED_ENV=1 make check_some ``` There is also an added test that checks that some data inserted into the database is or is not "visible" on disk. With `ENCRYPTED_ENV` active it must not find plain text strings, with `ENCRYPTED_ENV` unset, it must find the plain text strings. Closes https://github.com/facebook/rocksdb/pull/2424 Differential Revision: D5322178 Pulled By: sdwilsh fbshipit-source-id: 253b0a9c2c498cc98f580df7f2623cbf7678a27f
7 years ago
options.env = CurrentOptions().env;
options.write_buffer_size = k110KB;
#ifdef ROCKSDB_USING_THREAD_STATUS
options.enable_thread_tracking = true;
#endif // ROCKSDB_USING_THREAD_STATUS
for (auto atomic_flush : {false, true}) {
options.atomic_flush = atomic_flush;
options.create_if_missing = true;
DestroyAndReopen(options);
TestFlushListener* listener = new TestFlushListener(options.env, this);
options.listeners.emplace_back(listener);
options.table_properties_collector_factories.push_back(
std::make_shared<TestPropertiesCollectorFactory>());
std::vector<std::string> cf_names = {"pikachu", "ilya", "muromec",
"dobrynia", "nikitich", "alyosha",
"popovich"};
CreateAndReopenWithCF(cf_names, options);
ASSERT_OK(Put(1, "pikachu", std::string(90000, 'p')));
ASSERT_OK(Put(2, "ilya", std::string(90000, 'i')));
ASSERT_OK(Put(3, "muromec", std::string(90000, 'm')));
ASSERT_OK(Put(4, "dobrynia", std::string(90000, 'd')));
ASSERT_OK(Put(5, "nikitich", std::string(90000, 'n')));
ASSERT_OK(Put(6, "alyosha", std::string(90000, 'a')));
ASSERT_OK(Put(7, "popovich", std::string(90000, 'p')));
Fix TSAN data race in EventListenerTest.MultiCF (#9528) Summary: **Context:** `EventListenerTest.MultiCF` occasionally failed on TSAN data race as below: ``` WARNING: ThreadSanitizer: data race (pid=2047633) Read of size 8 at 0x7b6000001440 by main thread: #0 std::vector<rocksdb::DB*, std::allocator<rocksdb::DB*> >::size() const /usr/bin/../lib/gcc/x86_64-linux-gnu/9/../../../../include/c++/9/bits/stl_vector.h:916:40 (listener_test+0x52337c) https://github.com/facebook/rocksdb/issues/1 rocksdb::EventListenerTest_MultiCF_Test::TestBody() /home/circleci/project/db/listener_test.cc:384:7 (listener_test+0x52337c) Previous write of size 8 at 0x7b6000001440 by thread T2: #0 void std::vector<rocksdb::DB*, std::allocator<rocksdb::DB*> >::_M_realloc_insert<rocksdb::DB* const&>(__gnu_cxx::__normal_iterator<rocksdb::DB**, std::vector<rocksdb::DB*, std::allocator<rocksdb::DB*> > >, rocksdb::DB* const&) /usr/bin/../lib/gcc/x86_64-linux-gnu/9/../../../../include/c++/9/bits/vector.tcc:503:31 (listener_test+0x550654) https://github.com/facebook/rocksdb/issues/1 std::vector<rocksdb::DB*, std::allocator<rocksdb::DB*> >::push_back(rocksdb::DB* const&) /usr/bin/../lib/gcc/x86_64-linux-gnu/9/../../../../include/c++/9/bits/stl_vector.h:1195:4 (listener_test+0x550654) https://github.com/facebook/rocksdb/issues/2 rocksdb::TestFlushListener::OnFlushCompleted(rocksdb::DB*, rocksdb::FlushJobInfo const&) /home/circleci/project/db/listener_test.cc:255:18 (listener_test+0x550654) ``` After investigation, it is due to the following: (1) `ASSERT_OK(Flush(i));` before the read `std::vector::size()` is supposed to be [blocked on `DB::Impl::bg_cv_` for memtable flush to finish](https://github.com/facebook/rocksdb/blob/320d9a8e8a1b6998f92934f87fc71ad8bd6d4596/db/db_impl/db_impl_compaction_flush.cc#L2319) and get signaled [at the end of background flush ](https://github.com/facebook/rocksdb/blob/320d9a8e8a1b6998f92934f87fc71ad8bd6d4596/db/db_impl/db_impl_compaction_flush.cc#L2830), which happens after the write `std::vector::push_back()` . So the sequence of execution should have been synchronized as `call flush() -> write -> return from flush() -> read` and would not cause any TSAN data race. - The subsequent `ASSERT_OK(dbfull()->TEST_WaitForFlushMemTable());` serves a similar purpose based on [the previous attempt to deflake the test.](https://github.com/facebook/rocksdb/pull/9084) (2) However, there are multiple places in the code can signal this `DB::Impl::bg_cv_` and mistakenly wake up `ASSERT_OK(Flush(i));` (or `ASSERT_OK(dbfull()->TEST_WaitForFlushMemTable());`) too early (and with the lock available to them), resulting in non-synchronized read and write thus a TSAN data race. - Reproduced by the following, suggested by ajkr: ``` diff --git a/db/db_impl/db_impl_compaction_flush.cc b/db/db_impl/db_impl_compaction_flush.cc index 4ff87c1e4..52492e9cf 100644 --- a/db/db_impl/db_impl_compaction_flush.cc +++ b/db/db_impl/db_impl_compaction_flush.cc @@ -22,7 +22,7 @@ #include "test_util/sync_point.h" #include "util/cast_util.h" #include "util/concurrent_task_limiter_impl.h" namespace ROCKSDB_NAMESPACE { bool DBImpl::EnoughRoomForCompaction( @@ -855,6 +855,7 @@ void DBImpl::NotifyOnFlushCompleted( mutable_cf_options.level0_stop_writes_trigger); // release lock while notifying events mutex_.Unlock(); + bg_cv_.SignalAll(); ``` **Summary:** - Added synchornization between read and write by ` ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->LoadDependency()` mechanism Pull Request resolved: https://github.com/facebook/rocksdb/pull/9528 Test Plan: `./listener_test --gtest_filter=EventListenerTest.MultiCF --gtest_repeat=10` - pre-fix: ``` Repeating all tests (iteration 3) Note: Google Test filter = EventListenerTest.MultiCF [==========] Running 1 test from 1 test case. [----------] Global test environment set-up. [----------] 1 test from EventListenerTest [ RUN ] EventListenerTest.MultiCF ================== WARNING: ThreadSanitizer: data race (pid=3377137) Read of size 8 at 0x7b6000000840 by main thread: #0 std::vector<rocksdb::DB*, std::allocator<rocksdb::DB*> >::size() https://github.com/facebook/rocksdb/issues/1 rocksdb::EventListenerTest_MultiCF_Test::TestBody() db/listener_test.cc:384 (listener_test+0x4bb300) Previous write of size 8 at 0x7b6000000840 by thread T2: #0 void std::vector<rocksdb::DB*, std::allocator<rocksdb::DB*> >::_M_realloc_insert<rocksdb::DB* const&>(__gnu_cxx::__normal_iterator<rocksdb::DB**, std::vector<rocksdb::DB*, std::allocator<rocksdb::DB*> > >, rocksdb::DB* const&) https://github.com/facebook/rocksdb/issues/1 std::vector<rocksdb::DB*, std::allocator<rocksdb::DB*> >::push_back(rocksdb::DB* const&) https://github.com/facebook/rocksdb/issues/2 rocksdb::TestFlushListener::OnFlushCompleted(rocksdb::DB*, rocksdb::FlushJobInfo const&) db/listener_test.cc:255 (listener_test+0x4e820f) ``` - post-fix: `All passed` Reviewed By: ajkr Differential Revision: D34085791 Pulled By: hx235 fbshipit-source-id: f877aa687ea1d5cb6f31ef8c4772625d22868e8b
3 years ago
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->EnableProcessing();
for (int i = 1; i < 8; ++i) {
Fix TSAN data race in EventListenerTest.MultiCF (#9528) Summary: **Context:** `EventListenerTest.MultiCF` occasionally failed on TSAN data race as below: ``` WARNING: ThreadSanitizer: data race (pid=2047633) Read of size 8 at 0x7b6000001440 by main thread: #0 std::vector<rocksdb::DB*, std::allocator<rocksdb::DB*> >::size() const /usr/bin/../lib/gcc/x86_64-linux-gnu/9/../../../../include/c++/9/bits/stl_vector.h:916:40 (listener_test+0x52337c) https://github.com/facebook/rocksdb/issues/1 rocksdb::EventListenerTest_MultiCF_Test::TestBody() /home/circleci/project/db/listener_test.cc:384:7 (listener_test+0x52337c) Previous write of size 8 at 0x7b6000001440 by thread T2: #0 void std::vector<rocksdb::DB*, std::allocator<rocksdb::DB*> >::_M_realloc_insert<rocksdb::DB* const&>(__gnu_cxx::__normal_iterator<rocksdb::DB**, std::vector<rocksdb::DB*, std::allocator<rocksdb::DB*> > >, rocksdb::DB* const&) /usr/bin/../lib/gcc/x86_64-linux-gnu/9/../../../../include/c++/9/bits/vector.tcc:503:31 (listener_test+0x550654) https://github.com/facebook/rocksdb/issues/1 std::vector<rocksdb::DB*, std::allocator<rocksdb::DB*> >::push_back(rocksdb::DB* const&) /usr/bin/../lib/gcc/x86_64-linux-gnu/9/../../../../include/c++/9/bits/stl_vector.h:1195:4 (listener_test+0x550654) https://github.com/facebook/rocksdb/issues/2 rocksdb::TestFlushListener::OnFlushCompleted(rocksdb::DB*, rocksdb::FlushJobInfo const&) /home/circleci/project/db/listener_test.cc:255:18 (listener_test+0x550654) ``` After investigation, it is due to the following: (1) `ASSERT_OK(Flush(i));` before the read `std::vector::size()` is supposed to be [blocked on `DB::Impl::bg_cv_` for memtable flush to finish](https://github.com/facebook/rocksdb/blob/320d9a8e8a1b6998f92934f87fc71ad8bd6d4596/db/db_impl/db_impl_compaction_flush.cc#L2319) and get signaled [at the end of background flush ](https://github.com/facebook/rocksdb/blob/320d9a8e8a1b6998f92934f87fc71ad8bd6d4596/db/db_impl/db_impl_compaction_flush.cc#L2830), which happens after the write `std::vector::push_back()` . So the sequence of execution should have been synchronized as `call flush() -> write -> return from flush() -> read` and would not cause any TSAN data race. - The subsequent `ASSERT_OK(dbfull()->TEST_WaitForFlushMemTable());` serves a similar purpose based on [the previous attempt to deflake the test.](https://github.com/facebook/rocksdb/pull/9084) (2) However, there are multiple places in the code can signal this `DB::Impl::bg_cv_` and mistakenly wake up `ASSERT_OK(Flush(i));` (or `ASSERT_OK(dbfull()->TEST_WaitForFlushMemTable());`) too early (and with the lock available to them), resulting in non-synchronized read and write thus a TSAN data race. - Reproduced by the following, suggested by ajkr: ``` diff --git a/db/db_impl/db_impl_compaction_flush.cc b/db/db_impl/db_impl_compaction_flush.cc index 4ff87c1e4..52492e9cf 100644 --- a/db/db_impl/db_impl_compaction_flush.cc +++ b/db/db_impl/db_impl_compaction_flush.cc @@ -22,7 +22,7 @@ #include "test_util/sync_point.h" #include "util/cast_util.h" #include "util/concurrent_task_limiter_impl.h" namespace ROCKSDB_NAMESPACE { bool DBImpl::EnoughRoomForCompaction( @@ -855,6 +855,7 @@ void DBImpl::NotifyOnFlushCompleted( mutable_cf_options.level0_stop_writes_trigger); // release lock while notifying events mutex_.Unlock(); + bg_cv_.SignalAll(); ``` **Summary:** - Added synchornization between read and write by ` ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->LoadDependency()` mechanism Pull Request resolved: https://github.com/facebook/rocksdb/pull/9528 Test Plan: `./listener_test --gtest_filter=EventListenerTest.MultiCF --gtest_repeat=10` - pre-fix: ``` Repeating all tests (iteration 3) Note: Google Test filter = EventListenerTest.MultiCF [==========] Running 1 test from 1 test case. [----------] Global test environment set-up. [----------] 1 test from EventListenerTest [ RUN ] EventListenerTest.MultiCF ================== WARNING: ThreadSanitizer: data race (pid=3377137) Read of size 8 at 0x7b6000000840 by main thread: #0 std::vector<rocksdb::DB*, std::allocator<rocksdb::DB*> >::size() https://github.com/facebook/rocksdb/issues/1 rocksdb::EventListenerTest_MultiCF_Test::TestBody() db/listener_test.cc:384 (listener_test+0x4bb300) Previous write of size 8 at 0x7b6000000840 by thread T2: #0 void std::vector<rocksdb::DB*, std::allocator<rocksdb::DB*> >::_M_realloc_insert<rocksdb::DB* const&>(__gnu_cxx::__normal_iterator<rocksdb::DB**, std::vector<rocksdb::DB*, std::allocator<rocksdb::DB*> > >, rocksdb::DB* const&) https://github.com/facebook/rocksdb/issues/1 std::vector<rocksdb::DB*, std::allocator<rocksdb::DB*> >::push_back(rocksdb::DB* const&) https://github.com/facebook/rocksdb/issues/2 rocksdb::TestFlushListener::OnFlushCompleted(rocksdb::DB*, rocksdb::FlushJobInfo const&) db/listener_test.cc:255 (listener_test+0x4e820f) ``` - post-fix: `All passed` Reviewed By: ajkr Differential Revision: D34085791 Pulled By: hx235 fbshipit-source-id: f877aa687ea1d5cb6f31ef8c4772625d22868e8b
3 years ago
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->LoadDependency(
{{"DBImpl::NotifyOnFlushCompleted::PostAllOnFlushCompleted",
"EventListenerTest.MultiCF:PreVerifyListener"}});
ASSERT_OK(Flush(i));
Fix TSAN data race in EventListenerTest.MultiCF (#9528) Summary: **Context:** `EventListenerTest.MultiCF` occasionally failed on TSAN data race as below: ``` WARNING: ThreadSanitizer: data race (pid=2047633) Read of size 8 at 0x7b6000001440 by main thread: #0 std::vector<rocksdb::DB*, std::allocator<rocksdb::DB*> >::size() const /usr/bin/../lib/gcc/x86_64-linux-gnu/9/../../../../include/c++/9/bits/stl_vector.h:916:40 (listener_test+0x52337c) https://github.com/facebook/rocksdb/issues/1 rocksdb::EventListenerTest_MultiCF_Test::TestBody() /home/circleci/project/db/listener_test.cc:384:7 (listener_test+0x52337c) Previous write of size 8 at 0x7b6000001440 by thread T2: #0 void std::vector<rocksdb::DB*, std::allocator<rocksdb::DB*> >::_M_realloc_insert<rocksdb::DB* const&>(__gnu_cxx::__normal_iterator<rocksdb::DB**, std::vector<rocksdb::DB*, std::allocator<rocksdb::DB*> > >, rocksdb::DB* const&) /usr/bin/../lib/gcc/x86_64-linux-gnu/9/../../../../include/c++/9/bits/vector.tcc:503:31 (listener_test+0x550654) https://github.com/facebook/rocksdb/issues/1 std::vector<rocksdb::DB*, std::allocator<rocksdb::DB*> >::push_back(rocksdb::DB* const&) /usr/bin/../lib/gcc/x86_64-linux-gnu/9/../../../../include/c++/9/bits/stl_vector.h:1195:4 (listener_test+0x550654) https://github.com/facebook/rocksdb/issues/2 rocksdb::TestFlushListener::OnFlushCompleted(rocksdb::DB*, rocksdb::FlushJobInfo const&) /home/circleci/project/db/listener_test.cc:255:18 (listener_test+0x550654) ``` After investigation, it is due to the following: (1) `ASSERT_OK(Flush(i));` before the read `std::vector::size()` is supposed to be [blocked on `DB::Impl::bg_cv_` for memtable flush to finish](https://github.com/facebook/rocksdb/blob/320d9a8e8a1b6998f92934f87fc71ad8bd6d4596/db/db_impl/db_impl_compaction_flush.cc#L2319) and get signaled [at the end of background flush ](https://github.com/facebook/rocksdb/blob/320d9a8e8a1b6998f92934f87fc71ad8bd6d4596/db/db_impl/db_impl_compaction_flush.cc#L2830), which happens after the write `std::vector::push_back()` . So the sequence of execution should have been synchronized as `call flush() -> write -> return from flush() -> read` and would not cause any TSAN data race. - The subsequent `ASSERT_OK(dbfull()->TEST_WaitForFlushMemTable());` serves a similar purpose based on [the previous attempt to deflake the test.](https://github.com/facebook/rocksdb/pull/9084) (2) However, there are multiple places in the code can signal this `DB::Impl::bg_cv_` and mistakenly wake up `ASSERT_OK(Flush(i));` (or `ASSERT_OK(dbfull()->TEST_WaitForFlushMemTable());`) too early (and with the lock available to them), resulting in non-synchronized read and write thus a TSAN data race. - Reproduced by the following, suggested by ajkr: ``` diff --git a/db/db_impl/db_impl_compaction_flush.cc b/db/db_impl/db_impl_compaction_flush.cc index 4ff87c1e4..52492e9cf 100644 --- a/db/db_impl/db_impl_compaction_flush.cc +++ b/db/db_impl/db_impl_compaction_flush.cc @@ -22,7 +22,7 @@ #include "test_util/sync_point.h" #include "util/cast_util.h" #include "util/concurrent_task_limiter_impl.h" namespace ROCKSDB_NAMESPACE { bool DBImpl::EnoughRoomForCompaction( @@ -855,6 +855,7 @@ void DBImpl::NotifyOnFlushCompleted( mutable_cf_options.level0_stop_writes_trigger); // release lock while notifying events mutex_.Unlock(); + bg_cv_.SignalAll(); ``` **Summary:** - Added synchornization between read and write by ` ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->LoadDependency()` mechanism Pull Request resolved: https://github.com/facebook/rocksdb/pull/9528 Test Plan: `./listener_test --gtest_filter=EventListenerTest.MultiCF --gtest_repeat=10` - pre-fix: ``` Repeating all tests (iteration 3) Note: Google Test filter = EventListenerTest.MultiCF [==========] Running 1 test from 1 test case. [----------] Global test environment set-up. [----------] 1 test from EventListenerTest [ RUN ] EventListenerTest.MultiCF ================== WARNING: ThreadSanitizer: data race (pid=3377137) Read of size 8 at 0x7b6000000840 by main thread: #0 std::vector<rocksdb::DB*, std::allocator<rocksdb::DB*> >::size() https://github.com/facebook/rocksdb/issues/1 rocksdb::EventListenerTest_MultiCF_Test::TestBody() db/listener_test.cc:384 (listener_test+0x4bb300) Previous write of size 8 at 0x7b6000000840 by thread T2: #0 void std::vector<rocksdb::DB*, std::allocator<rocksdb::DB*> >::_M_realloc_insert<rocksdb::DB* const&>(__gnu_cxx::__normal_iterator<rocksdb::DB**, std::vector<rocksdb::DB*, std::allocator<rocksdb::DB*> > >, rocksdb::DB* const&) https://github.com/facebook/rocksdb/issues/1 std::vector<rocksdb::DB*, std::allocator<rocksdb::DB*> >::push_back(rocksdb::DB* const&) https://github.com/facebook/rocksdb/issues/2 rocksdb::TestFlushListener::OnFlushCompleted(rocksdb::DB*, rocksdb::FlushJobInfo const&) db/listener_test.cc:255 (listener_test+0x4e820f) ``` - post-fix: `All passed` Reviewed By: ajkr Differential Revision: D34085791 Pulled By: hx235 fbshipit-source-id: f877aa687ea1d5cb6f31ef8c4772625d22868e8b
3 years ago
TEST_SYNC_POINT("EventListenerTest.MultiCF:PreVerifyListener");
ASSERT_EQ(listener->flushed_dbs_.size(), i);
ASSERT_EQ(listener->flushed_column_family_names_.size(), i);
Fix TSAN data race in EventListenerTest.MultiCF (#9528) Summary: **Context:** `EventListenerTest.MultiCF` occasionally failed on TSAN data race as below: ``` WARNING: ThreadSanitizer: data race (pid=2047633) Read of size 8 at 0x7b6000001440 by main thread: #0 std::vector<rocksdb::DB*, std::allocator<rocksdb::DB*> >::size() const /usr/bin/../lib/gcc/x86_64-linux-gnu/9/../../../../include/c++/9/bits/stl_vector.h:916:40 (listener_test+0x52337c) https://github.com/facebook/rocksdb/issues/1 rocksdb::EventListenerTest_MultiCF_Test::TestBody() /home/circleci/project/db/listener_test.cc:384:7 (listener_test+0x52337c) Previous write of size 8 at 0x7b6000001440 by thread T2: #0 void std::vector<rocksdb::DB*, std::allocator<rocksdb::DB*> >::_M_realloc_insert<rocksdb::DB* const&>(__gnu_cxx::__normal_iterator<rocksdb::DB**, std::vector<rocksdb::DB*, std::allocator<rocksdb::DB*> > >, rocksdb::DB* const&) /usr/bin/../lib/gcc/x86_64-linux-gnu/9/../../../../include/c++/9/bits/vector.tcc:503:31 (listener_test+0x550654) https://github.com/facebook/rocksdb/issues/1 std::vector<rocksdb::DB*, std::allocator<rocksdb::DB*> >::push_back(rocksdb::DB* const&) /usr/bin/../lib/gcc/x86_64-linux-gnu/9/../../../../include/c++/9/bits/stl_vector.h:1195:4 (listener_test+0x550654) https://github.com/facebook/rocksdb/issues/2 rocksdb::TestFlushListener::OnFlushCompleted(rocksdb::DB*, rocksdb::FlushJobInfo const&) /home/circleci/project/db/listener_test.cc:255:18 (listener_test+0x550654) ``` After investigation, it is due to the following: (1) `ASSERT_OK(Flush(i));` before the read `std::vector::size()` is supposed to be [blocked on `DB::Impl::bg_cv_` for memtable flush to finish](https://github.com/facebook/rocksdb/blob/320d9a8e8a1b6998f92934f87fc71ad8bd6d4596/db/db_impl/db_impl_compaction_flush.cc#L2319) and get signaled [at the end of background flush ](https://github.com/facebook/rocksdb/blob/320d9a8e8a1b6998f92934f87fc71ad8bd6d4596/db/db_impl/db_impl_compaction_flush.cc#L2830), which happens after the write `std::vector::push_back()` . So the sequence of execution should have been synchronized as `call flush() -> write -> return from flush() -> read` and would not cause any TSAN data race. - The subsequent `ASSERT_OK(dbfull()->TEST_WaitForFlushMemTable());` serves a similar purpose based on [the previous attempt to deflake the test.](https://github.com/facebook/rocksdb/pull/9084) (2) However, there are multiple places in the code can signal this `DB::Impl::bg_cv_` and mistakenly wake up `ASSERT_OK(Flush(i));` (or `ASSERT_OK(dbfull()->TEST_WaitForFlushMemTable());`) too early (and with the lock available to them), resulting in non-synchronized read and write thus a TSAN data race. - Reproduced by the following, suggested by ajkr: ``` diff --git a/db/db_impl/db_impl_compaction_flush.cc b/db/db_impl/db_impl_compaction_flush.cc index 4ff87c1e4..52492e9cf 100644 --- a/db/db_impl/db_impl_compaction_flush.cc +++ b/db/db_impl/db_impl_compaction_flush.cc @@ -22,7 +22,7 @@ #include "test_util/sync_point.h" #include "util/cast_util.h" #include "util/concurrent_task_limiter_impl.h" namespace ROCKSDB_NAMESPACE { bool DBImpl::EnoughRoomForCompaction( @@ -855,6 +855,7 @@ void DBImpl::NotifyOnFlushCompleted( mutable_cf_options.level0_stop_writes_trigger); // release lock while notifying events mutex_.Unlock(); + bg_cv_.SignalAll(); ``` **Summary:** - Added synchornization between read and write by ` ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->LoadDependency()` mechanism Pull Request resolved: https://github.com/facebook/rocksdb/pull/9528 Test Plan: `./listener_test --gtest_filter=EventListenerTest.MultiCF --gtest_repeat=10` - pre-fix: ``` Repeating all tests (iteration 3) Note: Google Test filter = EventListenerTest.MultiCF [==========] Running 1 test from 1 test case. [----------] Global test environment set-up. [----------] 1 test from EventListenerTest [ RUN ] EventListenerTest.MultiCF ================== WARNING: ThreadSanitizer: data race (pid=3377137) Read of size 8 at 0x7b6000000840 by main thread: #0 std::vector<rocksdb::DB*, std::allocator<rocksdb::DB*> >::size() https://github.com/facebook/rocksdb/issues/1 rocksdb::EventListenerTest_MultiCF_Test::TestBody() db/listener_test.cc:384 (listener_test+0x4bb300) Previous write of size 8 at 0x7b6000000840 by thread T2: #0 void std::vector<rocksdb::DB*, std::allocator<rocksdb::DB*> >::_M_realloc_insert<rocksdb::DB* const&>(__gnu_cxx::__normal_iterator<rocksdb::DB**, std::vector<rocksdb::DB*, std::allocator<rocksdb::DB*> > >, rocksdb::DB* const&) https://github.com/facebook/rocksdb/issues/1 std::vector<rocksdb::DB*, std::allocator<rocksdb::DB*> >::push_back(rocksdb::DB* const&) https://github.com/facebook/rocksdb/issues/2 rocksdb::TestFlushListener::OnFlushCompleted(rocksdb::DB*, rocksdb::FlushJobInfo const&) db/listener_test.cc:255 (listener_test+0x4e820f) ``` - post-fix: `All passed` Reviewed By: ajkr Differential Revision: D34085791 Pulled By: hx235 fbshipit-source-id: f877aa687ea1d5cb6f31ef8c4772625d22868e8b
3 years ago
// make sure callback functions are called in the right order
if (i == 7) {
for (size_t j = 0; j < cf_names.size(); j++) {
ASSERT_EQ(listener->flushed_dbs_[j], db_);
ASSERT_EQ(listener->flushed_column_family_names_[j], cf_names[j]);
}
}
}
Fix TSAN data race in EventListenerTest.MultiCF (#9528) Summary: **Context:** `EventListenerTest.MultiCF` occasionally failed on TSAN data race as below: ``` WARNING: ThreadSanitizer: data race (pid=2047633) Read of size 8 at 0x7b6000001440 by main thread: #0 std::vector<rocksdb::DB*, std::allocator<rocksdb::DB*> >::size() const /usr/bin/../lib/gcc/x86_64-linux-gnu/9/../../../../include/c++/9/bits/stl_vector.h:916:40 (listener_test+0x52337c) https://github.com/facebook/rocksdb/issues/1 rocksdb::EventListenerTest_MultiCF_Test::TestBody() /home/circleci/project/db/listener_test.cc:384:7 (listener_test+0x52337c) Previous write of size 8 at 0x7b6000001440 by thread T2: #0 void std::vector<rocksdb::DB*, std::allocator<rocksdb::DB*> >::_M_realloc_insert<rocksdb::DB* const&>(__gnu_cxx::__normal_iterator<rocksdb::DB**, std::vector<rocksdb::DB*, std::allocator<rocksdb::DB*> > >, rocksdb::DB* const&) /usr/bin/../lib/gcc/x86_64-linux-gnu/9/../../../../include/c++/9/bits/vector.tcc:503:31 (listener_test+0x550654) https://github.com/facebook/rocksdb/issues/1 std::vector<rocksdb::DB*, std::allocator<rocksdb::DB*> >::push_back(rocksdb::DB* const&) /usr/bin/../lib/gcc/x86_64-linux-gnu/9/../../../../include/c++/9/bits/stl_vector.h:1195:4 (listener_test+0x550654) https://github.com/facebook/rocksdb/issues/2 rocksdb::TestFlushListener::OnFlushCompleted(rocksdb::DB*, rocksdb::FlushJobInfo const&) /home/circleci/project/db/listener_test.cc:255:18 (listener_test+0x550654) ``` After investigation, it is due to the following: (1) `ASSERT_OK(Flush(i));` before the read `std::vector::size()` is supposed to be [blocked on `DB::Impl::bg_cv_` for memtable flush to finish](https://github.com/facebook/rocksdb/blob/320d9a8e8a1b6998f92934f87fc71ad8bd6d4596/db/db_impl/db_impl_compaction_flush.cc#L2319) and get signaled [at the end of background flush ](https://github.com/facebook/rocksdb/blob/320d9a8e8a1b6998f92934f87fc71ad8bd6d4596/db/db_impl/db_impl_compaction_flush.cc#L2830), which happens after the write `std::vector::push_back()` . So the sequence of execution should have been synchronized as `call flush() -> write -> return from flush() -> read` and would not cause any TSAN data race. - The subsequent `ASSERT_OK(dbfull()->TEST_WaitForFlushMemTable());` serves a similar purpose based on [the previous attempt to deflake the test.](https://github.com/facebook/rocksdb/pull/9084) (2) However, there are multiple places in the code can signal this `DB::Impl::bg_cv_` and mistakenly wake up `ASSERT_OK(Flush(i));` (or `ASSERT_OK(dbfull()->TEST_WaitForFlushMemTable());`) too early (and with the lock available to them), resulting in non-synchronized read and write thus a TSAN data race. - Reproduced by the following, suggested by ajkr: ``` diff --git a/db/db_impl/db_impl_compaction_flush.cc b/db/db_impl/db_impl_compaction_flush.cc index 4ff87c1e4..52492e9cf 100644 --- a/db/db_impl/db_impl_compaction_flush.cc +++ b/db/db_impl/db_impl_compaction_flush.cc @@ -22,7 +22,7 @@ #include "test_util/sync_point.h" #include "util/cast_util.h" #include "util/concurrent_task_limiter_impl.h" namespace ROCKSDB_NAMESPACE { bool DBImpl::EnoughRoomForCompaction( @@ -855,6 +855,7 @@ void DBImpl::NotifyOnFlushCompleted( mutable_cf_options.level0_stop_writes_trigger); // release lock while notifying events mutex_.Unlock(); + bg_cv_.SignalAll(); ``` **Summary:** - Added synchornization between read and write by ` ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->LoadDependency()` mechanism Pull Request resolved: https://github.com/facebook/rocksdb/pull/9528 Test Plan: `./listener_test --gtest_filter=EventListenerTest.MultiCF --gtest_repeat=10` - pre-fix: ``` Repeating all tests (iteration 3) Note: Google Test filter = EventListenerTest.MultiCF [==========] Running 1 test from 1 test case. [----------] Global test environment set-up. [----------] 1 test from EventListenerTest [ RUN ] EventListenerTest.MultiCF ================== WARNING: ThreadSanitizer: data race (pid=3377137) Read of size 8 at 0x7b6000000840 by main thread: #0 std::vector<rocksdb::DB*, std::allocator<rocksdb::DB*> >::size() https://github.com/facebook/rocksdb/issues/1 rocksdb::EventListenerTest_MultiCF_Test::TestBody() db/listener_test.cc:384 (listener_test+0x4bb300) Previous write of size 8 at 0x7b6000000840 by thread T2: #0 void std::vector<rocksdb::DB*, std::allocator<rocksdb::DB*> >::_M_realloc_insert<rocksdb::DB* const&>(__gnu_cxx::__normal_iterator<rocksdb::DB**, std::vector<rocksdb::DB*, std::allocator<rocksdb::DB*> > >, rocksdb::DB* const&) https://github.com/facebook/rocksdb/issues/1 std::vector<rocksdb::DB*, std::allocator<rocksdb::DB*> >::push_back(rocksdb::DB* const&) https://github.com/facebook/rocksdb/issues/2 rocksdb::TestFlushListener::OnFlushCompleted(rocksdb::DB*, rocksdb::FlushJobInfo const&) db/listener_test.cc:255 (listener_test+0x4e820f) ``` - post-fix: `All passed` Reviewed By: ajkr Differential Revision: D34085791 Pulled By: hx235 fbshipit-source-id: f877aa687ea1d5cb6f31ef8c4772625d22868e8b
3 years ago
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->DisableProcessing();
Close();
}
}
rocksdb: switch to gtest Summary: Our existing test notation is very similar to what is used in gtest. It makes it easy to adopt what is different. In this diff I modify existing [[ https://code.google.com/p/googletest/wiki/Primer#Test_Fixtures:_Using_the_Same_Data_Configuration_for_Multiple_Te | test fixture ]] classes to inherit from `testing::Test`. Also for unit tests that use fixture class, `TEST` is replaced with `TEST_F` as required in gtest. There are several custom `main` functions in our existing tests. To make this transition easier, I modify all `main` functions to fallow gtest notation. But eventually we can remove them and use implementation of `main` that gtest provides. ```lang=bash % cat ~/transform #!/bin/sh files=$(git ls-files '*test\.cc') for file in $files do if grep -q "rocksdb::test::RunAllTests()" $file then if grep -Eq '^class \w+Test {' $file then perl -pi -e 's/^(class \w+Test) {/${1}: public testing::Test {/g' $file perl -pi -e 's/^(TEST)/${1}_F/g' $file fi perl -pi -e 's/(int main.*\{)/${1}::testing::InitGoogleTest(&argc, argv);/g' $file perl -pi -e 's/rocksdb::test::RunAllTests/RUN_ALL_TESTS/g' $file fi done % sh ~/transform % make format ``` Second iteration of this diff contains only scripted changes. Third iteration contains manual changes to fix last errors and make it compilable. Test Plan: Build and notice no errors. ```lang=bash % USE_CLANG=1 make check -j55 ``` Tests are still testing. Reviewers: meyering, sdong, rven, igor Reviewed By: igor Subscribers: dhruba, leveldb Differential Revision: https://reviews.facebook.net/D35157
10 years ago
TEST_F(EventListenerTest, MultiDBMultiListeners) {
Options options;
Encryption at rest support Summary: This PR adds support for encrypting data stored by RocksDB when written to disk. It adds an `EncryptedEnv` override of the `Env` class with matching overrides for sequential&random access files. The encryption itself is done through a configurable `EncryptionProvider`. This class creates is asked to create `BlockAccessCipherStream` for a file. This is where the actual encryption/decryption is being done. Currently there is a Counter mode implementation of `BlockAccessCipherStream` with a `ROT13` block cipher (NOTE the `ROT13` is for demo purposes only!!). The Counter operation mode uses an initial counter & random initialization vector (IV). Both are created randomly for each file and stored in a 4K (default size) block that is prefixed to that file. The `EncryptedEnv` implementation is such that clients of the `Env` class do not see this prefix (nor data, nor in filesize). The largest part of the prefix block is also encrypted, and there is room left for implementation specific settings/values/keys in there. To test the encryption, the `DBTestBase` class has been extended to consider a new environment variable called `ENCRYPTED_ENV`. If set, the test will setup a encrypted instance of the `Env` class to use for all tests. Typically you would run it like this: ``` ENCRYPTED_ENV=1 make check_some ``` There is also an added test that checks that some data inserted into the database is or is not "visible" on disk. With `ENCRYPTED_ENV` active it must not find plain text strings, with `ENCRYPTED_ENV` unset, it must find the plain text strings. Closes https://github.com/facebook/rocksdb/pull/2424 Differential Revision: D5322178 Pulled By: sdwilsh fbshipit-source-id: 253b0a9c2c498cc98f580df7f2623cbf7678a27f
7 years ago
options.env = CurrentOptions().env;
#ifdef ROCKSDB_USING_THREAD_STATUS
options.enable_thread_tracking = true;
#endif // ROCKSDB_USING_THREAD_STATUS
options.table_properties_collector_factories.push_back(
std::make_shared<TestPropertiesCollectorFactory>());
std::vector<TestFlushListener*> listeners;
const int kNumDBs = 5;
const int kNumListeners = 10;
for (int i = 0; i < kNumListeners; ++i) {
listeners.emplace_back(new TestFlushListener(options.env, this));
}
std::vector<std::string> cf_names = {
"pikachu", "ilya", "muromec", "dobrynia",
"nikitich", "alyosha", "popovich"};
options.create_if_missing = true;
for (int i = 0; i < kNumListeners; ++i) {
options.listeners.emplace_back(listeners[i]);
}
DBOptions db_opts(options);
ColumnFamilyOptions cf_opts(options);
std::vector<DB*> dbs;
std::vector<std::vector<ColumnFamilyHandle *>> vec_handles;
for (int d = 0; d < kNumDBs; ++d) {
ASSERT_OK(DestroyDB(dbname_ + ToString(d), options));
DB* db;
std::vector<ColumnFamilyHandle*> handles;
ASSERT_OK(DB::Open(options, dbname_ + ToString(d), &db));
for (size_t c = 0; c < cf_names.size(); ++c) {
ColumnFamilyHandle* handle;
ASSERT_OK(db->CreateColumnFamily(cf_opts, cf_names[c], &handle));
handles.push_back(handle);
}
vec_handles.push_back(std::move(handles));
dbs.push_back(db);
}
for (int d = 0; d < kNumDBs; ++d) {
for (size_t c = 0; c < cf_names.size(); ++c) {
ASSERT_OK(dbs[d]->Put(WriteOptions(), vec_handles[d][c],
cf_names[c], cf_names[c]));
}
}
for (size_t c = 0; c < cf_names.size(); ++c) {
for (int d = 0; d < kNumDBs; ++d) {
ASSERT_OK(dbs[d]->Flush(FlushOptions(), vec_handles[d][c]));
ASSERT_OK(
static_cast_with_check<DBImpl>(dbs[d])->TEST_WaitForFlushMemTable());
}
}
Fix test race conditions with OnFlushCompleted() (#9617) Summary: We often see flaky tests due to `DB::Flush()` or `DBImpl::TEST_WaitForFlushMemTable()` not waiting until event listeners complete. For example, https://github.com/facebook/rocksdb/issues/9084, https://github.com/facebook/rocksdb/issues/9400, https://github.com/facebook/rocksdb/issues/9528, plus two new ones this week: "EventListenerTest.OnSingleDBFlushTest" and "DBFlushTest.FireOnFlushCompletedAfterCommittedResult". I ran a `make check` with the below race condition-coercing patch and fixed issues it found besides old BlobDB. ``` diff --git a/db/db_impl/db_impl_compaction_flush.cc b/db/db_impl/db_impl_compaction_flush.cc index 0e1864788..aaba68c4a 100644 --- a/db/db_impl/db_impl_compaction_flush.cc +++ b/db/db_impl/db_impl_compaction_flush.cc @@ -861,6 +861,8 @@ void DBImpl::NotifyOnFlushCompleted( mutable_cf_options.level0_stop_writes_trigger); // release lock while notifying events mutex_.Unlock(); + bg_cv_.SignalAll(); + sleep(1); { for (auto& info : *flush_jobs_info) { info->triggered_writes_slowdown = triggered_writes_slowdown; ``` The reason I did not fix old BlobDB issues is because it appears to have a fundamental (non-test) issue. In particular, it uses an EventListener to keep track of the files. OnFlushCompleted() could be delayed until even after a compaction involving that flushed file completes, causing the compaction to unexpectedly delete an untracked file. Pull Request resolved: https://github.com/facebook/rocksdb/pull/9617 Test Plan: `make check` including the race condition coercing patch Reviewed By: hx235 Differential Revision: D34384022 Pulled By: ajkr fbshipit-source-id: 2652ded39b415277c5d6a628414345223930514e
3 years ago
for (int d = 0; d < kNumDBs; ++d) {
// Ensure background work is fully finished including listener callbacks
// before accessing listener state.
ASSERT_OK(
static_cast_with_check<DBImpl>(dbs[d])->TEST_WaitForBackgroundWork());
}
for (auto* listener : listeners) {
int pos = 0;
for (size_t c = 0; c < cf_names.size(); ++c) {
for (int d = 0; d < kNumDBs; ++d) {
ASSERT_EQ(listener->flushed_dbs_[pos], dbs[d]);
ASSERT_EQ(listener->flushed_column_family_names_[pos], cf_names[c]);
pos++;
}
}
}
for (auto handles : vec_handles) {
for (auto h : handles) {
delete h;
}
handles.clear();
}
vec_handles.clear();
for (auto db : dbs) {
delete db;
}
}
rocksdb: switch to gtest Summary: Our existing test notation is very similar to what is used in gtest. It makes it easy to adopt what is different. In this diff I modify existing [[ https://code.google.com/p/googletest/wiki/Primer#Test_Fixtures:_Using_the_Same_Data_Configuration_for_Multiple_Te | test fixture ]] classes to inherit from `testing::Test`. Also for unit tests that use fixture class, `TEST` is replaced with `TEST_F` as required in gtest. There are several custom `main` functions in our existing tests. To make this transition easier, I modify all `main` functions to fallow gtest notation. But eventually we can remove them and use implementation of `main` that gtest provides. ```lang=bash % cat ~/transform #!/bin/sh files=$(git ls-files '*test\.cc') for file in $files do if grep -q "rocksdb::test::RunAllTests()" $file then if grep -Eq '^class \w+Test {' $file then perl -pi -e 's/^(class \w+Test) {/${1}: public testing::Test {/g' $file perl -pi -e 's/^(TEST)/${1}_F/g' $file fi perl -pi -e 's/(int main.*\{)/${1}::testing::InitGoogleTest(&argc, argv);/g' $file perl -pi -e 's/rocksdb::test::RunAllTests/RUN_ALL_TESTS/g' $file fi done % sh ~/transform % make format ``` Second iteration of this diff contains only scripted changes. Third iteration contains manual changes to fix last errors and make it compilable. Test Plan: Build and notice no errors. ```lang=bash % USE_CLANG=1 make check -j55 ``` Tests are still testing. Reviewers: meyering, sdong, rven, igor Reviewed By: igor Subscribers: dhruba, leveldb Differential Revision: https://reviews.facebook.net/D35157
10 years ago
TEST_F(EventListenerTest, DisableBGCompaction) {
Options options;
Encryption at rest support Summary: This PR adds support for encrypting data stored by RocksDB when written to disk. It adds an `EncryptedEnv` override of the `Env` class with matching overrides for sequential&random access files. The encryption itself is done through a configurable `EncryptionProvider`. This class creates is asked to create `BlockAccessCipherStream` for a file. This is where the actual encryption/decryption is being done. Currently there is a Counter mode implementation of `BlockAccessCipherStream` with a `ROT13` block cipher (NOTE the `ROT13` is for demo purposes only!!). The Counter operation mode uses an initial counter & random initialization vector (IV). Both are created randomly for each file and stored in a 4K (default size) block that is prefixed to that file. The `EncryptedEnv` implementation is such that clients of the `Env` class do not see this prefix (nor data, nor in filesize). The largest part of the prefix block is also encrypted, and there is room left for implementation specific settings/values/keys in there. To test the encryption, the `DBTestBase` class has been extended to consider a new environment variable called `ENCRYPTED_ENV`. If set, the test will setup a encrypted instance of the `Env` class to use for all tests. Typically you would run it like this: ``` ENCRYPTED_ENV=1 make check_some ``` There is also an added test that checks that some data inserted into the database is or is not "visible" on disk. With `ENCRYPTED_ENV` active it must not find plain text strings, with `ENCRYPTED_ENV` unset, it must find the plain text strings. Closes https://github.com/facebook/rocksdb/pull/2424 Differential Revision: D5322178 Pulled By: sdwilsh fbshipit-source-id: 253b0a9c2c498cc98f580df7f2623cbf7678a27f
7 years ago
options.env = CurrentOptions().env;
#ifdef ROCKSDB_USING_THREAD_STATUS
options.enable_thread_tracking = true;
#endif // ROCKSDB_USING_THREAD_STATUS
TestFlushListener* listener = new TestFlushListener(options.env, this);
const int kCompactionTrigger = 1;
const int kSlowdownTrigger = 5;
const int kStopTrigger = 100;
options.level0_file_num_compaction_trigger = kCompactionTrigger;
options.level0_slowdown_writes_trigger = kSlowdownTrigger;
options.level0_stop_writes_trigger = kStopTrigger;
options.max_write_buffer_number = 10;
options.listeners.emplace_back(listener);
// BG compaction is disabled. Number of L0 files will simply keeps
// increasing in this test.
options.compaction_style = kCompactionStyleNone;
options.compression = kNoCompression;
options.write_buffer_size = 100000; // Small write buffer
options.table_properties_collector_factories.push_back(
std::make_shared<TestPropertiesCollectorFactory>());
CreateAndReopenWithCF({"pikachu"}, options);
ColumnFamilyMetaData cf_meta;
db_->GetColumnFamilyMetaData(handles_[1], &cf_meta);
// keep writing until writes are forced to stop.
for (int i = 0; static_cast<int>(cf_meta.file_count) < kSlowdownTrigger * 10;
++i) {
ASSERT_OK(Put(1, ToString(i), std::string(10000, 'x'), WriteOptions()));
FlushOptions fo;
fo.allow_write_stall = true;
ASSERT_OK(db_->Flush(fo, handles_[1]));
db_->GetColumnFamilyMetaData(handles_[1], &cf_meta);
}
Fix test race conditions with OnFlushCompleted() (#9617) Summary: We often see flaky tests due to `DB::Flush()` or `DBImpl::TEST_WaitForFlushMemTable()` not waiting until event listeners complete. For example, https://github.com/facebook/rocksdb/issues/9084, https://github.com/facebook/rocksdb/issues/9400, https://github.com/facebook/rocksdb/issues/9528, plus two new ones this week: "EventListenerTest.OnSingleDBFlushTest" and "DBFlushTest.FireOnFlushCompletedAfterCommittedResult". I ran a `make check` with the below race condition-coercing patch and fixed issues it found besides old BlobDB. ``` diff --git a/db/db_impl/db_impl_compaction_flush.cc b/db/db_impl/db_impl_compaction_flush.cc index 0e1864788..aaba68c4a 100644 --- a/db/db_impl/db_impl_compaction_flush.cc +++ b/db/db_impl/db_impl_compaction_flush.cc @@ -861,6 +861,8 @@ void DBImpl::NotifyOnFlushCompleted( mutable_cf_options.level0_stop_writes_trigger); // release lock while notifying events mutex_.Unlock(); + bg_cv_.SignalAll(); + sleep(1); { for (auto& info : *flush_jobs_info) { info->triggered_writes_slowdown = triggered_writes_slowdown; ``` The reason I did not fix old BlobDB issues is because it appears to have a fundamental (non-test) issue. In particular, it uses an EventListener to keep track of the files. OnFlushCompleted() could be delayed until even after a compaction involving that flushed file completes, causing the compaction to unexpectedly delete an untracked file. Pull Request resolved: https://github.com/facebook/rocksdb/pull/9617 Test Plan: `make check` including the race condition coercing patch Reviewed By: hx235 Differential Revision: D34384022 Pulled By: ajkr fbshipit-source-id: 2652ded39b415277c5d6a628414345223930514e
3 years ago
// Ensure background work is fully finished including listener callbacks
// before accessing listener state.
ASSERT_OK(dbfull()->TEST_WaitForBackgroundWork());
Fix test race conditions with OnFlushCompleted() (#9617) Summary: We often see flaky tests due to `DB::Flush()` or `DBImpl::TEST_WaitForFlushMemTable()` not waiting until event listeners complete. For example, https://github.com/facebook/rocksdb/issues/9084, https://github.com/facebook/rocksdb/issues/9400, https://github.com/facebook/rocksdb/issues/9528, plus two new ones this week: "EventListenerTest.OnSingleDBFlushTest" and "DBFlushTest.FireOnFlushCompletedAfterCommittedResult". I ran a `make check` with the below race condition-coercing patch and fixed issues it found besides old BlobDB. ``` diff --git a/db/db_impl/db_impl_compaction_flush.cc b/db/db_impl/db_impl_compaction_flush.cc index 0e1864788..aaba68c4a 100644 --- a/db/db_impl/db_impl_compaction_flush.cc +++ b/db/db_impl/db_impl_compaction_flush.cc @@ -861,6 +861,8 @@ void DBImpl::NotifyOnFlushCompleted( mutable_cf_options.level0_stop_writes_trigger); // release lock while notifying events mutex_.Unlock(); + bg_cv_.SignalAll(); + sleep(1); { for (auto& info : *flush_jobs_info) { info->triggered_writes_slowdown = triggered_writes_slowdown; ``` The reason I did not fix old BlobDB issues is because it appears to have a fundamental (non-test) issue. In particular, it uses an EventListener to keep track of the files. OnFlushCompleted() could be delayed until even after a compaction involving that flushed file completes, causing the compaction to unexpectedly delete an untracked file. Pull Request resolved: https://github.com/facebook/rocksdb/pull/9617 Test Plan: `make check` including the race condition coercing patch Reviewed By: hx235 Differential Revision: D34384022 Pulled By: ajkr fbshipit-source-id: 2652ded39b415277c5d6a628414345223930514e
3 years ago
ASSERT_GE(listener->slowdown_count, kSlowdownTrigger * 9);
}
class TestCompactionReasonListener : public EventListener {
public:
void OnCompactionCompleted(DB* /*db*/, const CompactionJobInfo& ci) override {
std::lock_guard<std::mutex> lock(mutex_);
compaction_reasons_.push_back(ci.compaction_reason);
}
std::vector<CompactionReason> compaction_reasons_;
std::mutex mutex_;
};
TEST_F(EventListenerTest, CompactionReasonLevel) {
Options options;
Encryption at rest support Summary: This PR adds support for encrypting data stored by RocksDB when written to disk. It adds an `EncryptedEnv` override of the `Env` class with matching overrides for sequential&random access files. The encryption itself is done through a configurable `EncryptionProvider`. This class creates is asked to create `BlockAccessCipherStream` for a file. This is where the actual encryption/decryption is being done. Currently there is a Counter mode implementation of `BlockAccessCipherStream` with a `ROT13` block cipher (NOTE the `ROT13` is for demo purposes only!!). The Counter operation mode uses an initial counter & random initialization vector (IV). Both are created randomly for each file and stored in a 4K (default size) block that is prefixed to that file. The `EncryptedEnv` implementation is such that clients of the `Env` class do not see this prefix (nor data, nor in filesize). The largest part of the prefix block is also encrypted, and there is room left for implementation specific settings/values/keys in there. To test the encryption, the `DBTestBase` class has been extended to consider a new environment variable called `ENCRYPTED_ENV`. If set, the test will setup a encrypted instance of the `Env` class to use for all tests. Typically you would run it like this: ``` ENCRYPTED_ENV=1 make check_some ``` There is also an added test that checks that some data inserted into the database is or is not "visible" on disk. With `ENCRYPTED_ENV` active it must not find plain text strings, with `ENCRYPTED_ENV` unset, it must find the plain text strings. Closes https://github.com/facebook/rocksdb/pull/2424 Differential Revision: D5322178 Pulled By: sdwilsh fbshipit-source-id: 253b0a9c2c498cc98f580df7f2623cbf7678a27f
7 years ago
options.env = CurrentOptions().env;
options.create_if_missing = true;
options.memtable_factory.reset(test::NewSpecialSkipListFactory(
DBTestBase::kNumKeysByGenerateNewRandomFile));
TestCompactionReasonListener* listener = new TestCompactionReasonListener();
options.listeners.emplace_back(listener);
options.level0_file_num_compaction_trigger = 4;
options.compaction_style = kCompactionStyleLevel;
DestroyAndReopen(options);
Random rnd(301);
// Write 4 files in L0
for (int i = 0; i < 4; i++) {
GenerateNewRandomFile(&rnd);
}
ASSERT_OK(dbfull()->TEST_WaitForCompact());
ASSERT_EQ(listener->compaction_reasons_.size(), 1);
ASSERT_EQ(listener->compaction_reasons_[0],
CompactionReason::kLevelL0FilesNum);
DestroyAndReopen(options);
// Write 3 non-overlapping files in L0
for (int k = 1; k <= 30; k++) {
ASSERT_OK(Put(Key(k), Key(k)));
if (k % 10 == 0) {
Flush();
}
}
// Do a trivial move from L0 -> L1
ASSERT_OK(db_->CompactRange(CompactRangeOptions(), nullptr, nullptr));
options.max_bytes_for_level_base = 1;
Close();
listener->compaction_reasons_.clear();
Reopen(options);
ASSERT_OK(dbfull()->TEST_WaitForCompact());
ASSERT_GT(listener->compaction_reasons_.size(), 1);
for (auto compaction_reason : listener->compaction_reasons_) {
ASSERT_EQ(compaction_reason, CompactionReason::kLevelMaxLevelSize);
}
options.disable_auto_compactions = true;
Close();
listener->compaction_reasons_.clear();
Reopen(options);
ASSERT_OK(Put("key", "value"));
CompactRangeOptions cro;
cro.bottommost_level_compaction = BottommostLevelCompaction::kForceOptimized;
ASSERT_OK(db_->CompactRange(cro, nullptr, nullptr));
ASSERT_GT(listener->compaction_reasons_.size(), 0);
for (auto compaction_reason : listener->compaction_reasons_) {
ASSERT_EQ(compaction_reason, CompactionReason::kManualCompaction);
}
}
TEST_F(EventListenerTest, CompactionReasonUniversal) {
Options options;
Encryption at rest support Summary: This PR adds support for encrypting data stored by RocksDB when written to disk. It adds an `EncryptedEnv` override of the `Env` class with matching overrides for sequential&random access files. The encryption itself is done through a configurable `EncryptionProvider`. This class creates is asked to create `BlockAccessCipherStream` for a file. This is where the actual encryption/decryption is being done. Currently there is a Counter mode implementation of `BlockAccessCipherStream` with a `ROT13` block cipher (NOTE the `ROT13` is for demo purposes only!!). The Counter operation mode uses an initial counter & random initialization vector (IV). Both are created randomly for each file and stored in a 4K (default size) block that is prefixed to that file. The `EncryptedEnv` implementation is such that clients of the `Env` class do not see this prefix (nor data, nor in filesize). The largest part of the prefix block is also encrypted, and there is room left for implementation specific settings/values/keys in there. To test the encryption, the `DBTestBase` class has been extended to consider a new environment variable called `ENCRYPTED_ENV`. If set, the test will setup a encrypted instance of the `Env` class to use for all tests. Typically you would run it like this: ``` ENCRYPTED_ENV=1 make check_some ``` There is also an added test that checks that some data inserted into the database is or is not "visible" on disk. With `ENCRYPTED_ENV` active it must not find plain text strings, with `ENCRYPTED_ENV` unset, it must find the plain text strings. Closes https://github.com/facebook/rocksdb/pull/2424 Differential Revision: D5322178 Pulled By: sdwilsh fbshipit-source-id: 253b0a9c2c498cc98f580df7f2623cbf7678a27f
7 years ago
options.env = CurrentOptions().env;
options.create_if_missing = true;
options.memtable_factory.reset(test::NewSpecialSkipListFactory(
DBTestBase::kNumKeysByGenerateNewRandomFile));
TestCompactionReasonListener* listener = new TestCompactionReasonListener();
options.listeners.emplace_back(listener);
options.compaction_style = kCompactionStyleUniversal;
Random rnd(301);
options.level0_file_num_compaction_trigger = 8;
options.compaction_options_universal.max_size_amplification_percent = 100000;
options.compaction_options_universal.size_ratio = 100000;
DestroyAndReopen(options);
listener->compaction_reasons_.clear();
// Write 8 files in L0
for (int i = 0; i < 8; i++) {
GenerateNewRandomFile(&rnd);
}
ASSERT_OK(dbfull()->TEST_WaitForCompact());
ASSERT_GT(listener->compaction_reasons_.size(), 0);
for (auto compaction_reason : listener->compaction_reasons_) {
ASSERT_EQ(compaction_reason, CompactionReason::kUniversalSizeRatio);
}
options.level0_file_num_compaction_trigger = 8;
options.compaction_options_universal.max_size_amplification_percent = 1;
options.compaction_options_universal.size_ratio = 100000;
DestroyAndReopen(options);
listener->compaction_reasons_.clear();
// Write 8 files in L0
for (int i = 0; i < 8; i++) {
GenerateNewRandomFile(&rnd);
}
ASSERT_OK(dbfull()->TEST_WaitForCompact());
ASSERT_GT(listener->compaction_reasons_.size(), 0);
for (auto compaction_reason : listener->compaction_reasons_) {
ASSERT_EQ(compaction_reason, CompactionReason::kUniversalSizeAmplification);
}
options.disable_auto_compactions = true;
Close();
listener->compaction_reasons_.clear();
Reopen(options);
ASSERT_OK(db_->CompactRange(CompactRangeOptions(), nullptr, nullptr));
ASSERT_GT(listener->compaction_reasons_.size(), 0);
for (auto compaction_reason : listener->compaction_reasons_) {
ASSERT_EQ(compaction_reason, CompactionReason::kManualCompaction);
}
}
TEST_F(EventListenerTest, CompactionReasonFIFO) {
Options options;
Encryption at rest support Summary: This PR adds support for encrypting data stored by RocksDB when written to disk. It adds an `EncryptedEnv` override of the `Env` class with matching overrides for sequential&random access files. The encryption itself is done through a configurable `EncryptionProvider`. This class creates is asked to create `BlockAccessCipherStream` for a file. This is where the actual encryption/decryption is being done. Currently there is a Counter mode implementation of `BlockAccessCipherStream` with a `ROT13` block cipher (NOTE the `ROT13` is for demo purposes only!!). The Counter operation mode uses an initial counter & random initialization vector (IV). Both are created randomly for each file and stored in a 4K (default size) block that is prefixed to that file. The `EncryptedEnv` implementation is such that clients of the `Env` class do not see this prefix (nor data, nor in filesize). The largest part of the prefix block is also encrypted, and there is room left for implementation specific settings/values/keys in there. To test the encryption, the `DBTestBase` class has been extended to consider a new environment variable called `ENCRYPTED_ENV`. If set, the test will setup a encrypted instance of the `Env` class to use for all tests. Typically you would run it like this: ``` ENCRYPTED_ENV=1 make check_some ``` There is also an added test that checks that some data inserted into the database is or is not "visible" on disk. With `ENCRYPTED_ENV` active it must not find plain text strings, with `ENCRYPTED_ENV` unset, it must find the plain text strings. Closes https://github.com/facebook/rocksdb/pull/2424 Differential Revision: D5322178 Pulled By: sdwilsh fbshipit-source-id: 253b0a9c2c498cc98f580df7f2623cbf7678a27f
7 years ago
options.env = CurrentOptions().env;
options.create_if_missing = true;
options.memtable_factory.reset(test::NewSpecialSkipListFactory(
DBTestBase::kNumKeysByGenerateNewRandomFile));
TestCompactionReasonListener* listener = new TestCompactionReasonListener();
options.listeners.emplace_back(listener);
options.level0_file_num_compaction_trigger = 4;
options.compaction_style = kCompactionStyleFIFO;
options.compaction_options_fifo.max_table_files_size = 1;
DestroyAndReopen(options);
Random rnd(301);
// Write 4 files in L0
for (int i = 0; i < 4; i++) {
GenerateNewRandomFile(&rnd);
}
ASSERT_OK(dbfull()->TEST_WaitForCompact());
ASSERT_GT(listener->compaction_reasons_.size(), 0);
for (auto compaction_reason : listener->compaction_reasons_) {
ASSERT_EQ(compaction_reason, CompactionReason::kFIFOMaxSize);
}
}
class TableFileCreationListener : public EventListener {
public:
class TestEnv : public EnvWrapper {
public:
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566) Summary: This PR does a few things: 1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation. 2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed: - The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated - The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory). 3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10). I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged. Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently. Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :) Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566 Reviewed By: zhichao-cao Differential Revision: D24408980 Pulled By: jay-zhuang fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
4 years ago
explicit TestEnv(Env* t) : EnvWrapper(t) {}
static const char* kClassName() { return "TestEnv"; }
const char* Name() const override { return kClassName(); }
void SetStatus(Status s) { status_ = s; }
Status NewWritableFile(const std::string& fname,
std::unique_ptr<WritableFile>* result,
const EnvOptions& options) override {
if (fname.size() > 4 && fname.substr(fname.size() - 4) == ".sst") {
if (!status_.ok()) {
return status_;
}
}
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566) Summary: This PR does a few things: 1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation. 2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed: - The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated - The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory). 3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10). I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged. Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently. Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :) Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566 Reviewed By: zhichao-cao Differential Revision: D24408980 Pulled By: jay-zhuang fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
4 years ago
return target()->NewWritableFile(fname, result, options);
}
private:
Status status_;
};
TableFileCreationListener() {
for (int i = 0; i < 2; i++) {
started_[i] = finished_[i] = failure_[i] = 0;
}
}
int Index(TableFileCreationReason reason) {
int idx;
switch (reason) {
case TableFileCreationReason::kFlush:
idx = 0;
break;
case TableFileCreationReason::kCompaction:
idx = 1;
break;
default:
idx = -1;
}
return idx;
}
void CheckAndResetCounters(int flush_started, int flush_finished,
int flush_failure, int compaction_started,
int compaction_finished, int compaction_failure) {
ASSERT_EQ(started_[0], flush_started);
ASSERT_EQ(finished_[0], flush_finished);
ASSERT_EQ(failure_[0], flush_failure);
ASSERT_EQ(started_[1], compaction_started);
ASSERT_EQ(finished_[1], compaction_finished);
ASSERT_EQ(failure_[1], compaction_failure);
for (int i = 0; i < 2; i++) {
started_[i] = finished_[i] = failure_[i] = 0;
}
}
void OnTableFileCreationStarted(
const TableFileCreationBriefInfo& info) override {
int idx = Index(info.reason);
if (idx >= 0) {
started_[idx]++;
}
ASSERT_GT(info.db_name.size(), 0U);
ASSERT_GT(info.cf_name.size(), 0U);
ASSERT_GT(info.file_path.size(), 0U);
ASSERT_GT(info.job_id, 0);
}
void OnTableFileCreated(const TableFileCreationInfo& info) override {
int idx = Index(info.reason);
if (idx >= 0) {
finished_[idx]++;
}
ASSERT_GT(info.db_name.size(), 0U);
ASSERT_GT(info.cf_name.size(), 0U);
ASSERT_GT(info.file_path.size(), 0U);
ASSERT_GT(info.job_id, 0);
ASSERT_EQ(info.file_checksum, kUnknownFileChecksum);
ASSERT_EQ(info.file_checksum_func_name, kUnknownFileChecksumFuncName);
if (info.status.ok()) {
if (info.table_properties.num_range_deletions == 0U) {
ASSERT_GT(info.table_properties.data_size, 0U);
ASSERT_GT(info.table_properties.raw_key_size, 0U);
ASSERT_GT(info.table_properties.raw_value_size, 0U);
ASSERT_GT(info.table_properties.num_data_blocks, 0U);
ASSERT_GT(info.table_properties.num_entries, 0U);
}
} else {
if (idx >= 0) {
failure_[idx]++;
last_failure_ = info.status;
}
}
}
int started_[2];
int finished_[2];
int failure_[2];
Status last_failure_;
};
TEST_F(EventListenerTest, TableFileCreationListenersTest) {
auto listener = std::make_shared<TableFileCreationListener>();
Options options;
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566) Summary: This PR does a few things: 1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation. 2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed: - The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated - The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory). 3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10). I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged. Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently. Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :) Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566 Reviewed By: zhichao-cao Differential Revision: D24408980 Pulled By: jay-zhuang fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
4 years ago
std::unique_ptr<TableFileCreationListener::TestEnv> test_env(
new TableFileCreationListener::TestEnv(CurrentOptions().env));
options.create_if_missing = true;
options.listeners.push_back(listener);
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566) Summary: This PR does a few things: 1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation. 2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed: - The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated - The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory). 3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10). I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged. Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently. Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :) Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566 Reviewed By: zhichao-cao Differential Revision: D24408980 Pulled By: jay-zhuang fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
4 years ago
options.env = test_env.get();
DestroyAndReopen(options);
ASSERT_OK(Put("foo", "aaa"));
ASSERT_OK(Put("bar", "bbb"));
ASSERT_OK(Flush());
ASSERT_OK(dbfull()->TEST_WaitForFlushMemTable());
listener->CheckAndResetCounters(1, 1, 0, 0, 0, 0);
ASSERT_OK(Put("foo", "aaa1"));
ASSERT_OK(Put("bar", "bbb1"));
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566) Summary: This PR does a few things: 1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation. 2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed: - The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated - The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory). 3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10). I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged. Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently. Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :) Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566 Reviewed By: zhichao-cao Differential Revision: D24408980 Pulled By: jay-zhuang fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
4 years ago
test_env->SetStatus(Status::NotSupported("not supported"));
ASSERT_NOK(Flush());
listener->CheckAndResetCounters(1, 1, 1, 0, 0, 0);
ASSERT_TRUE(listener->last_failure_.IsNotSupported());
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566) Summary: This PR does a few things: 1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation. 2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed: - The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated - The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory). 3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10). I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged. Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently. Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :) Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566 Reviewed By: zhichao-cao Differential Revision: D24408980 Pulled By: jay-zhuang fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
4 years ago
test_env->SetStatus(Status::OK());
Reopen(options);
ASSERT_OK(Put("foo", "aaa2"));
ASSERT_OK(Put("bar", "bbb2"));
ASSERT_OK(Flush());
ASSERT_OK(dbfull()->TEST_WaitForFlushMemTable());
listener->CheckAndResetCounters(1, 1, 0, 0, 0, 0);
const Slice kRangeStart = "a";
const Slice kRangeEnd = "z";
ASSERT_OK(
dbfull()->CompactRange(CompactRangeOptions(), &kRangeStart, &kRangeEnd));
ASSERT_OK(dbfull()->TEST_WaitForCompact());
listener->CheckAndResetCounters(0, 0, 0, 1, 1, 0);
ASSERT_OK(Put("foo", "aaa3"));
ASSERT_OK(Put("bar", "bbb3"));
ASSERT_OK(Flush());
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566) Summary: This PR does a few things: 1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation. 2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed: - The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated - The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory). 3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10). I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged. Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently. Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :) Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566 Reviewed By: zhichao-cao Differential Revision: D24408980 Pulled By: jay-zhuang fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
4 years ago
test_env->SetStatus(Status::NotSupported("not supported"));
ASSERT_NOK(
dbfull()->CompactRange(CompactRangeOptions(), &kRangeStart, &kRangeEnd));
ASSERT_NOK(dbfull()->TEST_WaitForCompact());
listener->CheckAndResetCounters(1, 1, 0, 1, 1, 1);
ASSERT_TRUE(listener->last_failure_.IsNotSupported());
// Reset
test_env->SetStatus(Status::OK());
DestroyAndReopen(options);
// Verify that an empty table file that is immediately deleted gives Aborted
// status to listener.
ASSERT_OK(Put("baz", "z"));
ASSERT_OK(SingleDelete("baz"));
ASSERT_OK(Flush());
listener->CheckAndResetCounters(1, 1, 1, 0, 0, 0);
ASSERT_TRUE(listener->last_failure_.IsAborted());
// Also in compaction
ASSERT_OK(Put("baz", "z"));
ASSERT_OK(Flush());
ASSERT_OK(db_->DeleteRange(WriteOptions(), db_->DefaultColumnFamily(),
kRangeStart, kRangeEnd));
ASSERT_OK(Flush());
ASSERT_OK(db_->CompactRange(CompactRangeOptions(), nullptr, nullptr));
ASSERT_OK(dbfull()->TEST_WaitForCompact());
listener->CheckAndResetCounters(2, 2, 0, 1, 1, 1);
ASSERT_TRUE(listener->last_failure_.IsAborted());
Close(); // Avoid UAF on listener
}
class MemTableSealedListener : public EventListener {
private:
SequenceNumber latest_seq_number_;
public:
MemTableSealedListener() {}
void OnMemTableSealed(const MemTableInfo& info) override {
latest_seq_number_ = info.first_seqno;
}
void OnFlushCompleted(DB* /*db*/,
const FlushJobInfo& flush_job_info) override {
ASSERT_LE(flush_job_info.smallest_seqno, latest_seq_number_);
}
};
TEST_F(EventListenerTest, MemTableSealedListenerTest) {
auto listener = std::make_shared<MemTableSealedListener>();
Options options;
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566) Summary: This PR does a few things: 1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation. 2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed: - The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated - The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory). 3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10). I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged. Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently. Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :) Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566 Reviewed By: zhichao-cao Differential Revision: D24408980 Pulled By: jay-zhuang fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
4 years ago
options.env = CurrentOptions().env;
options.create_if_missing = true;
options.listeners.push_back(listener);
DestroyAndReopen(options);
for (unsigned int i = 0; i < 10; i++) {
std::string tag = std::to_string(i);
ASSERT_OK(Put("foo"+tag, "aaa"));
ASSERT_OK(Put("bar"+tag, "bbb"));
ASSERT_OK(Flush());
}
}
class ColumnFamilyHandleDeletionStartedListener : public EventListener {
private:
std::vector<std::string> cfs_;
int counter;
public:
explicit ColumnFamilyHandleDeletionStartedListener(
const std::vector<std::string>& cfs)
: cfs_(cfs), counter(0) {
cfs_.insert(cfs_.begin(), kDefaultColumnFamilyName);
}
void OnColumnFamilyHandleDeletionStarted(
ColumnFamilyHandle* handle) override {
ASSERT_EQ(cfs_[handle->GetID()], handle->GetName());
counter++;
}
int getCounter() { return counter; }
};
TEST_F(EventListenerTest, ColumnFamilyHandleDeletionStartedListenerTest) {
std::vector<std::string> cfs{"pikachu", "eevee", "Mewtwo"};
auto listener =
std::make_shared<ColumnFamilyHandleDeletionStartedListener>(cfs);
Options options;
Encryption at rest support Summary: This PR adds support for encrypting data stored by RocksDB when written to disk. It adds an `EncryptedEnv` override of the `Env` class with matching overrides for sequential&random access files. The encryption itself is done through a configurable `EncryptionProvider`. This class creates is asked to create `BlockAccessCipherStream` for a file. This is where the actual encryption/decryption is being done. Currently there is a Counter mode implementation of `BlockAccessCipherStream` with a `ROT13` block cipher (NOTE the `ROT13` is for demo purposes only!!). The Counter operation mode uses an initial counter & random initialization vector (IV). Both are created randomly for each file and stored in a 4K (default size) block that is prefixed to that file. The `EncryptedEnv` implementation is such that clients of the `Env` class do not see this prefix (nor data, nor in filesize). The largest part of the prefix block is also encrypted, and there is room left for implementation specific settings/values/keys in there. To test the encryption, the `DBTestBase` class has been extended to consider a new environment variable called `ENCRYPTED_ENV`. If set, the test will setup a encrypted instance of the `Env` class to use for all tests. Typically you would run it like this: ``` ENCRYPTED_ENV=1 make check_some ``` There is also an added test that checks that some data inserted into the database is or is not "visible" on disk. With `ENCRYPTED_ENV` active it must not find plain text strings, with `ENCRYPTED_ENV` unset, it must find the plain text strings. Closes https://github.com/facebook/rocksdb/pull/2424 Differential Revision: D5322178 Pulled By: sdwilsh fbshipit-source-id: 253b0a9c2c498cc98f580df7f2623cbf7678a27f
7 years ago
options.env = CurrentOptions().env;
options.create_if_missing = true;
options.listeners.push_back(listener);
CreateAndReopenWithCF(cfs, options);
ASSERT_EQ(handles_.size(), 4);
delete handles_[3];
delete handles_[2];
delete handles_[1];
handles_.resize(1);
ASSERT_EQ(listener->getCounter(), 3);
}
class BackgroundErrorListener : public EventListener {
private:
SpecialEnv* env_;
int counter_;
public:
BackgroundErrorListener(SpecialEnv* env) : env_(env), counter_(0) {}
void OnBackgroundError(BackgroundErrorReason /*reason*/,
Status* bg_error) override {
if (counter_ == 0) {
// suppress the first error and disable write-dropping such that a retry
// can succeed.
*bg_error = Status::OK();
env_->drop_writes_.store(false, std::memory_order_release);
Fix+clean up handling of mock sleeps (#7101) Summary: We have a number of tests hanging on MacOS and windows due to mishandling of code for mock sleeps. In addition, the code was in terrible shape because the same variable (addon_time_) would sometimes refer to microseconds and sometimes to seconds. One test even assumed it was nanoseconds but was written to pass anyway. This has been cleaned up so that DB tests generally use a SpecialEnv function to mock sleep, for either some number of microseconds or seconds depending on the function called. But to call one of these, the test must first call SetMockSleep (precondition enforced with assertion), which also turns sleeps in RocksDB into mock sleeps. To also removes accounting for actual clock time, call SetTimeElapseOnlySleepOnReopen, which implies SetMockSleep (on DB re-open). This latter setting only works by applying on DB re-open, otherwise havoc can ensue if Env goes back in time with DB open. More specifics: Removed some unused test classes, and updated comments on the general problem. Fixed DBSSTTest.GetTotalSstFilesSize using a sync point callback instead of mock time. For this we have the only modification to production code, inserting a sync point callback in flush_job.cc, which is not a change to production behavior. Removed unnecessary resetting of mock times to 0 in many tests. RocksDB deals in relative time. Any behaviors relying on absolute date/time are likely a bug. (The above test DBSSTTest.GetTotalSstFilesSize was the only one clearly injecting a specific absolute time for actual testing convenience.) Just in case I misunderstood some test, I put this note in each replacement: // NOTE: Presumed unnecessary and removed: resetting mock time in env Strengthened some tests like MergeTestTime, MergeCompactionTimeTest, and FilterCompactionTimeTest in db_test.cc stats_history_test and blob_db_test are each their own beast, rather deeply dependent on MockTimeEnv. Each gets its own variant of a work-around for TimedWait in a mock time environment. (Reduces redundancy and inconsistency in stats_history_test.) Intended follow-up: Remove TimedWait from the public API of InstrumentedCondVar, and only make that accessible through Env by passing in an InstrumentedCondVar and a deadline. Then the Env implementations mocking time can fix this problem without using sync points. (Test infrastructure using sync points interferes with individual tests' control over sync points.) With that change, we can simplify/consolidate the scattered work-arounds. Pull Request resolved: https://github.com/facebook/rocksdb/pull/7101 Test Plan: make check on Linux and MacOS Reviewed By: zhichao-cao Differential Revision: D23032815 Pulled By: pdillinger fbshipit-source-id: 7f33967ada8b83011fb54e8279365c008bd6610b
4 years ago
env_->SetMockSleep(false);
}
++counter_;
}
int counter() { return counter_; }
};
TEST_F(EventListenerTest, BackgroundErrorListenerFailedFlushTest) {
auto listener = std::make_shared<BackgroundErrorListener>(env_);
Options options;
options.create_if_missing = true;
options.env = env_;
options.listeners.push_back(listener);
options.memtable_factory.reset(test::NewSpecialSkipListFactory(1));
options.paranoid_checks = true;
DestroyAndReopen(options);
// the usual TEST_WaitForFlushMemTable() doesn't work for failed flushes, so
// forge a custom one for the failed flush case.
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->LoadDependency(
{{"DBImpl::BGWorkFlush:done",
"EventListenerTest:BackgroundErrorListenerFailedFlushTest:1"}});
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->EnableProcessing();
env_->drop_writes_.store(true, std::memory_order_release);
Fix+clean up handling of mock sleeps (#7101) Summary: We have a number of tests hanging on MacOS and windows due to mishandling of code for mock sleeps. In addition, the code was in terrible shape because the same variable (addon_time_) would sometimes refer to microseconds and sometimes to seconds. One test even assumed it was nanoseconds but was written to pass anyway. This has been cleaned up so that DB tests generally use a SpecialEnv function to mock sleep, for either some number of microseconds or seconds depending on the function called. But to call one of these, the test must first call SetMockSleep (precondition enforced with assertion), which also turns sleeps in RocksDB into mock sleeps. To also removes accounting for actual clock time, call SetTimeElapseOnlySleepOnReopen, which implies SetMockSleep (on DB re-open). This latter setting only works by applying on DB re-open, otherwise havoc can ensue if Env goes back in time with DB open. More specifics: Removed some unused test classes, and updated comments on the general problem. Fixed DBSSTTest.GetTotalSstFilesSize using a sync point callback instead of mock time. For this we have the only modification to production code, inserting a sync point callback in flush_job.cc, which is not a change to production behavior. Removed unnecessary resetting of mock times to 0 in many tests. RocksDB deals in relative time. Any behaviors relying on absolute date/time are likely a bug. (The above test DBSSTTest.GetTotalSstFilesSize was the only one clearly injecting a specific absolute time for actual testing convenience.) Just in case I misunderstood some test, I put this note in each replacement: // NOTE: Presumed unnecessary and removed: resetting mock time in env Strengthened some tests like MergeTestTime, MergeCompactionTimeTest, and FilterCompactionTimeTest in db_test.cc stats_history_test and blob_db_test are each their own beast, rather deeply dependent on MockTimeEnv. Each gets its own variant of a work-around for TimedWait in a mock time environment. (Reduces redundancy and inconsistency in stats_history_test.) Intended follow-up: Remove TimedWait from the public API of InstrumentedCondVar, and only make that accessible through Env by passing in an InstrumentedCondVar and a deadline. Then the Env implementations mocking time can fix this problem without using sync points. (Test infrastructure using sync points interferes with individual tests' control over sync points.) With that change, we can simplify/consolidate the scattered work-arounds. Pull Request resolved: https://github.com/facebook/rocksdb/pull/7101 Test Plan: make check on Linux and MacOS Reviewed By: zhichao-cao Differential Revision: D23032815 Pulled By: pdillinger fbshipit-source-id: 7f33967ada8b83011fb54e8279365c008bd6610b
4 years ago
env_->SetMockSleep();
ASSERT_OK(Put("key0", "val"));
ASSERT_OK(Put("key1", "val"));
TEST_SYNC_POINT("EventListenerTest:BackgroundErrorListenerFailedFlushTest:1");
ASSERT_EQ(1, listener->counter());
ASSERT_OK(Put("key2", "val"));
ASSERT_OK(dbfull()->TEST_WaitForFlushMemTable());
ASSERT_EQ(1, NumTableFilesAtLevel(0));
}
TEST_F(EventListenerTest, BackgroundErrorListenerFailedCompactionTest) {
auto listener = std::make_shared<BackgroundErrorListener>(env_);
Options options;
options.create_if_missing = true;
options.disable_auto_compactions = true;
options.env = env_;
options.level0_file_num_compaction_trigger = 2;
options.listeners.push_back(listener);
options.memtable_factory.reset(test::NewSpecialSkipListFactory(2));
options.paranoid_checks = true;
DestroyAndReopen(options);
// third iteration triggers the second memtable's flush
for (int i = 0; i < 3; ++i) {
ASSERT_OK(Put("key0", "val"));
if (i > 0) {
ASSERT_OK(dbfull()->TEST_WaitForFlushMemTable());
}
ASSERT_OK(Put("key1", "val"));
}
ASSERT_EQ(2, NumTableFilesAtLevel(0));
env_->drop_writes_.store(true, std::memory_order_release);
Fix+clean up handling of mock sleeps (#7101) Summary: We have a number of tests hanging on MacOS and windows due to mishandling of code for mock sleeps. In addition, the code was in terrible shape because the same variable (addon_time_) would sometimes refer to microseconds and sometimes to seconds. One test even assumed it was nanoseconds but was written to pass anyway. This has been cleaned up so that DB tests generally use a SpecialEnv function to mock sleep, for either some number of microseconds or seconds depending on the function called. But to call one of these, the test must first call SetMockSleep (precondition enforced with assertion), which also turns sleeps in RocksDB into mock sleeps. To also removes accounting for actual clock time, call SetTimeElapseOnlySleepOnReopen, which implies SetMockSleep (on DB re-open). This latter setting only works by applying on DB re-open, otherwise havoc can ensue if Env goes back in time with DB open. More specifics: Removed some unused test classes, and updated comments on the general problem. Fixed DBSSTTest.GetTotalSstFilesSize using a sync point callback instead of mock time. For this we have the only modification to production code, inserting a sync point callback in flush_job.cc, which is not a change to production behavior. Removed unnecessary resetting of mock times to 0 in many tests. RocksDB deals in relative time. Any behaviors relying on absolute date/time are likely a bug. (The above test DBSSTTest.GetTotalSstFilesSize was the only one clearly injecting a specific absolute time for actual testing convenience.) Just in case I misunderstood some test, I put this note in each replacement: // NOTE: Presumed unnecessary and removed: resetting mock time in env Strengthened some tests like MergeTestTime, MergeCompactionTimeTest, and FilterCompactionTimeTest in db_test.cc stats_history_test and blob_db_test are each their own beast, rather deeply dependent on MockTimeEnv. Each gets its own variant of a work-around for TimedWait in a mock time environment. (Reduces redundancy and inconsistency in stats_history_test.) Intended follow-up: Remove TimedWait from the public API of InstrumentedCondVar, and only make that accessible through Env by passing in an InstrumentedCondVar and a deadline. Then the Env implementations mocking time can fix this problem without using sync points. (Test infrastructure using sync points interferes with individual tests' control over sync points.) With that change, we can simplify/consolidate the scattered work-arounds. Pull Request resolved: https://github.com/facebook/rocksdb/pull/7101 Test Plan: make check on Linux and MacOS Reviewed By: zhichao-cao Differential Revision: D23032815 Pulled By: pdillinger fbshipit-source-id: 7f33967ada8b83011fb54e8279365c008bd6610b
4 years ago
env_->SetMockSleep();
ASSERT_OK(dbfull()->SetOptions({{"disable_auto_compactions", "false"}}));
ASSERT_OK(dbfull()->TEST_WaitForCompact());
ASSERT_EQ(1, listener->counter());
// trigger flush so compaction is triggered again; this time it succeeds
Auto recovery from out of space errors (#4164) Summary: This commit implements automatic recovery from a Status::NoSpace() error during background operations such as write callback, flush and compaction. The broad design is as follows - 1. Compaction errors are treated as soft errors and don't put the database in read-only mode. A compaction is delayed until enough free disk space is available to accomodate the compaction outputs, which is estimated based on the input size. This means that users can continue to write, and we rely on the WriteController to delay or stop writes if the compaction debt becomes too high due to persistent low disk space condition 2. Errors during write callback and flush are treated as hard errors, i.e the database is put in read-only mode and goes back to read-write only fater certain recovery actions are taken. 3. Both types of recovery rely on the SstFileManagerImpl to poll for sufficient disk space. We assume that there is a 1-1 mapping between an SFM and the underlying OS storage container. For cases where multiple DBs are hosted on a single storage container, the user is expected to allocate a single SFM instance and use the same one for all the DBs. If no SFM is specified by the user, DBImpl::Open() will allocate one, but this will be one per DB and each DB will recover independently. The recovery implemented by SFM is as follows - a) On the first occurance of an out of space error during compaction, subsequent compactions will be delayed until the disk free space check indicates enough available space. The required space is computed as the sum of input sizes. b) The free space check requirement will be removed once the amount of free space is greater than the size reserved by in progress compactions when the first error occured c) If the out of space error is a hard error, a background thread in SFM will poll for sufficient headroom before triggering the recovery of the database and putting it in write-only mode. The headroom is calculated as the sum of the write_buffer_size of all the DB instances associated with the SFM 4. EventListener callbacks will be called at the start and completion of automatic recovery. Users can disable the auto recov ery in the start callback, and later initiate it manually by calling DB::Resume() Todo: 1. More extensive testing 2. Add disk full condition to db_stress (follow-on PR) Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164 Differential Revision: D9846378 Pulled By: anand1976 fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
6 years ago
// The previous failed compaction may get retried automatically, so we may
// be left with 0 or 1 files in level 1, depending on when the retry gets
// scheduled
ASSERT_OK(Put("key0", "val"));
ASSERT_OK(dbfull()->TEST_WaitForFlushMemTable());
ASSERT_OK(dbfull()->TEST_WaitForCompact());
Auto recovery from out of space errors (#4164) Summary: This commit implements automatic recovery from a Status::NoSpace() error during background operations such as write callback, flush and compaction. The broad design is as follows - 1. Compaction errors are treated as soft errors and don't put the database in read-only mode. A compaction is delayed until enough free disk space is available to accomodate the compaction outputs, which is estimated based on the input size. This means that users can continue to write, and we rely on the WriteController to delay or stop writes if the compaction debt becomes too high due to persistent low disk space condition 2. Errors during write callback and flush are treated as hard errors, i.e the database is put in read-only mode and goes back to read-write only fater certain recovery actions are taken. 3. Both types of recovery rely on the SstFileManagerImpl to poll for sufficient disk space. We assume that there is a 1-1 mapping between an SFM and the underlying OS storage container. For cases where multiple DBs are hosted on a single storage container, the user is expected to allocate a single SFM instance and use the same one for all the DBs. If no SFM is specified by the user, DBImpl::Open() will allocate one, but this will be one per DB and each DB will recover independently. The recovery implemented by SFM is as follows - a) On the first occurance of an out of space error during compaction, subsequent compactions will be delayed until the disk free space check indicates enough available space. The required space is computed as the sum of input sizes. b) The free space check requirement will be removed once the amount of free space is greater than the size reserved by in progress compactions when the first error occured c) If the out of space error is a hard error, a background thread in SFM will poll for sufficient headroom before triggering the recovery of the database and putting it in write-only mode. The headroom is calculated as the sum of the write_buffer_size of all the DB instances associated with the SFM 4. EventListener callbacks will be called at the start and completion of automatic recovery. Users can disable the auto recov ery in the start callback, and later initiate it manually by calling DB::Resume() Todo: 1. More extensive testing 2. Add disk full condition to db_stress (follow-on PR) Pull Request resolved: https://github.com/facebook/rocksdb/pull/4164 Differential Revision: D9846378 Pulled By: anand1976 fbshipit-source-id: 80ea875dbd7f00205e19c82215ff6e37da10da4a
6 years ago
ASSERT_LE(1, NumTableFilesAtLevel(0));
}
class TestFileOperationListener : public EventListener {
public:
TestFileOperationListener() {
file_reads_.store(0);
file_reads_success_.store(0);
file_writes_.store(0);
file_writes_success_.store(0);
file_flushes_.store(0);
file_flushes_success_.store(0);
file_closes_.store(0);
file_closes_success_.store(0);
file_syncs_.store(0);
file_syncs_success_.store(0);
file_truncates_.store(0);
file_truncates_success_.store(0);
file_seq_reads_.store(0);
blob_file_reads_.store(0);
blob_file_writes_.store(0);
blob_file_flushes_.store(0);
blob_file_closes_.store(0);
blob_file_syncs_.store(0);
blob_file_truncates_.store(0);
}
void OnFileReadFinish(const FileOperationInfo& info) override {
++file_reads_;
if (info.status.ok()) {
++file_reads_success_;
}
if (info.path.find("MANIFEST") != std::string::npos) {
++file_seq_reads_;
}
if (EndsWith(info.path, ".blob")) {
++blob_file_reads_;
}
ReportDuration(info);
}
void OnFileWriteFinish(const FileOperationInfo& info) override {
++file_writes_;
if (info.status.ok()) {
++file_writes_success_;
}
if (EndsWith(info.path, ".blob")) {
++blob_file_writes_;
}
ReportDuration(info);
}
void OnFileFlushFinish(const FileOperationInfo& info) override {
++file_flushes_;
if (info.status.ok()) {
++file_flushes_success_;
}
if (EndsWith(info.path, ".blob")) {
++blob_file_flushes_;
}
ReportDuration(info);
}
void OnFileCloseFinish(const FileOperationInfo& info) override {
++file_closes_;
if (info.status.ok()) {
++file_closes_success_;
}
if (EndsWith(info.path, ".blob")) {
++blob_file_closes_;
}
ReportDuration(info);
}
void OnFileSyncFinish(const FileOperationInfo& info) override {
++file_syncs_;
if (info.status.ok()) {
++file_syncs_success_;
}
if (EndsWith(info.path, ".blob")) {
++blob_file_syncs_;
}
ReportDuration(info);
}
void OnFileTruncateFinish(const FileOperationInfo& info) override {
++file_truncates_;
if (info.status.ok()) {
++file_truncates_success_;
}
if (EndsWith(info.path, ".blob")) {
++blob_file_truncates_;
}
ReportDuration(info);
}
bool ShouldBeNotifiedOnFileIO() override { return true; }
std::atomic<size_t> file_reads_;
std::atomic<size_t> file_reads_success_;
std::atomic<size_t> file_writes_;
std::atomic<size_t> file_writes_success_;
std::atomic<size_t> file_flushes_;
std::atomic<size_t> file_flushes_success_;
std::atomic<size_t> file_closes_;
std::atomic<size_t> file_closes_success_;
std::atomic<size_t> file_syncs_;
std::atomic<size_t> file_syncs_success_;
std::atomic<size_t> file_truncates_;
std::atomic<size_t> file_truncates_success_;
std::atomic<size_t> file_seq_reads_;
std::atomic<size_t> blob_file_reads_;
std::atomic<size_t> blob_file_writes_;
std::atomic<size_t> blob_file_flushes_;
std::atomic<size_t> blob_file_closes_;
std::atomic<size_t> blob_file_syncs_;
std::atomic<size_t> blob_file_truncates_;
private:
void ReportDuration(const FileOperationInfo& info) const {
ASSERT_GT(info.duration.count(), 0);
}
};
TEST_F(EventListenerTest, OnFileOperationTest) {
Options options;
options.env = CurrentOptions().env;
options.create_if_missing = true;
TestFileOperationListener* listener = new TestFileOperationListener();
options.listeners.emplace_back(listener);
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566) Summary: This PR does a few things: 1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation. 2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed: - The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated - The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory). 3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10). I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged. Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently. Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :) Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566 Reviewed By: zhichao-cao Differential Revision: D24408980 Pulled By: jay-zhuang fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
4 years ago
options.use_direct_io_for_flush_and_compaction = false;
Status s = TryReopen(options);
if (s.IsInvalidArgument()) {
options.use_direct_io_for_flush_and_compaction = false;
} else {
ASSERT_OK(s);
}
DestroyAndReopen(options);
ASSERT_OK(Put("foo", "aaa"));
ASSERT_OK(dbfull()->Flush(FlushOptions()));
ASSERT_OK(dbfull()->TEST_WaitForFlushMemTable());
ASSERT_GE(listener->file_writes_.load(),
listener->file_writes_success_.load());
ASSERT_GT(listener->file_writes_.load(), 0);
ASSERT_GE(listener->file_flushes_.load(),
listener->file_flushes_success_.load());
ASSERT_GT(listener->file_flushes_.load(), 0);
Close();
Reopen(options);
ASSERT_GE(listener->file_reads_.load(), listener->file_reads_success_.load());
ASSERT_GT(listener->file_reads_.load(), 0);
ASSERT_GE(listener->file_closes_.load(),
listener->file_closes_success_.load());
ASSERT_GT(listener->file_closes_.load(), 0);
ASSERT_GE(listener->file_syncs_.load(), listener->file_syncs_success_.load());
ASSERT_GT(listener->file_syncs_.load(), 0);
if (true == options.use_direct_io_for_flush_and_compaction) {
ASSERT_GE(listener->file_truncates_.load(),
listener->file_truncates_success_.load());
ASSERT_GT(listener->file_truncates_.load(), 0);
}
}
TEST_F(EventListenerTest, OnBlobFileOperationTest) {
Options options;
options.env = CurrentOptions().env;
options.create_if_missing = true;
TestFileOperationListener* listener = new TestFileOperationListener();
options.listeners.emplace_back(listener);
options.disable_auto_compactions = true;
options.enable_blob_files = true;
options.min_blob_size = 0;
options.enable_blob_garbage_collection = true;
options.blob_garbage_collection_age_cutoff = 0.5;
DestroyAndReopen(options);
ASSERT_OK(Put("Key1", "blob_value1"));
ASSERT_OK(Put("Key2", "blob_value2"));
ASSERT_OK(Put("Key3", "blob_value3"));
ASSERT_OK(Put("Key4", "blob_value4"));
ASSERT_OK(Flush());
ASSERT_OK(Put("Key3", "new_blob_value3"));
ASSERT_OK(Put("Key4", "new_blob_value4"));
ASSERT_OK(Flush());
ASSERT_OK(Put("Key5", "blob_value5"));
ASSERT_OK(Put("Key6", "blob_value6"));
ASSERT_OK(Flush());
ASSERT_GT(listener->blob_file_writes_.load(), 0U);
ASSERT_GT(listener->blob_file_flushes_.load(), 0U);
Close();
Reopen(options);
ASSERT_GT(listener->blob_file_closes_.load(), 0U);
ASSERT_GT(listener->blob_file_syncs_.load(), 0U);
if (true == options.use_direct_io_for_flush_and_compaction) {
ASSERT_GT(listener->blob_file_truncates_.load(), 0U);
}
}
TEST_F(EventListenerTest, ReadManifestAndWALOnRecovery) {
Options options;
options.env = CurrentOptions().env;
options.create_if_missing = true;
TestFileOperationListener* listener = new TestFileOperationListener();
options.listeners.emplace_back(listener);
options.use_direct_io_for_flush_and_compaction = false;
Status s = TryReopen(options);
if (s.IsInvalidArgument()) {
options.use_direct_io_for_flush_and_compaction = false;
} else {
ASSERT_OK(s);
}
DestroyAndReopen(options);
ASSERT_OK(Put("foo", "aaa"));
Close();
size_t seq_reads = listener->file_seq_reads_.load();
Reopen(options);
ASSERT_GT(listener->file_seq_reads_.load(), seq_reads);
}
class BlobDBJobLevelEventListenerTest : public EventListener {
public:
explicit BlobDBJobLevelEventListenerTest(EventListenerTest* test)
: test_(test), call_count_(0) {}
Use a sorted vector instead of a map to store blob file metadata (#9526) Summary: The patch replaces `std::map` with a sorted `std::vector` for `VersionStorageInfo::blob_files_` and preallocates the space for the `vector` before saving the `BlobFileMetaData` into the new `VersionStorageInfo` in `VersionBuilder::Rep::SaveBlobFilesTo`. These changes reduce the time the DB mutex is held while saving new `Version`s, and using a sorted `vector` also makes lookups faster thanks to better memory locality. In addition, the patch introduces helper methods `VersionStorageInfo::GetBlobFileMetaData` and `VersionStorageInfo::GetBlobFileMetaDataLB` that can be used by clients to perform lookups in the `vector`, and does some general cleanup in the parts of code where blob file metadata are used. Pull Request resolved: https://github.com/facebook/rocksdb/pull/9526 Test Plan: Ran `make check` and the crash test script for a while. Performance was tested using a load-optimized benchmark (`fillseq` with vector memtable, no WAL) and small file sizes so that a significant number of files are produced: ``` numactl --interleave=all ./db_bench --benchmarks=fillseq --allow_concurrent_memtable_write=false --level0_file_num_compaction_trigger=4 --level0_slowdown_writes_trigger=20 --level0_stop_writes_trigger=30 --max_background_jobs=8 --max_write_buffer_number=8 --db=/data/ltamasi-dbbench --wal_dir=/data/ltamasi-dbbench --num=800000000 --num_levels=8 --key_size=20 --value_size=400 --block_size=8192 --cache_size=51539607552 --cache_numshardbits=6 --compression_max_dict_bytes=0 --compression_ratio=0.5 --compression_type=lz4 --bytes_per_sync=8388608 --cache_index_and_filter_blocks=1 --cache_high_pri_pool_ratio=0.5 --benchmark_write_rate_limit=0 --write_buffer_size=16777216 --target_file_size_base=16777216 --max_bytes_for_level_base=67108864 --verify_checksum=1 --delete_obsolete_files_period_micros=62914560 --max_bytes_for_level_multiplier=8 --statistics=0 --stats_per_interval=1 --stats_interval_seconds=20 --histogram=1 --memtablerep=skip_list --bloom_bits=10 --open_files=-1 --subcompactions=1 --compaction_style=0 --min_level_to_compress=3 --level_compaction_dynamic_level_bytes=true --pin_l0_filter_and_index_blocks_in_cache=1 --soft_pending_compaction_bytes_limit=167503724544 --hard_pending_compaction_bytes_limit=335007449088 --min_level_to_compress=0 --use_existing_db=0 --sync=0 --threads=1 --memtablerep=vector --allow_concurrent_memtable_write=false --disable_wal=1 --enable_blob_files=1 --blob_file_size=16777216 --min_blob_size=0 --blob_compression_type=lz4 --enable_blob_garbage_collection=1 --seed=<some value> ``` Final statistics before the patch: ``` Cumulative writes: 0 writes, 700M keys, 0 commit groups, 0.0 writes per commit group, ingest: 284.62 GB, 121.27 MB/s Interval writes: 0 writes, 334K keys, 0 commit groups, 0.0 writes per commit group, ingest: 139.28 MB, 72.46 MB/s ``` With the patch: ``` Cumulative writes: 0 writes, 760M keys, 0 commit groups, 0.0 writes per commit group, ingest: 308.66 GB, 131.52 MB/s Interval writes: 0 writes, 445K keys, 0 commit groups, 0.0 writes per commit group, ingest: 185.35 MB, 93.15 MB/s ``` Total time to complete the benchmark is 2611 seconds with the patch, down from 2986 secs. Reviewed By: riversand963 Differential Revision: D34082728 Pulled By: ltamasi fbshipit-source-id: fc598abf676dce436734d06bb9d2d99a26a004fc
3 years ago
const VersionStorageInfo* GetVersionStorageInfo() const {
VersionSet* const versions = test_->dbfull()->GetVersionSet();
assert(versions);
ColumnFamilyData* const cfd = versions->GetColumnFamilySet()->GetDefault();
EXPECT_NE(cfd, nullptr);
Version* const current = cfd->current();
EXPECT_NE(current, nullptr);
const VersionStorageInfo* const storage_info = current->storage_info();
EXPECT_NE(storage_info, nullptr);
Use a sorted vector instead of a map to store blob file metadata (#9526) Summary: The patch replaces `std::map` with a sorted `std::vector` for `VersionStorageInfo::blob_files_` and preallocates the space for the `vector` before saving the `BlobFileMetaData` into the new `VersionStorageInfo` in `VersionBuilder::Rep::SaveBlobFilesTo`. These changes reduce the time the DB mutex is held while saving new `Version`s, and using a sorted `vector` also makes lookups faster thanks to better memory locality. In addition, the patch introduces helper methods `VersionStorageInfo::GetBlobFileMetaData` and `VersionStorageInfo::GetBlobFileMetaDataLB` that can be used by clients to perform lookups in the `vector`, and does some general cleanup in the parts of code where blob file metadata are used. Pull Request resolved: https://github.com/facebook/rocksdb/pull/9526 Test Plan: Ran `make check` and the crash test script for a while. Performance was tested using a load-optimized benchmark (`fillseq` with vector memtable, no WAL) and small file sizes so that a significant number of files are produced: ``` numactl --interleave=all ./db_bench --benchmarks=fillseq --allow_concurrent_memtable_write=false --level0_file_num_compaction_trigger=4 --level0_slowdown_writes_trigger=20 --level0_stop_writes_trigger=30 --max_background_jobs=8 --max_write_buffer_number=8 --db=/data/ltamasi-dbbench --wal_dir=/data/ltamasi-dbbench --num=800000000 --num_levels=8 --key_size=20 --value_size=400 --block_size=8192 --cache_size=51539607552 --cache_numshardbits=6 --compression_max_dict_bytes=0 --compression_ratio=0.5 --compression_type=lz4 --bytes_per_sync=8388608 --cache_index_and_filter_blocks=1 --cache_high_pri_pool_ratio=0.5 --benchmark_write_rate_limit=0 --write_buffer_size=16777216 --target_file_size_base=16777216 --max_bytes_for_level_base=67108864 --verify_checksum=1 --delete_obsolete_files_period_micros=62914560 --max_bytes_for_level_multiplier=8 --statistics=0 --stats_per_interval=1 --stats_interval_seconds=20 --histogram=1 --memtablerep=skip_list --bloom_bits=10 --open_files=-1 --subcompactions=1 --compaction_style=0 --min_level_to_compress=3 --level_compaction_dynamic_level_bytes=true --pin_l0_filter_and_index_blocks_in_cache=1 --soft_pending_compaction_bytes_limit=167503724544 --hard_pending_compaction_bytes_limit=335007449088 --min_level_to_compress=0 --use_existing_db=0 --sync=0 --threads=1 --memtablerep=vector --allow_concurrent_memtable_write=false --disable_wal=1 --enable_blob_files=1 --blob_file_size=16777216 --min_blob_size=0 --blob_compression_type=lz4 --enable_blob_garbage_collection=1 --seed=<some value> ``` Final statistics before the patch: ``` Cumulative writes: 0 writes, 700M keys, 0 commit groups, 0.0 writes per commit group, ingest: 284.62 GB, 121.27 MB/s Interval writes: 0 writes, 334K keys, 0 commit groups, 0.0 writes per commit group, ingest: 139.28 MB, 72.46 MB/s ``` With the patch: ``` Cumulative writes: 0 writes, 760M keys, 0 commit groups, 0.0 writes per commit group, ingest: 308.66 GB, 131.52 MB/s Interval writes: 0 writes, 445K keys, 0 commit groups, 0.0 writes per commit group, ingest: 185.35 MB, 93.15 MB/s ``` Total time to complete the benchmark is 2611 seconds with the patch, down from 2986 secs. Reviewed By: riversand963 Differential Revision: D34082728 Pulled By: ltamasi fbshipit-source-id: fc598abf676dce436734d06bb9d2d99a26a004fc
3 years ago
return storage_info;
}
void CheckBlobFileAdditions(
const std::vector<BlobFileAdditionInfo>& blob_file_addition_infos) const {
const auto* vstorage = GetVersionStorageInfo();
EXPECT_FALSE(blob_file_addition_infos.empty());
for (const auto& blob_file_addition_info : blob_file_addition_infos) {
const auto meta = vstorage->GetBlobFileMetaData(
blob_file_addition_info.blob_file_number);
EXPECT_NE(meta, nullptr);
EXPECT_EQ(meta->GetBlobFileNumber(),
blob_file_addition_info.blob_file_number);
EXPECT_EQ(meta->GetTotalBlobBytes(),
blob_file_addition_info.total_blob_bytes);
EXPECT_EQ(meta->GetTotalBlobCount(),
blob_file_addition_info.total_blob_count);
EXPECT_FALSE(blob_file_addition_info.blob_file_path.empty());
}
}
std::vector<std::string> GetFlushedFiles() {
std::lock_guard<std::mutex> lock(mutex_);
std::vector<std::string> result;
for (const auto& fname : flushed_files_) {
result.push_back(fname);
}
return result;
}
void OnFlushCompleted(DB* /*db*/, const FlushJobInfo& info) override {
call_count_++;
Use a sorted vector instead of a map to store blob file metadata (#9526) Summary: The patch replaces `std::map` with a sorted `std::vector` for `VersionStorageInfo::blob_files_` and preallocates the space for the `vector` before saving the `BlobFileMetaData` into the new `VersionStorageInfo` in `VersionBuilder::Rep::SaveBlobFilesTo`. These changes reduce the time the DB mutex is held while saving new `Version`s, and using a sorted `vector` also makes lookups faster thanks to better memory locality. In addition, the patch introduces helper methods `VersionStorageInfo::GetBlobFileMetaData` and `VersionStorageInfo::GetBlobFileMetaDataLB` that can be used by clients to perform lookups in the `vector`, and does some general cleanup in the parts of code where blob file metadata are used. Pull Request resolved: https://github.com/facebook/rocksdb/pull/9526 Test Plan: Ran `make check` and the crash test script for a while. Performance was tested using a load-optimized benchmark (`fillseq` with vector memtable, no WAL) and small file sizes so that a significant number of files are produced: ``` numactl --interleave=all ./db_bench --benchmarks=fillseq --allow_concurrent_memtable_write=false --level0_file_num_compaction_trigger=4 --level0_slowdown_writes_trigger=20 --level0_stop_writes_trigger=30 --max_background_jobs=8 --max_write_buffer_number=8 --db=/data/ltamasi-dbbench --wal_dir=/data/ltamasi-dbbench --num=800000000 --num_levels=8 --key_size=20 --value_size=400 --block_size=8192 --cache_size=51539607552 --cache_numshardbits=6 --compression_max_dict_bytes=0 --compression_ratio=0.5 --compression_type=lz4 --bytes_per_sync=8388608 --cache_index_and_filter_blocks=1 --cache_high_pri_pool_ratio=0.5 --benchmark_write_rate_limit=0 --write_buffer_size=16777216 --target_file_size_base=16777216 --max_bytes_for_level_base=67108864 --verify_checksum=1 --delete_obsolete_files_period_micros=62914560 --max_bytes_for_level_multiplier=8 --statistics=0 --stats_per_interval=1 --stats_interval_seconds=20 --histogram=1 --memtablerep=skip_list --bloom_bits=10 --open_files=-1 --subcompactions=1 --compaction_style=0 --min_level_to_compress=3 --level_compaction_dynamic_level_bytes=true --pin_l0_filter_and_index_blocks_in_cache=1 --soft_pending_compaction_bytes_limit=167503724544 --hard_pending_compaction_bytes_limit=335007449088 --min_level_to_compress=0 --use_existing_db=0 --sync=0 --threads=1 --memtablerep=vector --allow_concurrent_memtable_write=false --disable_wal=1 --enable_blob_files=1 --blob_file_size=16777216 --min_blob_size=0 --blob_compression_type=lz4 --enable_blob_garbage_collection=1 --seed=<some value> ``` Final statistics before the patch: ``` Cumulative writes: 0 writes, 700M keys, 0 commit groups, 0.0 writes per commit group, ingest: 284.62 GB, 121.27 MB/s Interval writes: 0 writes, 334K keys, 0 commit groups, 0.0 writes per commit group, ingest: 139.28 MB, 72.46 MB/s ``` With the patch: ``` Cumulative writes: 0 writes, 760M keys, 0 commit groups, 0.0 writes per commit group, ingest: 308.66 GB, 131.52 MB/s Interval writes: 0 writes, 445K keys, 0 commit groups, 0.0 writes per commit group, ingest: 185.35 MB, 93.15 MB/s ``` Total time to complete the benchmark is 2611 seconds with the patch, down from 2986 secs. Reviewed By: riversand963 Differential Revision: D34082728 Pulled By: ltamasi fbshipit-source-id: fc598abf676dce436734d06bb9d2d99a26a004fc
3 years ago
{
std::lock_guard<std::mutex> lock(mutex_);
flushed_files_.push_back(info.file_path);
}
Use a sorted vector instead of a map to store blob file metadata (#9526) Summary: The patch replaces `std::map` with a sorted `std::vector` for `VersionStorageInfo::blob_files_` and preallocates the space for the `vector` before saving the `BlobFileMetaData` into the new `VersionStorageInfo` in `VersionBuilder::Rep::SaveBlobFilesTo`. These changes reduce the time the DB mutex is held while saving new `Version`s, and using a sorted `vector` also makes lookups faster thanks to better memory locality. In addition, the patch introduces helper methods `VersionStorageInfo::GetBlobFileMetaData` and `VersionStorageInfo::GetBlobFileMetaDataLB` that can be used by clients to perform lookups in the `vector`, and does some general cleanup in the parts of code where blob file metadata are used. Pull Request resolved: https://github.com/facebook/rocksdb/pull/9526 Test Plan: Ran `make check` and the crash test script for a while. Performance was tested using a load-optimized benchmark (`fillseq` with vector memtable, no WAL) and small file sizes so that a significant number of files are produced: ``` numactl --interleave=all ./db_bench --benchmarks=fillseq --allow_concurrent_memtable_write=false --level0_file_num_compaction_trigger=4 --level0_slowdown_writes_trigger=20 --level0_stop_writes_trigger=30 --max_background_jobs=8 --max_write_buffer_number=8 --db=/data/ltamasi-dbbench --wal_dir=/data/ltamasi-dbbench --num=800000000 --num_levels=8 --key_size=20 --value_size=400 --block_size=8192 --cache_size=51539607552 --cache_numshardbits=6 --compression_max_dict_bytes=0 --compression_ratio=0.5 --compression_type=lz4 --bytes_per_sync=8388608 --cache_index_and_filter_blocks=1 --cache_high_pri_pool_ratio=0.5 --benchmark_write_rate_limit=0 --write_buffer_size=16777216 --target_file_size_base=16777216 --max_bytes_for_level_base=67108864 --verify_checksum=1 --delete_obsolete_files_period_micros=62914560 --max_bytes_for_level_multiplier=8 --statistics=0 --stats_per_interval=1 --stats_interval_seconds=20 --histogram=1 --memtablerep=skip_list --bloom_bits=10 --open_files=-1 --subcompactions=1 --compaction_style=0 --min_level_to_compress=3 --level_compaction_dynamic_level_bytes=true --pin_l0_filter_and_index_blocks_in_cache=1 --soft_pending_compaction_bytes_limit=167503724544 --hard_pending_compaction_bytes_limit=335007449088 --min_level_to_compress=0 --use_existing_db=0 --sync=0 --threads=1 --memtablerep=vector --allow_concurrent_memtable_write=false --disable_wal=1 --enable_blob_files=1 --blob_file_size=16777216 --min_blob_size=0 --blob_compression_type=lz4 --enable_blob_garbage_collection=1 --seed=<some value> ``` Final statistics before the patch: ``` Cumulative writes: 0 writes, 700M keys, 0 commit groups, 0.0 writes per commit group, ingest: 284.62 GB, 121.27 MB/s Interval writes: 0 writes, 334K keys, 0 commit groups, 0.0 writes per commit group, ingest: 139.28 MB, 72.46 MB/s ``` With the patch: ``` Cumulative writes: 0 writes, 760M keys, 0 commit groups, 0.0 writes per commit group, ingest: 308.66 GB, 131.52 MB/s Interval writes: 0 writes, 445K keys, 0 commit groups, 0.0 writes per commit group, ingest: 185.35 MB, 93.15 MB/s ``` Total time to complete the benchmark is 2611 seconds with the patch, down from 2986 secs. Reviewed By: riversand963 Differential Revision: D34082728 Pulled By: ltamasi fbshipit-source-id: fc598abf676dce436734d06bb9d2d99a26a004fc
3 years ago
EXPECT_EQ(info.blob_compression_type, kNoCompression);
Use a sorted vector instead of a map to store blob file metadata (#9526) Summary: The patch replaces `std::map` with a sorted `std::vector` for `VersionStorageInfo::blob_files_` and preallocates the space for the `vector` before saving the `BlobFileMetaData` into the new `VersionStorageInfo` in `VersionBuilder::Rep::SaveBlobFilesTo`. These changes reduce the time the DB mutex is held while saving new `Version`s, and using a sorted `vector` also makes lookups faster thanks to better memory locality. In addition, the patch introduces helper methods `VersionStorageInfo::GetBlobFileMetaData` and `VersionStorageInfo::GetBlobFileMetaDataLB` that can be used by clients to perform lookups in the `vector`, and does some general cleanup in the parts of code where blob file metadata are used. Pull Request resolved: https://github.com/facebook/rocksdb/pull/9526 Test Plan: Ran `make check` and the crash test script for a while. Performance was tested using a load-optimized benchmark (`fillseq` with vector memtable, no WAL) and small file sizes so that a significant number of files are produced: ``` numactl --interleave=all ./db_bench --benchmarks=fillseq --allow_concurrent_memtable_write=false --level0_file_num_compaction_trigger=4 --level0_slowdown_writes_trigger=20 --level0_stop_writes_trigger=30 --max_background_jobs=8 --max_write_buffer_number=8 --db=/data/ltamasi-dbbench --wal_dir=/data/ltamasi-dbbench --num=800000000 --num_levels=8 --key_size=20 --value_size=400 --block_size=8192 --cache_size=51539607552 --cache_numshardbits=6 --compression_max_dict_bytes=0 --compression_ratio=0.5 --compression_type=lz4 --bytes_per_sync=8388608 --cache_index_and_filter_blocks=1 --cache_high_pri_pool_ratio=0.5 --benchmark_write_rate_limit=0 --write_buffer_size=16777216 --target_file_size_base=16777216 --max_bytes_for_level_base=67108864 --verify_checksum=1 --delete_obsolete_files_period_micros=62914560 --max_bytes_for_level_multiplier=8 --statistics=0 --stats_per_interval=1 --stats_interval_seconds=20 --histogram=1 --memtablerep=skip_list --bloom_bits=10 --open_files=-1 --subcompactions=1 --compaction_style=0 --min_level_to_compress=3 --level_compaction_dynamic_level_bytes=true --pin_l0_filter_and_index_blocks_in_cache=1 --soft_pending_compaction_bytes_limit=167503724544 --hard_pending_compaction_bytes_limit=335007449088 --min_level_to_compress=0 --use_existing_db=0 --sync=0 --threads=1 --memtablerep=vector --allow_concurrent_memtable_write=false --disable_wal=1 --enable_blob_files=1 --blob_file_size=16777216 --min_blob_size=0 --blob_compression_type=lz4 --enable_blob_garbage_collection=1 --seed=<some value> ``` Final statistics before the patch: ``` Cumulative writes: 0 writes, 700M keys, 0 commit groups, 0.0 writes per commit group, ingest: 284.62 GB, 121.27 MB/s Interval writes: 0 writes, 334K keys, 0 commit groups, 0.0 writes per commit group, ingest: 139.28 MB, 72.46 MB/s ``` With the patch: ``` Cumulative writes: 0 writes, 760M keys, 0 commit groups, 0.0 writes per commit group, ingest: 308.66 GB, 131.52 MB/s Interval writes: 0 writes, 445K keys, 0 commit groups, 0.0 writes per commit group, ingest: 185.35 MB, 93.15 MB/s ``` Total time to complete the benchmark is 2611 seconds with the patch, down from 2986 secs. Reviewed By: riversand963 Differential Revision: D34082728 Pulled By: ltamasi fbshipit-source-id: fc598abf676dce436734d06bb9d2d99a26a004fc
3 years ago
CheckBlobFileAdditions(info.blob_file_addition_infos);
}
Use a sorted vector instead of a map to store blob file metadata (#9526) Summary: The patch replaces `std::map` with a sorted `std::vector` for `VersionStorageInfo::blob_files_` and preallocates the space for the `vector` before saving the `BlobFileMetaData` into the new `VersionStorageInfo` in `VersionBuilder::Rep::SaveBlobFilesTo`. These changes reduce the time the DB mutex is held while saving new `Version`s, and using a sorted `vector` also makes lookups faster thanks to better memory locality. In addition, the patch introduces helper methods `VersionStorageInfo::GetBlobFileMetaData` and `VersionStorageInfo::GetBlobFileMetaDataLB` that can be used by clients to perform lookups in the `vector`, and does some general cleanup in the parts of code where blob file metadata are used. Pull Request resolved: https://github.com/facebook/rocksdb/pull/9526 Test Plan: Ran `make check` and the crash test script for a while. Performance was tested using a load-optimized benchmark (`fillseq` with vector memtable, no WAL) and small file sizes so that a significant number of files are produced: ``` numactl --interleave=all ./db_bench --benchmarks=fillseq --allow_concurrent_memtable_write=false --level0_file_num_compaction_trigger=4 --level0_slowdown_writes_trigger=20 --level0_stop_writes_trigger=30 --max_background_jobs=8 --max_write_buffer_number=8 --db=/data/ltamasi-dbbench --wal_dir=/data/ltamasi-dbbench --num=800000000 --num_levels=8 --key_size=20 --value_size=400 --block_size=8192 --cache_size=51539607552 --cache_numshardbits=6 --compression_max_dict_bytes=0 --compression_ratio=0.5 --compression_type=lz4 --bytes_per_sync=8388608 --cache_index_and_filter_blocks=1 --cache_high_pri_pool_ratio=0.5 --benchmark_write_rate_limit=0 --write_buffer_size=16777216 --target_file_size_base=16777216 --max_bytes_for_level_base=67108864 --verify_checksum=1 --delete_obsolete_files_period_micros=62914560 --max_bytes_for_level_multiplier=8 --statistics=0 --stats_per_interval=1 --stats_interval_seconds=20 --histogram=1 --memtablerep=skip_list --bloom_bits=10 --open_files=-1 --subcompactions=1 --compaction_style=0 --min_level_to_compress=3 --level_compaction_dynamic_level_bytes=true --pin_l0_filter_and_index_blocks_in_cache=1 --soft_pending_compaction_bytes_limit=167503724544 --hard_pending_compaction_bytes_limit=335007449088 --min_level_to_compress=0 --use_existing_db=0 --sync=0 --threads=1 --memtablerep=vector --allow_concurrent_memtable_write=false --disable_wal=1 --enable_blob_files=1 --blob_file_size=16777216 --min_blob_size=0 --blob_compression_type=lz4 --enable_blob_garbage_collection=1 --seed=<some value> ``` Final statistics before the patch: ``` Cumulative writes: 0 writes, 700M keys, 0 commit groups, 0.0 writes per commit group, ingest: 284.62 GB, 121.27 MB/s Interval writes: 0 writes, 334K keys, 0 commit groups, 0.0 writes per commit group, ingest: 139.28 MB, 72.46 MB/s ``` With the patch: ``` Cumulative writes: 0 writes, 760M keys, 0 commit groups, 0.0 writes per commit group, ingest: 308.66 GB, 131.52 MB/s Interval writes: 0 writes, 445K keys, 0 commit groups, 0.0 writes per commit group, ingest: 185.35 MB, 93.15 MB/s ``` Total time to complete the benchmark is 2611 seconds with the patch, down from 2986 secs. Reviewed By: riversand963 Differential Revision: D34082728 Pulled By: ltamasi fbshipit-source-id: fc598abf676dce436734d06bb9d2d99a26a004fc
3 years ago
void OnCompactionCompleted(DB* /*db*/,
const CompactionJobInfo& info) override {
call_count_++;
Use a sorted vector instead of a map to store blob file metadata (#9526) Summary: The patch replaces `std::map` with a sorted `std::vector` for `VersionStorageInfo::blob_files_` and preallocates the space for the `vector` before saving the `BlobFileMetaData` into the new `VersionStorageInfo` in `VersionBuilder::Rep::SaveBlobFilesTo`. These changes reduce the time the DB mutex is held while saving new `Version`s, and using a sorted `vector` also makes lookups faster thanks to better memory locality. In addition, the patch introduces helper methods `VersionStorageInfo::GetBlobFileMetaData` and `VersionStorageInfo::GetBlobFileMetaDataLB` that can be used by clients to perform lookups in the `vector`, and does some general cleanup in the parts of code where blob file metadata are used. Pull Request resolved: https://github.com/facebook/rocksdb/pull/9526 Test Plan: Ran `make check` and the crash test script for a while. Performance was tested using a load-optimized benchmark (`fillseq` with vector memtable, no WAL) and small file sizes so that a significant number of files are produced: ``` numactl --interleave=all ./db_bench --benchmarks=fillseq --allow_concurrent_memtable_write=false --level0_file_num_compaction_trigger=4 --level0_slowdown_writes_trigger=20 --level0_stop_writes_trigger=30 --max_background_jobs=8 --max_write_buffer_number=8 --db=/data/ltamasi-dbbench --wal_dir=/data/ltamasi-dbbench --num=800000000 --num_levels=8 --key_size=20 --value_size=400 --block_size=8192 --cache_size=51539607552 --cache_numshardbits=6 --compression_max_dict_bytes=0 --compression_ratio=0.5 --compression_type=lz4 --bytes_per_sync=8388608 --cache_index_and_filter_blocks=1 --cache_high_pri_pool_ratio=0.5 --benchmark_write_rate_limit=0 --write_buffer_size=16777216 --target_file_size_base=16777216 --max_bytes_for_level_base=67108864 --verify_checksum=1 --delete_obsolete_files_period_micros=62914560 --max_bytes_for_level_multiplier=8 --statistics=0 --stats_per_interval=1 --stats_interval_seconds=20 --histogram=1 --memtablerep=skip_list --bloom_bits=10 --open_files=-1 --subcompactions=1 --compaction_style=0 --min_level_to_compress=3 --level_compaction_dynamic_level_bytes=true --pin_l0_filter_and_index_blocks_in_cache=1 --soft_pending_compaction_bytes_limit=167503724544 --hard_pending_compaction_bytes_limit=335007449088 --min_level_to_compress=0 --use_existing_db=0 --sync=0 --threads=1 --memtablerep=vector --allow_concurrent_memtable_write=false --disable_wal=1 --enable_blob_files=1 --blob_file_size=16777216 --min_blob_size=0 --blob_compression_type=lz4 --enable_blob_garbage_collection=1 --seed=<some value> ``` Final statistics before the patch: ``` Cumulative writes: 0 writes, 700M keys, 0 commit groups, 0.0 writes per commit group, ingest: 284.62 GB, 121.27 MB/s Interval writes: 0 writes, 334K keys, 0 commit groups, 0.0 writes per commit group, ingest: 139.28 MB, 72.46 MB/s ``` With the patch: ``` Cumulative writes: 0 writes, 760M keys, 0 commit groups, 0.0 writes per commit group, ingest: 308.66 GB, 131.52 MB/s Interval writes: 0 writes, 445K keys, 0 commit groups, 0.0 writes per commit group, ingest: 185.35 MB, 93.15 MB/s ``` Total time to complete the benchmark is 2611 seconds with the patch, down from 2986 secs. Reviewed By: riversand963 Differential Revision: D34082728 Pulled By: ltamasi fbshipit-source-id: fc598abf676dce436734d06bb9d2d99a26a004fc
3 years ago
EXPECT_EQ(info.blob_compression_type, kNoCompression);
CheckBlobFileAdditions(info.blob_file_addition_infos);
EXPECT_FALSE(info.blob_file_garbage_infos.empty());
Use a sorted vector instead of a map to store blob file metadata (#9526) Summary: The patch replaces `std::map` with a sorted `std::vector` for `VersionStorageInfo::blob_files_` and preallocates the space for the `vector` before saving the `BlobFileMetaData` into the new `VersionStorageInfo` in `VersionBuilder::Rep::SaveBlobFilesTo`. These changes reduce the time the DB mutex is held while saving new `Version`s, and using a sorted `vector` also makes lookups faster thanks to better memory locality. In addition, the patch introduces helper methods `VersionStorageInfo::GetBlobFileMetaData` and `VersionStorageInfo::GetBlobFileMetaDataLB` that can be used by clients to perform lookups in the `vector`, and does some general cleanup in the parts of code where blob file metadata are used. Pull Request resolved: https://github.com/facebook/rocksdb/pull/9526 Test Plan: Ran `make check` and the crash test script for a while. Performance was tested using a load-optimized benchmark (`fillseq` with vector memtable, no WAL) and small file sizes so that a significant number of files are produced: ``` numactl --interleave=all ./db_bench --benchmarks=fillseq --allow_concurrent_memtable_write=false --level0_file_num_compaction_trigger=4 --level0_slowdown_writes_trigger=20 --level0_stop_writes_trigger=30 --max_background_jobs=8 --max_write_buffer_number=8 --db=/data/ltamasi-dbbench --wal_dir=/data/ltamasi-dbbench --num=800000000 --num_levels=8 --key_size=20 --value_size=400 --block_size=8192 --cache_size=51539607552 --cache_numshardbits=6 --compression_max_dict_bytes=0 --compression_ratio=0.5 --compression_type=lz4 --bytes_per_sync=8388608 --cache_index_and_filter_blocks=1 --cache_high_pri_pool_ratio=0.5 --benchmark_write_rate_limit=0 --write_buffer_size=16777216 --target_file_size_base=16777216 --max_bytes_for_level_base=67108864 --verify_checksum=1 --delete_obsolete_files_period_micros=62914560 --max_bytes_for_level_multiplier=8 --statistics=0 --stats_per_interval=1 --stats_interval_seconds=20 --histogram=1 --memtablerep=skip_list --bloom_bits=10 --open_files=-1 --subcompactions=1 --compaction_style=0 --min_level_to_compress=3 --level_compaction_dynamic_level_bytes=true --pin_l0_filter_and_index_blocks_in_cache=1 --soft_pending_compaction_bytes_limit=167503724544 --hard_pending_compaction_bytes_limit=335007449088 --min_level_to_compress=0 --use_existing_db=0 --sync=0 --threads=1 --memtablerep=vector --allow_concurrent_memtable_write=false --disable_wal=1 --enable_blob_files=1 --blob_file_size=16777216 --min_blob_size=0 --blob_compression_type=lz4 --enable_blob_garbage_collection=1 --seed=<some value> ``` Final statistics before the patch: ``` Cumulative writes: 0 writes, 700M keys, 0 commit groups, 0.0 writes per commit group, ingest: 284.62 GB, 121.27 MB/s Interval writes: 0 writes, 334K keys, 0 commit groups, 0.0 writes per commit group, ingest: 139.28 MB, 72.46 MB/s ``` With the patch: ``` Cumulative writes: 0 writes, 760M keys, 0 commit groups, 0.0 writes per commit group, ingest: 308.66 GB, 131.52 MB/s Interval writes: 0 writes, 445K keys, 0 commit groups, 0.0 writes per commit group, ingest: 185.35 MB, 93.15 MB/s ``` Total time to complete the benchmark is 2611 seconds with the patch, down from 2986 secs. Reviewed By: riversand963 Differential Revision: D34082728 Pulled By: ltamasi fbshipit-source-id: fc598abf676dce436734d06bb9d2d99a26a004fc
3 years ago
for (const auto& blob_file_garbage_info : info.blob_file_garbage_infos) {
EXPECT_GT(blob_file_garbage_info.blob_file_number, 0U);
EXPECT_GT(blob_file_garbage_info.garbage_blob_count, 0U);
EXPECT_GT(blob_file_garbage_info.garbage_blob_bytes, 0U);
EXPECT_FALSE(blob_file_garbage_info.blob_file_path.empty());
}
}
EventListenerTest* test_;
uint32_t call_count_;
private:
std::vector<std::string> flushed_files_;
std::mutex mutex_;
};
// Test OnFlushCompleted EventListener called for blob files
TEST_F(EventListenerTest, BlobDBOnFlushCompleted) {
Options options;
options.env = CurrentOptions().env;
options.enable_blob_files = true;
options.create_if_missing = true;
options.disable_auto_compactions = true;
options.min_blob_size = 0;
BlobDBJobLevelEventListenerTest* blob_event_listener =
new BlobDBJobLevelEventListenerTest(this);
options.listeners.emplace_back(blob_event_listener);
DestroyAndReopen(options);
ASSERT_OK(Put("Key1", "blob_value1"));
ASSERT_OK(Put("Key2", "blob_value2"));
ASSERT_OK(Flush());
ASSERT_OK(Put("Key3", "blob_value3"));
ASSERT_OK(Flush());
ASSERT_EQ(Get("Key1"), "blob_value1");
ASSERT_EQ(Get("Key2"), "blob_value2");
ASSERT_EQ(Get("Key3"), "blob_value3");
ASSERT_GT(blob_event_listener->call_count_, 0U);
}
// Test OnCompactionCompleted EventListener called for blob files
TEST_F(EventListenerTest, BlobDBOnCompactionCompleted) {
Options options;
options.env = CurrentOptions().env;
options.enable_blob_files = true;
options.create_if_missing = true;
options.disable_auto_compactions = true;
options.min_blob_size = 0;
BlobDBJobLevelEventListenerTest* blob_event_listener =
new BlobDBJobLevelEventListenerTest(this);
options.listeners.emplace_back(blob_event_listener);
options.enable_blob_garbage_collection = true;
options.blob_garbage_collection_age_cutoff = 0.5;
DestroyAndReopen(options);
ASSERT_OK(Put("Key1", "blob_value1"));
ASSERT_OK(Put("Key2", "blob_value2"));
ASSERT_OK(Put("Key3", "blob_value3"));
ASSERT_OK(Put("Key4", "blob_value4"));
ASSERT_OK(Flush());
ASSERT_OK(Put("Key3", "new_blob_value3"));
ASSERT_OK(Put("Key4", "new_blob_value4"));
ASSERT_OK(Flush());
ASSERT_OK(Put("Key5", "blob_value5"));
ASSERT_OK(Put("Key6", "blob_value6"));
ASSERT_OK(Flush());
blob_event_listener->call_count_ = 0;
constexpr Slice* begin = nullptr;
constexpr Slice* end = nullptr;
// On compaction, because of blob_garbage_collection_age_cutoff, it will
// delete the oldest blob file and create new blob file during compaction.
ASSERT_OK(db_->CompactRange(CompactRangeOptions(), begin, end));
// Make sure, OnCompactionCompleted is called.
ASSERT_GT(blob_event_listener->call_count_, 0U);
}
// Test CompactFiles calls OnCompactionCompleted EventListener for blob files
// and populate the blob files info.
TEST_F(EventListenerTest, BlobDBCompactFiles) {
Options options;
options.env = CurrentOptions().env;
options.enable_blob_files = true;
options.create_if_missing = true;
options.disable_auto_compactions = true;
options.min_blob_size = 0;
options.enable_blob_garbage_collection = true;
options.blob_garbage_collection_age_cutoff = 0.5;
BlobDBJobLevelEventListenerTest* blob_event_listener =
new BlobDBJobLevelEventListenerTest(this);
options.listeners.emplace_back(blob_event_listener);
DestroyAndReopen(options);
ASSERT_OK(Put("Key1", "blob_value1"));
ASSERT_OK(Put("Key2", "blob_value2"));
ASSERT_OK(Put("Key3", "blob_value3"));
ASSERT_OK(Put("Key4", "blob_value4"));
ASSERT_OK(Flush());
ASSERT_OK(Put("Key3", "new_blob_value3"));
ASSERT_OK(Put("Key4", "new_blob_value4"));
ASSERT_OK(Flush());
ASSERT_OK(Put("Key5", "blob_value5"));
ASSERT_OK(Put("Key6", "blob_value6"));
ASSERT_OK(Flush());
std::vector<std::string> output_file_names;
CompactionJobInfo compaction_job_info;
// On compaction, because of blob_garbage_collection_age_cutoff, it will
// delete the oldest blob file and create new blob file during compaction
// which will be populated in output_files_names.
ASSERT_OK(dbfull()->CompactFiles(
CompactionOptions(), blob_event_listener->GetFlushedFiles(), 1, -1,
&output_file_names, &compaction_job_info));
bool is_blob_in_output = false;
for (const auto& file : output_file_names) {
if (EndsWith(file, ".blob")) {
is_blob_in_output = true;
}
}
ASSERT_TRUE(is_blob_in_output);
for (const auto& blob_file_addition_info :
compaction_job_info.blob_file_addition_infos) {
EXPECT_GT(blob_file_addition_info.blob_file_number, 0U);
EXPECT_GT(blob_file_addition_info.total_blob_bytes, 0U);
EXPECT_GT(blob_file_addition_info.total_blob_count, 0U);
EXPECT_FALSE(blob_file_addition_info.blob_file_path.empty());
}
for (const auto& blob_file_garbage_info :
compaction_job_info.blob_file_garbage_infos) {
EXPECT_GT(blob_file_garbage_info.blob_file_number, 0U);
EXPECT_GT(blob_file_garbage_info.garbage_blob_count, 0U);
EXPECT_GT(blob_file_garbage_info.garbage_blob_bytes, 0U);
EXPECT_FALSE(blob_file_garbage_info.blob_file_path.empty());
}
}
class BlobDBFileLevelEventListener : public EventListener {
public:
void OnBlobFileCreationStarted(
const BlobFileCreationBriefInfo& info) override {
files_started_++;
EXPECT_FALSE(info.db_name.empty());
EXPECT_FALSE(info.cf_name.empty());
EXPECT_FALSE(info.file_path.empty());
EXPECT_GT(info.job_id, 0);
}
void OnBlobFileCreated(const BlobFileCreationInfo& info) override {
files_created_++;
EXPECT_FALSE(info.db_name.empty());
EXPECT_FALSE(info.cf_name.empty());
EXPECT_FALSE(info.file_path.empty());
EXPECT_GT(info.job_id, 0);
EXPECT_GT(info.total_blob_count, 0U);
EXPECT_GT(info.total_blob_bytes, 0U);
EXPECT_EQ(info.file_checksum, kUnknownFileChecksum);
EXPECT_EQ(info.file_checksum_func_name, kUnknownFileChecksumFuncName);
EXPECT_TRUE(info.status.ok());
}
void OnBlobFileDeleted(const BlobFileDeletionInfo& info) override {
files_deleted_++;
EXPECT_FALSE(info.db_name.empty());
EXPECT_FALSE(info.file_path.empty());
EXPECT_GT(info.job_id, 0);
EXPECT_TRUE(info.status.ok());
}
void CheckCounters() {
EXPECT_EQ(files_started_, files_created_);
EXPECT_GT(files_started_, 0U);
EXPECT_GT(files_deleted_, 0U);
EXPECT_LT(files_deleted_, files_created_);
}
private:
std::atomic<uint32_t> files_started_{};
std::atomic<uint32_t> files_created_{};
std::atomic<uint32_t> files_deleted_{};
};
TEST_F(EventListenerTest, BlobDBFileTest) {
Options options;
options.env = CurrentOptions().env;
options.enable_blob_files = true;
options.create_if_missing = true;
options.disable_auto_compactions = true;
options.min_blob_size = 0;
options.enable_blob_garbage_collection = true;
options.blob_garbage_collection_age_cutoff = 0.5;
BlobDBFileLevelEventListener* blob_event_listener =
new BlobDBFileLevelEventListener();
options.listeners.emplace_back(blob_event_listener);
DestroyAndReopen(options);
ASSERT_OK(Put("Key1", "blob_value1"));
ASSERT_OK(Put("Key2", "blob_value2"));
ASSERT_OK(Put("Key3", "blob_value3"));
ASSERT_OK(Put("Key4", "blob_value4"));
ASSERT_OK(Flush());
ASSERT_OK(Put("Key3", "new_blob_value3"));
ASSERT_OK(Put("Key4", "new_blob_value4"));
ASSERT_OK(Flush());
ASSERT_OK(Put("Key5", "blob_value5"));
ASSERT_OK(Put("Key6", "blob_value6"));
ASSERT_OK(Flush());
constexpr Slice* begin = nullptr;
constexpr Slice* end = nullptr;
// On compaction, because of blob_garbage_collection_age_cutoff, it will
// delete the oldest blob file and create new blob file during compaction.
ASSERT_OK(db_->CompactRange(CompactRangeOptions(), begin, end));
ASSERT_OK(dbfull()->TEST_WaitForCompact());
blob_event_listener->CheckCounters();
}
} // namespace ROCKSDB_NAMESPACE
#endif // ROCKSDB_LITE
int main(int argc, char** argv) {
rocksdb: switch to gtest Summary: Our existing test notation is very similar to what is used in gtest. It makes it easy to adopt what is different. In this diff I modify existing [[ https://code.google.com/p/googletest/wiki/Primer#Test_Fixtures:_Using_the_Same_Data_Configuration_for_Multiple_Te | test fixture ]] classes to inherit from `testing::Test`. Also for unit tests that use fixture class, `TEST` is replaced with `TEST_F` as required in gtest. There are several custom `main` functions in our existing tests. To make this transition easier, I modify all `main` functions to fallow gtest notation. But eventually we can remove them and use implementation of `main` that gtest provides. ```lang=bash % cat ~/transform #!/bin/sh files=$(git ls-files '*test\.cc') for file in $files do if grep -q "rocksdb::test::RunAllTests()" $file then if grep -Eq '^class \w+Test {' $file then perl -pi -e 's/^(class \w+Test) {/${1}: public testing::Test {/g' $file perl -pi -e 's/^(TEST)/${1}_F/g' $file fi perl -pi -e 's/(int main.*\{)/${1}::testing::InitGoogleTest(&argc, argv);/g' $file perl -pi -e 's/rocksdb::test::RunAllTests/RUN_ALL_TESTS/g' $file fi done % sh ~/transform % make format ``` Second iteration of this diff contains only scripted changes. Third iteration contains manual changes to fix last errors and make it compilable. Test Plan: Build and notice no errors. ```lang=bash % USE_CLANG=1 make check -j55 ``` Tests are still testing. Reviewers: meyering, sdong, rven, igor Reviewed By: igor Subscribers: dhruba, leveldb Differential Revision: https://reviews.facebook.net/D35157
10 years ago
::testing::InitGoogleTest(&argc, argv);
return RUN_ALL_TESTS();
}