rocksdb/db_stress_tool/db_stress_test_base.cc

2982 lines
110 KiB
C++
Raw Normal View History

// Copyright (c) 2011-present, Facebook, Inc. All rights reserved.
// This source code is licensed under both the GPLv2 (found in the
// COPYING file in the root directory) and Apache 2.0 License
// (found in the LICENSE.Apache file in the root directory).
//
// Copyright (c) 2011 The LevelDB Authors. All rights reserved.
// Use of this source code is governed by a BSD-style license that can be
// found in the LICENSE file. See the AUTHORS file for names of contributors.
//
#ifdef GFLAGS
#include "db_stress_tool/db_stress_common.h"
#include "db_stress_tool/db_stress_compaction_filter.h"
#include "db_stress_tool/db_stress_driver.h"
#include "db_stress_tool/db_stress_table_properties_collector.h"
#include "rocksdb/convenience.h"
#include "rocksdb/filter_policy.h"
#include "rocksdb/secondary_cache.h"
#include "rocksdb/sst_file_manager.h"
#include "rocksdb/types.h"
#include "rocksdb/utilities/object_registry.h"
#include "test_util/testutil.h"
#include "util/cast_util.h"
#include "utilities/backup/backup_engine_impl.h"
#include "utilities/fault_injection_fs.h"
#include "utilities/fault_injection_secondary_cache.h"
namespace ROCKSDB_NAMESPACE {
namespace {
std::shared_ptr<const FilterPolicy> CreateFilterPolicy() {
if (FLAGS_bloom_bits < 0) {
return BlockBasedTableOptions().filter_policy;
}
const FilterPolicy* new_policy;
Add Bloom/Ribbon hybrid API support (#8679) Summary: This is essentially resurrection and fixing of the part of https://github.com/facebook/rocksdb/issues/8198 that was reverted in https://github.com/facebook/rocksdb/issues/8212, using data added in https://github.com/facebook/rocksdb/issues/8246. Basically, when configuring Ribbon filter, you can specify an LSM level before which Bloom will be used instead of Ribbon. But Bloom is only considered for Leveled and Universal compaction styles and file going into a known LSM level. This way, SST file writer, FIFO compaction, etc. use Ribbon filter as you would expect with NewRibbonFilterPolicy. So that this can be controlled with a single int value and so that flushes can be distinguished from intra-L0, we consider flush to go to level -1 for the purposes of this option. (Explained in API comment.) I also expect the most common and recommended Ribbon configuration to use Bloom during flush, to minimize slowing down writes and because according to my estimates, Ribbon only pays off if the structure lives in memory for more than an hour. Thus, I have changed the default for NewRibbonFilterPolicy to be this mild hybrid configuration. I don't really want to add something like NewHybridFilterPolicy because at least the mild hybrid configuration (Bloom for flush, Ribbon otherwise) should be considered a natural choice. C APIs also updated, but because they don't support overloading, rocksdb_filterpolicy_create_ribbon is kept pure ribbon for clarity and rocksdb_filterpolicy_create_ribbon_hybrid must be called for a hybrid configuration. While touching C API, I changed bits per key options from int to double. BuiltinFilterPolicy is needed so that LevelThresholdFilterPolicy doesn't inherit unused fields from BloomFilterPolicy. Pull Request resolved: https://github.com/facebook/rocksdb/pull/8679 Test Plan: new + updated tests, including crash test Reviewed By: jay-zhuang Differential Revision: D30445797 Pulled By: pdillinger fbshipit-source-id: 6f5aeddfd6d79f7e55493b563c2d1d2d568892e1
2021-08-21 02:59:24 +02:00
if (FLAGS_use_block_based_filter) {
if (FLAGS_ribbon_starting_level < 999) {
fprintf(
stderr,
"Cannot combine use_block_based_filter and ribbon_starting_level\n");
exit(1);
} else {
new_policy = NewBloomFilterPolicy(FLAGS_bloom_bits, true);
}
Add Bloom/Ribbon hybrid API support (#8679) Summary: This is essentially resurrection and fixing of the part of https://github.com/facebook/rocksdb/issues/8198 that was reverted in https://github.com/facebook/rocksdb/issues/8212, using data added in https://github.com/facebook/rocksdb/issues/8246. Basically, when configuring Ribbon filter, you can specify an LSM level before which Bloom will be used instead of Ribbon. But Bloom is only considered for Leveled and Universal compaction styles and file going into a known LSM level. This way, SST file writer, FIFO compaction, etc. use Ribbon filter as you would expect with NewRibbonFilterPolicy. So that this can be controlled with a single int value and so that flushes can be distinguished from intra-L0, we consider flush to go to level -1 for the purposes of this option. (Explained in API comment.) I also expect the most common and recommended Ribbon configuration to use Bloom during flush, to minimize slowing down writes and because according to my estimates, Ribbon only pays off if the structure lives in memory for more than an hour. Thus, I have changed the default for NewRibbonFilterPolicy to be this mild hybrid configuration. I don't really want to add something like NewHybridFilterPolicy because at least the mild hybrid configuration (Bloom for flush, Ribbon otherwise) should be considered a natural choice. C APIs also updated, but because they don't support overloading, rocksdb_filterpolicy_create_ribbon is kept pure ribbon for clarity and rocksdb_filterpolicy_create_ribbon_hybrid must be called for a hybrid configuration. While touching C API, I changed bits per key options from int to double. BuiltinFilterPolicy is needed so that LevelThresholdFilterPolicy doesn't inherit unused fields from BloomFilterPolicy. Pull Request resolved: https://github.com/facebook/rocksdb/pull/8679 Test Plan: new + updated tests, including crash test Reviewed By: jay-zhuang Differential Revision: D30445797 Pulled By: pdillinger fbshipit-source-id: 6f5aeddfd6d79f7e55493b563c2d1d2d568892e1
2021-08-21 02:59:24 +02:00
} else if (FLAGS_ribbon_starting_level >= 999) {
// Use Bloom API
new_policy = NewBloomFilterPolicy(FLAGS_bloom_bits, false);
} else {
new_policy = NewRibbonFilterPolicy(
FLAGS_bloom_bits, /* bloom_before_level */ FLAGS_ribbon_starting_level);
}
return std::shared_ptr<const FilterPolicy>(new_policy);
}
} // namespace
StressTest::StressTest()
: cache_(NewCache(FLAGS_cache_size, FLAGS_cache_numshardbits)),
compressed_cache_(NewLRUCache(FLAGS_compressed_cache_size)),
filter_policy_(CreateFilterPolicy()),
db_(nullptr),
#ifndef ROCKSDB_LITE
txn_db_(nullptr),
#endif
clock_(db_stress_env->GetSystemClock().get()),
new_column_family_name_(1),
num_times_reopened_(0),
db_preload_finished_(false),
cmp_db_(nullptr),
is_db_stopped_(false) {
if (FLAGS_destroy_db_initially) {
std::vector<std::string> files;
db_stress_env->GetChildren(FLAGS_db, &files);
for (unsigned int i = 0; i < files.size(); i++) {
if (Slice(files[i]).starts_with("heap-")) {
db_stress_env->DeleteFile(FLAGS_db + "/" + files[i]);
}
}
Options options;
options.env = db_stress_env;
// Remove files without preserving manfiest files
#ifndef ROCKSDB_LITE
const Status s = !FLAGS_use_blob_db
? DestroyDB(FLAGS_db, options)
: blob_db::DestroyBlobDB(FLAGS_db, options,
blob_db::BlobDBOptions());
#else
const Status s = DestroyDB(FLAGS_db, options);
#endif // !ROCKSDB_LITE
if (!s.ok()) {
fprintf(stderr, "Cannot destroy original db: %s\n", s.ToString().c_str());
exit(1);
}
}
}
StressTest::~StressTest() {
for (auto cf : column_families_) {
delete cf;
}
column_families_.clear();
delete db_;
assert(secondaries_.size() == secondary_cfh_lists_.size());
size_t n = secondaries_.size();
for (size_t i = 0; i != n; ++i) {
for (auto* cf : secondary_cfh_lists_[i]) {
delete cf;
}
secondary_cfh_lists_[i].clear();
delete secondaries_[i];
}
secondaries_.clear();
for (auto* cf : cmp_cfhs_) {
delete cf;
}
cmp_cfhs_.clear();
delete cmp_db_;
}
std::shared_ptr<Cache> StressTest::NewCache(size_t capacity,
int32_t num_shard_bits) {
ConfigOptions config_options;
if (capacity <= 0) {
return nullptr;
}
if (FLAGS_use_clock_cache) {
auto cache = NewClockCache((size_t)capacity);
if (!cache) {
fprintf(stderr, "Clock cache not supported.");
exit(1);
}
return cache;
} else {
LRUCacheOptions opts;
opts.capacity = capacity;
opts.num_shard_bits = num_shard_bits;
#ifndef ROCKSDB_LITE
std::shared_ptr<SecondaryCache> secondary_cache;
if (!FLAGS_secondary_cache_uri.empty()) {
Status s = SecondaryCache::CreateFromString(
config_options, FLAGS_secondary_cache_uri, &secondary_cache);
if (secondary_cache == nullptr) {
fprintf(stderr,
"No secondary cache registered matching string: %s status=%s\n",
FLAGS_secondary_cache_uri.c_str(), s.ToString().c_str());
exit(1);
}
if (FLAGS_secondary_cache_fault_one_in > 0) {
secondary_cache = std::make_shared<FaultInjectionSecondaryCache>(
secondary_cache, static_cast<uint32_t>(FLAGS_seed),
FLAGS_secondary_cache_fault_one_in);
}
opts.secondary_cache = secondary_cache;
}
#endif
return NewLRUCache(opts);
}
}
std::vector<std::string> StressTest::GetBlobCompressionTags() {
std::vector<std::string> compression_tags{"kNoCompression"};
if (Snappy_Supported()) {
compression_tags.emplace_back("kSnappyCompression");
}
if (LZ4_Supported()) {
compression_tags.emplace_back("kLZ4Compression");
}
if (ZSTD_Supported()) {
compression_tags.emplace_back("kZSTD");
}
return compression_tags;
}
bool StressTest::BuildOptionsTable() {
if (FLAGS_set_options_one_in <= 0) {
return true;
}
std::unordered_map<std::string, std::vector<std::string>> options_tbl = {
{"write_buffer_size",
{ToString(options_.write_buffer_size),
ToString(options_.write_buffer_size * 2),
ToString(options_.write_buffer_size * 4)}},
{"max_write_buffer_number",
{ToString(options_.max_write_buffer_number),
ToString(options_.max_write_buffer_number * 2),
ToString(options_.max_write_buffer_number * 4)}},
{"arena_block_size",
{
ToString(options_.arena_block_size),
ToString(options_.write_buffer_size / 4),
ToString(options_.write_buffer_size / 8),
}},
{"memtable_huge_page_size", {"0", ToString(2 * 1024 * 1024)}},
{"max_successive_merges", {"0", "2", "4"}},
{"inplace_update_num_locks", {"100", "200", "300"}},
// TODO(ljin): enable test for this option
// {"disable_auto_compactions", {"100", "200", "300"}},
{"level0_file_num_compaction_trigger",
{
ToString(options_.level0_file_num_compaction_trigger),
ToString(options_.level0_file_num_compaction_trigger + 2),
ToString(options_.level0_file_num_compaction_trigger + 4),
}},
{"level0_slowdown_writes_trigger",
{
ToString(options_.level0_slowdown_writes_trigger),
ToString(options_.level0_slowdown_writes_trigger + 2),
ToString(options_.level0_slowdown_writes_trigger + 4),
}},
{"level0_stop_writes_trigger",
{
ToString(options_.level0_stop_writes_trigger),
ToString(options_.level0_stop_writes_trigger + 2),
ToString(options_.level0_stop_writes_trigger + 4),
}},
{"max_compaction_bytes",
{
ToString(options_.target_file_size_base * 5),
ToString(options_.target_file_size_base * 15),
ToString(options_.target_file_size_base * 100),
}},
{"target_file_size_base",
{
ToString(options_.target_file_size_base),
ToString(options_.target_file_size_base * 2),
ToString(options_.target_file_size_base * 4),
}},
{"target_file_size_multiplier",
{
ToString(options_.target_file_size_multiplier),
"1",
"2",
}},
{"max_bytes_for_level_base",
{
ToString(options_.max_bytes_for_level_base / 2),
ToString(options_.max_bytes_for_level_base),
ToString(options_.max_bytes_for_level_base * 2),
}},
{"max_bytes_for_level_multiplier",
{
ToString(options_.max_bytes_for_level_multiplier),
"1",
"2",
}},
{"max_sequential_skip_in_iterations", {"4", "8", "12"}},
};
if (FLAGS_allow_setting_blob_options_dynamically) {
options_tbl.emplace("enable_blob_files",
std::vector<std::string>{"false", "true"});
options_tbl.emplace("min_blob_size",
std::vector<std::string>{"0", "8", "16"});
options_tbl.emplace("blob_file_size",
std::vector<std::string>{"1M", "16M", "256M", "1G"});
options_tbl.emplace("blob_compression_type", GetBlobCompressionTags());
options_tbl.emplace("enable_blob_garbage_collection",
std::vector<std::string>{"false", "true"});
options_tbl.emplace(
"blob_garbage_collection_age_cutoff",
std::vector<std::string>{"0.0", "0.25", "0.5", "0.75", "1.0"});
Make it possible to force the garbage collection of the oldest blob files (#8994) Summary: The current BlobDB garbage collection logic works by relocating the valid blobs from the oldest blob files as they are encountered during compaction, and cleaning up blob files once they contain nothing but garbage. However, with sufficiently skewed workloads, it is theoretically possible to end up in a situation when few or no compactions get scheduled for the SST files that contain references to the oldest blob files, which can lead to increased space amp due to the lack of GC. In order to efficiently handle such workloads, the patch adds a new BlobDB configuration option called `blob_garbage_collection_force_threshold`, which signals to BlobDB to schedule targeted compactions for the SST files that keep alive the oldest batch of blob files if the overall ratio of garbage in the given blob files meets the threshold *and* all the given blob files are eligible for GC based on `blob_garbage_collection_age_cutoff`. (For example, if the new option is set to 0.9, targeted compactions will get scheduled if the sum of garbage bytes meets or exceeds 90% of the sum of total bytes in the oldest blob files, assuming all affected blob files are below the age-based cutoff.) The net result of these targeted compactions is that the valid blobs in the oldest blob files are relocated and the oldest blob files themselves cleaned up (since *all* SST files that rely on them get compacted away). These targeted compactions are similar to periodic compactions in the sense that they force certain SST files that otherwise would not get picked up to undergo compaction and also in the sense that instead of merging files from multiple levels, they target a single file. (Note: such compactions might still include neighboring files from the same level due to the need of having a "clean cut" boundary but they never include any files from any other level.) This functionality is currently only supported with the leveled compaction style and is inactive by default (since the default value is set to 1.0, i.e. 100%). Pull Request resolved: https://github.com/facebook/rocksdb/pull/8994 Test Plan: Ran `make check` and tested using `db_bench` and the stress/crash tests. Reviewed By: riversand963 Differential Revision: D31489850 Pulled By: ltamasi fbshipit-source-id: 44057d511726a0e2a03c5d9313d7511b3f0c4eab
2021-10-12 03:00:44 +02:00
options_tbl.emplace("blob_garbage_collection_force_threshold",
std::vector<std::string>{"0.5", "0.75", "1.0"});
options_tbl.emplace("blob_compaction_readahead_size",
std::vector<std::string>{"0", "1M", "4M"});
}
options_table_ = std::move(options_tbl);
for (const auto& iter : options_table_) {
options_index_.push_back(iter.first);
}
return true;
}
void StressTest::InitDb() {
uint64_t now = clock_->NowMicros();
fprintf(stdout, "%s Initializing db_stress\n",
clock_->TimeToString(now / 1000000).c_str());
PrintEnv();
Open();
BuildOptionsTable();
}
void StressTest::FinishInitDb(SharedState* shared) {
if (FLAGS_read_only) {
uint64_t now = clock_->NowMicros();
fprintf(stdout, "%s Preloading db with %" PRIu64 " KVs\n",
clock_->TimeToString(now / 1000000).c_str(), FLAGS_max_key);
PreloadDbAndReopenAsReadOnly(FLAGS_max_key, shared);
}
Fix unsynced data loss correctness test with mixed `-test_batches_snapshots` (#9302) Summary: This fixes two bugs in the recently committed DB verification following crash-recovery with unsynced data loss (https://github.com/facebook/rocksdb/issues/8966): The first bug was in crash test runs involving mixed values for `-test_batches_snapshots`. The problem was we were neither restoring expected values nor enabling tracing when `-test_batches_snapshots=1`. This caused a future `-test_batches_snapshots=0` run to not find enough trace data to restore expected values. The fix is to restore expected values at the start of `-test_batches_snapshots=1` runs, but still leave tracing disabled as we do not need to track those KVs. The second bug was in `db_stress` runs that restore the expected values file and use compaction filter. The compaction filter was initialized to use the pre-restore expected values, which would be `munmap()`'d during `FileExpectedStateManager::Restore()`. Then compaction filter would run into a segfault. The fix is just to reorder compaction filter init after expected values restore. Pull Request resolved: https://github.com/facebook/rocksdb/pull/9302 Test Plan: - To verify the first problem, the below sequence used to fail; now it passes. ``` $ ./db_stress --db=./test-db/ --expected_values_dir=./test-db-expected/ --max_key=100000 --ops_per_thread=1000 --sync_fault_injection=1 --clear_column_family_one_in=0 --destroy_db_initially=0 -reopen=0 -test_batches_snapshots=0 $ ./db_stress --db=./test-db/ --expected_values_dir=./test-db-expected/ --max_key=100000 --ops_per_thread=1000 --sync_fault_injection=1 --clear_column_family_one_in=0 --destroy_db_initially=0 -reopen=0 -test_batches_snapshots=1 $ ./db_stress --db=./test-db/ --expected_values_dir=./test-db-expected/ --max_key=100000 --ops_per_thread=1000 --sync_fault_injection=1 --clear_column_family_one_in=0 --destroy_db_initially=0 -reopen=0 -test_batches_snapshots=0 ``` - The second problem occurred rarely in the form of a SIGSEGV on a file that was `munmap()`d. I have not seen it after this PR though this doesn't prove much. Reviewed By: jay-zhuang Differential Revision: D33155283 Pulled By: ajkr fbshipit-source-id: 66fd0f0edf34015a010c30015f14f104734e964e
2021-12-18 07:04:05 +01:00
if (shared->HasHistory()) {
// The way it works right now is, if there's any history, that means the
// previous run mutating the DB had all its operations traced, in which case
// we should always be able to `Restore()` the expected values to match the
// `db_`'s current seqno.
Status s = shared->Restore(db_);
if (!s.ok()) {
fprintf(stderr, "Error restoring historical expected values: %s\n",
s.ToString().c_str());
exit(1);
}
}
Fix unsynced data loss correctness test with mixed `-test_batches_snapshots` (#9302) Summary: This fixes two bugs in the recently committed DB verification following crash-recovery with unsynced data loss (https://github.com/facebook/rocksdb/issues/8966): The first bug was in crash test runs involving mixed values for `-test_batches_snapshots`. The problem was we were neither restoring expected values nor enabling tracing when `-test_batches_snapshots=1`. This caused a future `-test_batches_snapshots=0` run to not find enough trace data to restore expected values. The fix is to restore expected values at the start of `-test_batches_snapshots=1` runs, but still leave tracing disabled as we do not need to track those KVs. The second bug was in `db_stress` runs that restore the expected values file and use compaction filter. The compaction filter was initialized to use the pre-restore expected values, which would be `munmap()`'d during `FileExpectedStateManager::Restore()`. Then compaction filter would run into a segfault. The fix is just to reorder compaction filter init after expected values restore. Pull Request resolved: https://github.com/facebook/rocksdb/pull/9302 Test Plan: - To verify the first problem, the below sequence used to fail; now it passes. ``` $ ./db_stress --db=./test-db/ --expected_values_dir=./test-db-expected/ --max_key=100000 --ops_per_thread=1000 --sync_fault_injection=1 --clear_column_family_one_in=0 --destroy_db_initially=0 -reopen=0 -test_batches_snapshots=0 $ ./db_stress --db=./test-db/ --expected_values_dir=./test-db-expected/ --max_key=100000 --ops_per_thread=1000 --sync_fault_injection=1 --clear_column_family_one_in=0 --destroy_db_initially=0 -reopen=0 -test_batches_snapshots=1 $ ./db_stress --db=./test-db/ --expected_values_dir=./test-db-expected/ --max_key=100000 --ops_per_thread=1000 --sync_fault_injection=1 --clear_column_family_one_in=0 --destroy_db_initially=0 -reopen=0 -test_batches_snapshots=0 ``` - The second problem occurred rarely in the form of a SIGSEGV on a file that was `munmap()`d. I have not seen it after this PR though this doesn't prove much. Reviewed By: jay-zhuang Differential Revision: D33155283 Pulled By: ajkr fbshipit-source-id: 66fd0f0edf34015a010c30015f14f104734e964e
2021-12-18 07:04:05 +01:00
if (FLAGS_enable_compaction_filter) {
auto* compaction_filter_factory =
reinterpret_cast<DbStressCompactionFilterFactory*>(
options_.compaction_filter_factory.get());
assert(compaction_filter_factory);
// This must be called only after any potential `SharedState::Restore()` has
// completed in order for the `compaction_filter_factory` to operate on the
// correct latest values file.
compaction_filter_factory->SetSharedState(shared);
fprintf(stdout, "Compaction filter factory: %s\n",
compaction_filter_factory->Name());
2022-01-31 22:31:00 +01:00
}
}
void StressTest::TrackExpectedState(SharedState* shared) {
if ((FLAGS_sync_fault_injection || FLAGS_disable_wal) && IsStateTracked()) {
Status s = shared->SaveAtAndAfter(db_);
if (!s.ok()) {
fprintf(stderr, "Error enabling history tracing: %s\n",
s.ToString().c_str());
exit(1);
}
Fix unsynced data loss correctness test with mixed `-test_batches_snapshots` (#9302) Summary: This fixes two bugs in the recently committed DB verification following crash-recovery with unsynced data loss (https://github.com/facebook/rocksdb/issues/8966): The first bug was in crash test runs involving mixed values for `-test_batches_snapshots`. The problem was we were neither restoring expected values nor enabling tracing when `-test_batches_snapshots=1`. This caused a future `-test_batches_snapshots=0` run to not find enough trace data to restore expected values. The fix is to restore expected values at the start of `-test_batches_snapshots=1` runs, but still leave tracing disabled as we do not need to track those KVs. The second bug was in `db_stress` runs that restore the expected values file and use compaction filter. The compaction filter was initialized to use the pre-restore expected values, which would be `munmap()`'d during `FileExpectedStateManager::Restore()`. Then compaction filter would run into a segfault. The fix is just to reorder compaction filter init after expected values restore. Pull Request resolved: https://github.com/facebook/rocksdb/pull/9302 Test Plan: - To verify the first problem, the below sequence used to fail; now it passes. ``` $ ./db_stress --db=./test-db/ --expected_values_dir=./test-db-expected/ --max_key=100000 --ops_per_thread=1000 --sync_fault_injection=1 --clear_column_family_one_in=0 --destroy_db_initially=0 -reopen=0 -test_batches_snapshots=0 $ ./db_stress --db=./test-db/ --expected_values_dir=./test-db-expected/ --max_key=100000 --ops_per_thread=1000 --sync_fault_injection=1 --clear_column_family_one_in=0 --destroy_db_initially=0 -reopen=0 -test_batches_snapshots=1 $ ./db_stress --db=./test-db/ --expected_values_dir=./test-db-expected/ --max_key=100000 --ops_per_thread=1000 --sync_fault_injection=1 --clear_column_family_one_in=0 --destroy_db_initially=0 -reopen=0 -test_batches_snapshots=0 ``` - The second problem occurred rarely in the form of a SIGSEGV on a file that was `munmap()`d. I have not seen it after this PR though this doesn't prove much. Reviewed By: jay-zhuang Differential Revision: D33155283 Pulled By: ajkr fbshipit-source-id: 66fd0f0edf34015a010c30015f14f104734e964e
2021-12-18 07:04:05 +01:00
}
}
bool StressTest::VerifySecondaries() {
#ifndef ROCKSDB_LITE
if (FLAGS_test_secondary) {
uint64_t now = clock_->NowMicros();
fprintf(stdout, "%s Start to verify secondaries against primary\n",
clock_->TimeToString(static_cast<uint64_t>(now) / 1000000).c_str());
}
for (size_t k = 0; k != secondaries_.size(); ++k) {
Status s = secondaries_[k]->TryCatchUpWithPrimary();
if (!s.ok()) {
fprintf(stderr, "Secondary failed to catch up with primary\n");
return false;
}
Add rate limiter priority to ReadOptions (#9424) Summary: Users can set the priority for file reads associated with their operation by setting `ReadOptions::rate_limiter_priority` to something other than `Env::IO_TOTAL`. Rate limiting `VerifyChecksum()` and `VerifyFileChecksums()` is the motivation for this PR, so it also includes benchmarks and minor bug fixes to get that working. `RandomAccessFileReader::Read()` already had support for rate limiting compaction reads. I changed that rate limiting to be non-specific to compaction, but rather performed according to the passed in `Env::IOPriority`. Now the compaction read rate limiting is supported by setting `rate_limiter_priority = Env::IO_LOW` on its `ReadOptions`. There is no default value for the new `Env::IOPriority` parameter to `RandomAccessFileReader::Read()`. That means this PR goes through all callers (in some cases multiple layers up the call stack) to find a `ReadOptions` to provide the priority. There are TODOs for cases I believe it would be good to let user control the priority some day (e.g., file footer reads), and no TODO in cases I believe it doesn't matter (e.g., trace file reads). The API doc only lists the missing cases where a file read associated with a provided `ReadOptions` cannot be rate limited. For cases like file ingestion checksum calculation, there is no API to provide `ReadOptions` or `Env::IOPriority`, so I didn't count that as missing. Pull Request resolved: https://github.com/facebook/rocksdb/pull/9424 Test Plan: - new unit tests - new benchmarks on ~50MB database with 1MB/s read rate limit and 100ms refill interval; verified with strace reads are chunked (at 0.1MB per chunk) and spaced roughly 100ms apart. - setup command: `./db_bench -benchmarks=fillrandom,compact -db=/tmp/testdb -target_file_size_base=1048576 -disable_auto_compactions=true -file_checksum=true` - benchmarks command: `strace -ttfe pread64 ./db_bench -benchmarks=verifychecksum,verifyfilechecksums -use_existing_db=true -db=/tmp/testdb -rate_limiter_bytes_per_sec=1048576 -rate_limit_bg_reads=1 -rate_limit_user_ops=true -file_checksum=true` - crash test using IO_USER priority on non-validation reads with https://github.com/facebook/rocksdb/issues/9567 reverted: `python3 tools/db_crashtest.py blackbox --max_key=1000000 --write_buffer_size=524288 --target_file_size_base=524288 --level_compaction_dynamic_level_bytes=true --duration=3600 --rate_limit_bg_reads=true --rate_limit_user_ops=true --rate_limiter_bytes_per_sec=10485760 --interval=10` Reviewed By: hx235 Differential Revision: D33747386 Pulled By: ajkr fbshipit-source-id: a2d985e97912fba8c54763798e04f006ccc56e0c
2022-02-17 08:17:03 +01:00
// This `ReadOptions` is for validation purposes. Ignore
// `FLAGS_rate_limit_user_ops` to avoid slowing any validation.
ReadOptions ropts;
ropts.total_order_seek = true;
// Verify only the default column family since the primary may have
// dropped other column families after most recent reopen.
std::unique_ptr<Iterator> iter1(db_->NewIterator(ropts));
std::unique_ptr<Iterator> iter2(secondaries_[k]->NewIterator(ropts));
for (iter1->SeekToFirst(), iter2->SeekToFirst();
iter1->Valid() && iter2->Valid(); iter1->Next(), iter2->Next()) {
if (iter1->key().compare(iter2->key()) != 0 ||
iter1->value().compare(iter2->value())) {
fprintf(stderr,
"Secondary %d contains different data from "
"primary.\nPrimary: %s : %s\nSecondary: %s : %s\n",
static_cast<int>(k),
iter1->key().ToString(/*hex=*/true).c_str(),
iter1->value().ToString(/*hex=*/true).c_str(),
iter2->key().ToString(/*hex=*/true).c_str(),
iter2->value().ToString(/*hex=*/true).c_str());
return false;
}
}
if (iter1->Valid() && !iter2->Valid()) {
fprintf(stderr,
"Secondary %d record count is smaller than that of primary\n",
static_cast<int>(k));
return false;
} else if (!iter1->Valid() && iter2->Valid()) {
fprintf(stderr,
"Secondary %d record count is larger than that of primary\n",
static_cast<int>(k));
return false;
}
}
if (FLAGS_test_secondary) {
uint64_t now = clock_->NowMicros();
fprintf(stdout, "%s Verification of secondaries succeeded\n",
clock_->TimeToString(static_cast<uint64_t>(now) / 1000000).c_str());
}
#endif // ROCKSDB_LITE
return true;
}
Status StressTest::AssertSame(DB* db, ColumnFamilyHandle* cf,
ThreadState::SnapshotState& snap_state) {
Status s;
if (cf->GetName() != snap_state.cf_at_name) {
return s;
}
Add rate limiter priority to ReadOptions (#9424) Summary: Users can set the priority for file reads associated with their operation by setting `ReadOptions::rate_limiter_priority` to something other than `Env::IO_TOTAL`. Rate limiting `VerifyChecksum()` and `VerifyFileChecksums()` is the motivation for this PR, so it also includes benchmarks and minor bug fixes to get that working. `RandomAccessFileReader::Read()` already had support for rate limiting compaction reads. I changed that rate limiting to be non-specific to compaction, but rather performed according to the passed in `Env::IOPriority`. Now the compaction read rate limiting is supported by setting `rate_limiter_priority = Env::IO_LOW` on its `ReadOptions`. There is no default value for the new `Env::IOPriority` parameter to `RandomAccessFileReader::Read()`. That means this PR goes through all callers (in some cases multiple layers up the call stack) to find a `ReadOptions` to provide the priority. There are TODOs for cases I believe it would be good to let user control the priority some day (e.g., file footer reads), and no TODO in cases I believe it doesn't matter (e.g., trace file reads). The API doc only lists the missing cases where a file read associated with a provided `ReadOptions` cannot be rate limited. For cases like file ingestion checksum calculation, there is no API to provide `ReadOptions` or `Env::IOPriority`, so I didn't count that as missing. Pull Request resolved: https://github.com/facebook/rocksdb/pull/9424 Test Plan: - new unit tests - new benchmarks on ~50MB database with 1MB/s read rate limit and 100ms refill interval; verified with strace reads are chunked (at 0.1MB per chunk) and spaced roughly 100ms apart. - setup command: `./db_bench -benchmarks=fillrandom,compact -db=/tmp/testdb -target_file_size_base=1048576 -disable_auto_compactions=true -file_checksum=true` - benchmarks command: `strace -ttfe pread64 ./db_bench -benchmarks=verifychecksum,verifyfilechecksums -use_existing_db=true -db=/tmp/testdb -rate_limiter_bytes_per_sec=1048576 -rate_limit_bg_reads=1 -rate_limit_user_ops=true -file_checksum=true` - crash test using IO_USER priority on non-validation reads with https://github.com/facebook/rocksdb/issues/9567 reverted: `python3 tools/db_crashtest.py blackbox --max_key=1000000 --write_buffer_size=524288 --target_file_size_base=524288 --level_compaction_dynamic_level_bytes=true --duration=3600 --rate_limit_bg_reads=true --rate_limit_user_ops=true --rate_limiter_bytes_per_sec=10485760 --interval=10` Reviewed By: hx235 Differential Revision: D33747386 Pulled By: ajkr fbshipit-source-id: a2d985e97912fba8c54763798e04f006ccc56e0c
2022-02-17 08:17:03 +01:00
// This `ReadOptions` is for validation purposes. Ignore
// `FLAGS_rate_limit_user_ops` to avoid slowing any validation.
ReadOptions ropt;
ropt.snapshot = snap_state.snapshot;
Slice ts;
if (!snap_state.timestamp.empty()) {
ts = snap_state.timestamp;
ropt.timestamp = &ts;
}
PinnableSlice exp_v(&snap_state.value);
exp_v.PinSelf();
PinnableSlice v;
s = db->Get(ropt, cf, snap_state.key, &v);
if (!s.ok() && !s.IsNotFound()) {
return s;
}
if (snap_state.status != s) {
return Status::Corruption(
"The snapshot gave inconsistent results for key " +
ToString(Hash(snap_state.key.c_str(), snap_state.key.size(), 0)) +
" in cf " + cf->GetName() + ": (" + snap_state.status.ToString() +
") vs. (" + s.ToString() + ")");
}
if (s.ok()) {
if (exp_v != v) {
return Status::Corruption("The snapshot gave inconsistent values: (" +
exp_v.ToString() + ") vs. (" + v.ToString() +
")");
}
}
if (snap_state.key_vec != nullptr) {
// When `prefix_extractor` is set, seeking to beginning and scanning
// across prefixes are only supported with `total_order_seek` set.
ropt.total_order_seek = true;
std::unique_ptr<Iterator> iterator(db->NewIterator(ropt));
std::unique_ptr<std::vector<bool>> tmp_bitvec(
new std::vector<bool>(FLAGS_max_key));
for (iterator->SeekToFirst(); iterator->Valid(); iterator->Next()) {
uint64_t key_val;
if (GetIntVal(iterator->key().ToString(), &key_val)) {
(*tmp_bitvec.get())[key_val] = true;
}
}
if (!std::equal(snap_state.key_vec->begin(), snap_state.key_vec->end(),
tmp_bitvec.get()->begin())) {
return Status::Corruption("Found inconsistent keys at this snapshot");
}
}
return Status::OK();
}
void StressTest::VerificationAbort(SharedState* shared, std::string msg,
Status s) const {
fprintf(stderr, "Verification failed: %s. Status is %s\n", msg.c_str(),
s.ToString().c_str());
shared->SetVerificationFailure();
}
void StressTest::VerificationAbort(SharedState* shared, std::string msg, int cf,
int64_t key) const {
auto key_str = Key(key);
Slice key_slice = key_str;
fprintf(stderr,
"Verification failed for column family %d key %s (%" PRIi64 "): %s\n",
cf, key_slice.ToString(true).c_str(), key, msg.c_str());
shared->SetVerificationFailure();
}
void StressTest::PrintStatistics() {
if (dbstats) {
fprintf(stdout, "STATISTICS:\n%s\n", dbstats->ToString().c_str());
}
if (dbstats_secondaries) {
fprintf(stdout, "Secondary instances STATISTICS:\n%s\n",
dbstats_secondaries->ToString().c_str());
}
}
// Currently PreloadDb has to be single-threaded.
void StressTest::PreloadDbAndReopenAsReadOnly(int64_t number_of_keys,
SharedState* shared) {
WriteOptions write_opts;
write_opts.disableWAL = FLAGS_disable_wal;
if (FLAGS_sync) {
write_opts.sync = true;
}
Rate-limit automatic WAL flush after each user write (#9607) Summary: **Context:** WAL flush is currently not rate-limited by `Options::rate_limiter`. This PR is to provide rate-limiting to auto WAL flush, the one that automatically happen after each user write operation (i.e, `Options::manual_wal_flush == false`), by adding `WriteOptions::rate_limiter_options`. Note that we are NOT rate-limiting WAL flush that do NOT automatically happen after each user write, such as `Options::manual_wal_flush == true + manual FlushWAL()` (rate-limiting multiple WAL flushes), for the benefits of: - being consistent with [ReadOptions::rate_limiter_priority](https://github.com/facebook/rocksdb/blob/7.0.fb/include/rocksdb/options.h#L515) - being able to turn off some WAL flush's rate-limiting but not all (e.g, turn off specific the WAL flush of a critical user write like a service's heartbeat) `WriteOptions::rate_limiter_options` only accept `Env::IO_USER` and `Env::IO_TOTAL` currently due to an implementation constraint. - The constraint is that we currently queue parallel writes (including WAL writes) based on FIFO policy which does not factor rate limiter priority into this layer's scheduling. If we allow lower priorities such as `Env::IO_HIGH/MID/LOW` and such writes specified with lower priorities occurs before ones specified with higher priorities (even just by a tiny bit in arrival time), the former would have blocked the latter, leading to a "priority inversion" issue and contradictory to what we promise for rate-limiting priority. Therefore we only allow `Env::IO_USER` and `Env::IO_TOTAL` right now before improving that scheduling. A pre-requisite to this feature is to support operation-level rate limiting in `WritableFileWriter`, which is also included in this PR. **Summary:** - Renamed test suite `DBRateLimiterTest to DBRateLimiterOnReadTest` for adding a new test suite - Accept `rate_limiter_priority` in `WritableFileWriter`'s private and public write functions - Passed `WriteOptions::rate_limiter_options` to `WritableFileWriter` in the path of automatic WAL flush. Pull Request resolved: https://github.com/facebook/rocksdb/pull/9607 Test Plan: - Added new unit test to verify existing flush/compaction rate-limiting does not break, since `DBTest, RateLimitingTest` is disabled and current db-level rate-limiting tests focus on read only (e.g, `db_rate_limiter_test`, `DBTest2, RateLimitedCompactionReads`). - Added new unit test `DBRateLimiterOnWriteWALTest, AutoWalFlush` - `strace -ftt -e trace=write ./db_bench -benchmarks=fillseq -db=/dev/shm/testdb -rate_limit_auto_wal_flush=1 -rate_limiter_bytes_per_sec=15 -rate_limiter_refill_period_us=1000000 -write_buffer_size=100000000 -disable_auto_compactions=1 -num=100` - verified that WAL flush(i.e, system-call _write_) were chunked into 15 bytes and each _write_ was roughly 1 second apart - verified the chunking disappeared when `-rate_limit_auto_wal_flush=0` - crash test: `python3 tools/db_crashtest.py blackbox --disable_wal=0 --rate_limit_auto_wal_flush=1 --rate_limiter_bytes_per_sec=10485760 --interval=10` killed as normal **Benchmarked on flush/compaction to ensure no performance regression:** - compaction with rate-limiting (see table 1, avg over 1280-run): pre-change: **915635 micros/op**; post-change: **907350 micros/op (improved by 0.106%)** ``` #!/bin/bash TEST_TMPDIR=/dev/shm/testdb START=1 NUM_DATA_ENTRY=8 N=10 rm -f compact_bmk_output.txt compact_bmk_output_2.txt dont_care_output.txt for i in $(eval echo "{$START..$NUM_DATA_ENTRY}") do NUM_RUN=$(($N*(2**($i-1)))) for j in $(eval echo "{$START..$NUM_RUN}") do ./db_bench --benchmarks=fillrandom -db=$TEST_TMPDIR -disable_auto_compactions=1 -write_buffer_size=6710886 > dont_care_output.txt && ./db_bench --benchmarks=compact -use_existing_db=1 -db=$TEST_TMPDIR -level0_file_num_compaction_trigger=1 -rate_limiter_bytes_per_sec=100000000 | egrep 'compact' done > compact_bmk_output.txt && awk -v NUM_RUN=$NUM_RUN '{sum+=$3;sum_sqrt+=$3^2}END{print sum/NUM_RUN, sqrt(sum_sqrt/NUM_RUN-(sum/NUM_RUN)^2)}' compact_bmk_output.txt >> compact_bmk_output_2.txt done ``` - compaction w/o rate-limiting (see table 2, avg over 640-run): pre-change: **822197 micros/op**; post-change: **823148 micros/op (regressed by 0.12%)** ``` Same as above script, except that -rate_limiter_bytes_per_sec=0 ``` - flush with rate-limiting (see table 3, avg over 320-run, run on the [patch](https://github.com/hx235/rocksdb/commit/ee5c6023a9f6533fab9afdc681568daa21da4953) to augment current db_bench ): pre-change: **745752 micros/op**; post-change: **745331 micros/op (regressed by 0.06 %)** ``` #!/bin/bash TEST_TMPDIR=/dev/shm/testdb START=1 NUM_DATA_ENTRY=8 N=10 rm -f flush_bmk_output.txt flush_bmk_output_2.txt for i in $(eval echo "{$START..$NUM_DATA_ENTRY}") do NUM_RUN=$(($N*(2**($i-1)))) for j in $(eval echo "{$START..$NUM_RUN}") do ./db_bench -db=$TEST_TMPDIR -write_buffer_size=1048576000 -num=1000000 -rate_limiter_bytes_per_sec=100000000 -benchmarks=fillseq,flush | egrep 'flush' done > flush_bmk_output.txt && awk -v NUM_RUN=$NUM_RUN '{sum+=$3;sum_sqrt+=$3^2}END{print sum/NUM_RUN, sqrt(sum_sqrt/NUM_RUN-(sum/NUM_RUN)^2)}' flush_bmk_output.txt >> flush_bmk_output_2.txt done ``` - flush w/o rate-limiting (see table 4, avg over 320-run, run on the [patch](https://github.com/hx235/rocksdb/commit/ee5c6023a9f6533fab9afdc681568daa21da4953) to augment current db_bench): pre-change: **487512 micros/op**, post-change: **485856 micors/ops (improved by 0.34%)** ``` Same as above script, except that -rate_limiter_bytes_per_sec=0 ``` | table 1 - compact with rate-limiting| #-run | (pre-change) avg micros/op | std micros/op | (post-change) avg micros/op | std micros/op | change in avg micros/op (%) -- | -- | -- | -- | -- | -- 10 | 896978 | 16046.9 | 901242 | 15670.9 | 0.475373978 20 | 893718 | 15813 | 886505 | 17544.7 | -0.8070778478 40 | 900426 | 23882.2 | 894958 | 15104.5 | -0.6072681153 80 | 906635 | 21761.5 | 903332 | 23948.3 | -0.3643141948 160 | 898632 | 21098.9 | 907583 | 21145 | 0.9960695813 3.20E+02 | 905252 | 22785.5 | 908106 | 25325.5 | 0.3152713278 6.40E+02 | 905213 | 23598.6 | 906741 | 21370.5 | 0.1688000504 **1.28E+03** | **908316** | **23533.1** | **907350** | **24626.8** | **-0.1063506533** average over #-run | 901896.25 | 21064.9625 | 901977.125 | 20592.025 | 0.008967217682 | table 2 - compact w/o rate-limiting| #-run | (pre-change) avg micros/op | std micros/op | (post-change) avg micros/op | std micros/op | change in avg micros/op (%) -- | -- | -- | -- | -- | -- 10 | 811211 | 26996.7 | 807586 | 28456.4 | -0.4468627768 20 | 815465 | 14803.7 | 814608 | 28719.7 | -0.105093413 40 | 809203 | 26187.1 | 797835 | 25492.1 | -1.404839082 80 | 822088 | 28765.3 | 822192 | 32840.4 | 0.01265071379 160 | 821719 | 36344.7 | 821664 | 29544.9 | -0.006693285661 3.20E+02 | 820921 | 27756.4 | 821403 | 28347.7 | 0.05871454135 **6.40E+02** | **822197** | **28960.6** | **823148** | **30055.1** | **0.1156657103** average over #-run | 8.18E+05 | 2.71E+04 | 8.15E+05 | 2.91E+04 | -0.25 | table 3 - flush with rate-limiting| #-run | (pre-change) avg micros/op | std micros/op | (post-change) avg micros/op | std micros/op | change in avg micros/op (%) -- | -- | -- | -- | -- | -- 10 | 741721 | 11770.8 | 740345 | 5949.76 | -0.1855144994 20 | 735169 | 3561.83 | 743199 | 9755.77 | 1.09226586 40 | 743368 | 8891.03 | 742102 | 8683.22 | -0.1703059588 80 | 742129 | 8148.51 | 743417 | 9631.58| 0.1735547324 160 | 749045 | 9757.21 | 746256 | 9191.86 | -0.3723407806 **3.20E+02** | **745752** | **9819.65** | **745331** | **9840.62** | **-0.0564530836** 6.40E+02 | 749006 | 11080.5 | 748173 | 10578.7 | -0.1112140624 average over #-run | 743741.4286 | 9004.218571 | 744117.5714 | 9090.215714 | 0.05057441238 | table 4 - flush w/o rate-limiting| #-run | (pre-change) avg micros/op | std micros/op | (post-change) avg micros/op | std micros/op | change in avg micros/op (%) -- | -- | -- | -- | -- | -- 10 | 477283 | 24719.6 | 473864 | 12379 | -0.7163464863 20 | 486743 | 20175.2 | 502296 | 23931.3 | 3.195320734 40 | 482846 | 15309.2 | 489820 | 22259.5 | 1.444352858 80 | 491490 | 21883.1 | 490071 | 23085.7 | -0.2887139108 160 | 493347 | 28074.3 | 483609 | 21211.7 | -1.973864238 **3.20E+02** | **487512** | **21401.5** | **485856** | **22195.2** | **-0.3396839462** 6.40E+02 | 490307 | 25418.6 | 485435 | 22405.2 | -0.9936631539 average over #-run | 4.87E+05 | 2.24E+04 | 4.87E+05 | 2.11E+04 | 0.00E+00 Reviewed By: ajkr Differential Revision: D34442441 Pulled By: hx235 fbshipit-source-id: 4790f13e1e5c0a95ae1d1cc93ffcf69dc6e78bdd
2022-03-08 22:19:39 +01:00
if (FLAGS_rate_limit_auto_wal_flush) {
write_opts.rate_limiter_priority = Env::IO_USER;
}
char value[100];
int cf_idx = 0;
Status s;
for (auto cfh : column_families_) {
for (int64_t k = 0; k != number_of_keys; ++k) {
std::string key_str = Key(k);
Slice key = key_str;
size_t sz = GenerateValue(0 /*value_base*/, value, sizeof(value));
Slice v(value, sz);
shared->Put(cf_idx, k, 0, true /* pending */);
if (FLAGS_use_merge) {
if (!FLAGS_use_txn) {
s = db_->Merge(write_opts, cfh, key, v);
} else {
#ifndef ROCKSDB_LITE
Transaction* txn;
s = NewTxn(write_opts, &txn);
if (s.ok()) {
s = txn->Merge(cfh, key, v);
if (s.ok()) {
s = CommitTxn(txn);
}
}
#endif
}
} else {
if (!FLAGS_use_txn) {
std::string ts_str;
Slice ts;
if (FLAGS_user_timestamp_size > 0) {
ts_str = NowNanosStr();
ts = ts_str;
Revise APIs related to user-defined timestamp (#8946) Summary: ajkr reminded me that we have a rule of not including per-kv related data in `WriteOptions`. Namely, `WriteOptions` should not include information about "what-to-write", but should just include information about "how-to-write". According to this rule, `WriteOptions::timestamp` (experimental) is clearly a violation. Therefore, this PR removes `WriteOptions::timestamp` for compliance. After the removal, we need to pass timestamp info via another set of APIs. This PR proposes a set of overloaded functions `Put(write_opts, key, value, ts)`, `Delete(write_opts, key, ts)`, and `SingleDelete(write_opts, key, ts)`. Planned to add `Write(write_opts, batch, ts)`, but its complexity made me reconsider doing it in another PR (maybe). For better checking and returning error early, we also add a new set of APIs to `WriteBatch` that take extra `timestamp` information when writing to `WriteBatch`es. These set of APIs in `WriteBatchWithIndex` are currently not supported, and are on our TODO list. Removed `WriteBatch::AssignTimestamps()` and renamed `WriteBatch::AssignTimestamp()` to `WriteBatch::UpdateTimestamps()` since this method require that all keys have space for timestamps allocated already and multiple timestamps can be updated. The constructor of `WriteBatch` now takes a fourth argument `default_cf_ts_sz` which is the timestamp size of the default column family. This will be used to allocate space when calling APIs that do not specify a column family handle. Also, updated `DB::Get()`, `DB::MultiGet()`, `DB::NewIterator()`, `DB::NewIterators()` methods, replacing some assertions about timestamp to returning Status code. Pull Request resolved: https://github.com/facebook/rocksdb/pull/8946 Test Plan: make check ./db_bench -benchmarks=fillseq,fillrandom,readrandom,readseq,deleterandom -user_timestamp_size=8 ./db_stress --user_timestamp_size=8 -nooverwritepercent=0 -test_secondary=0 -secondary_catch_up_one_in=0 -continuous_verification_interval=0 Make sure there is no perf regression by running the following ``` ./db_bench_opt -db=/dev/shm/rocksdb -use_existing_db=0 -level0_stop_writes_trigger=256 -level0_slowdown_writes_trigger=256 -level0_file_num_compaction_trigger=256 -disable_wal=1 -duration=10 -benchmarks=fillrandom ``` Before this PR ``` DB path: [/dev/shm/rocksdb] fillrandom : 1.831 micros/op 546235 ops/sec; 60.4 MB/s ``` After this PR ``` DB path: [/dev/shm/rocksdb] fillrandom : 1.820 micros/op 549404 ops/sec; 60.8 MB/s ``` Reviewed By: ltamasi Differential Revision: D33721359 Pulled By: riversand963 fbshipit-source-id: c131561534272c120ffb80711d42748d21badf09
2022-02-02 07:17:46 +01:00
s = db_->Put(write_opts, cfh, key, ts, v);
} else {
s = db_->Put(write_opts, cfh, key, v);
}
} else {
#ifndef ROCKSDB_LITE
Transaction* txn;
s = NewTxn(write_opts, &txn);
if (s.ok()) {
s = txn->Put(cfh, key, v);
if (s.ok()) {
s = CommitTxn(txn);
}
}
#endif
}
}
shared->Put(cf_idx, k, 0, false /* pending */);
if (!s.ok()) {
break;
}
}
if (!s.ok()) {
break;
}
++cf_idx;
}
if (s.ok()) {
s = db_->Flush(FlushOptions(), column_families_);
}
if (s.ok()) {
for (auto cf : column_families_) {
delete cf;
}
column_families_.clear();
delete db_;
db_ = nullptr;
#ifndef ROCKSDB_LITE
txn_db_ = nullptr;
#endif
db_preload_finished_.store(true);
auto now = clock_->NowMicros();
fprintf(stdout, "%s Reopening database in read-only\n",
clock_->TimeToString(now / 1000000).c_str());
// Reopen as read-only, can ignore all options related to updates
Open();
} else {
fprintf(stderr, "Failed to preload db");
exit(1);
}
}
Status StressTest::SetOptions(ThreadState* thread) {
assert(FLAGS_set_options_one_in > 0);
std::unordered_map<std::string, std::string> opts;
std::string name =
options_index_[thread->rand.Next() % options_index_.size()];
int value_idx = thread->rand.Next() % options_table_[name].size();
if (name == "level0_file_num_compaction_trigger" ||
name == "level0_slowdown_writes_trigger" ||
name == "level0_stop_writes_trigger") {
opts["level0_file_num_compaction_trigger"] =
options_table_["level0_file_num_compaction_trigger"][value_idx];
opts["level0_slowdown_writes_trigger"] =
options_table_["level0_slowdown_writes_trigger"][value_idx];
opts["level0_stop_writes_trigger"] =
options_table_["level0_stop_writes_trigger"][value_idx];
} else {
opts[name] = options_table_[name][value_idx];
}
int rand_cf_idx = thread->rand.Next() % FLAGS_column_families;
auto cfh = column_families_[rand_cf_idx];
return db_->SetOptions(cfh, opts);
}
#ifndef ROCKSDB_LITE
Status StressTest::NewTxn(WriteOptions& write_opts, Transaction** txn) {
if (!FLAGS_use_txn) {
return Status::InvalidArgument("NewTxn when FLAGS_use_txn is not set");
}
Improve stress test for transactions (#9568) Summary: Test only, no change to functionality. Extremely low risk of library regression. Update test key generation by maintaining existing and non-existing keys. Update db_crashtest.py to drive multiops_txn stress test for both write-committed and write-prepared. Add a make target 'blackbox_crash_test_with_multiops_txn'. Running the following commands caught the bug exposed in https://github.com/facebook/rocksdb/issues/9571. ``` $rm -rf /tmp/rocksdbtest/* $./db_stress -progress_reports=0 -test_multi_ops_txns -use_txn -clear_column_family_one_in=0 \ -column_families=1 -writepercent=0 -delpercent=0 -delrangepercent=0 -customopspercent=60 \ -readpercent=20 -prefixpercent=0 -iterpercent=20 -reopen=0 -ops_per_thread=1000 -ub_a=10000 \ -ub_c=100 -destroy_db_initially=0 -key_spaces_path=/dev/shm/key_spaces_desc -threads=32 -read_fault_one_in=0 $./db_stress -progress_reports=0 -test_multi_ops_txns -use_txn -clear_column_family_one_in=0 -column_families=1 -writepercent=0 -delpercent=0 -delrangepercent=0 -customopspercent=60 -readpercent=20 \ -prefixpercent=0 -iterpercent=20 -reopen=0 -ops_per_thread=1000 -ub_a=10000 -ub_c=100 -destroy_db_initially=0 \ -key_spaces_path=/dev/shm/key_spaces_desc -threads=32 -read_fault_one_in=0 ``` Running the following command caught a bug which will be fixed in https://github.com/facebook/rocksdb/issues/9648 . ``` $TEST_TMPDIR=/dev/shm make blackbox_crash_test_with_multiops_wc_txn ``` Pull Request resolved: https://github.com/facebook/rocksdb/pull/9568 Reviewed By: jay-zhuang Differential Revision: D34308154 Pulled By: riversand963 fbshipit-source-id: 99ff1b65c19b46c471d2f2d3b47adcd342a1b9e7
2022-03-17 03:00:04 +01:00
write_opts.disableWAL = FLAGS_disable_wal;
static std::atomic<uint64_t> txn_id = {0};
TransactionOptions txn_options;
Improve stress test for transactions (#9568) Summary: Test only, no change to functionality. Extremely low risk of library regression. Update test key generation by maintaining existing and non-existing keys. Update db_crashtest.py to drive multiops_txn stress test for both write-committed and write-prepared. Add a make target 'blackbox_crash_test_with_multiops_txn'. Running the following commands caught the bug exposed in https://github.com/facebook/rocksdb/issues/9571. ``` $rm -rf /tmp/rocksdbtest/* $./db_stress -progress_reports=0 -test_multi_ops_txns -use_txn -clear_column_family_one_in=0 \ -column_families=1 -writepercent=0 -delpercent=0 -delrangepercent=0 -customopspercent=60 \ -readpercent=20 -prefixpercent=0 -iterpercent=20 -reopen=0 -ops_per_thread=1000 -ub_a=10000 \ -ub_c=100 -destroy_db_initially=0 -key_spaces_path=/dev/shm/key_spaces_desc -threads=32 -read_fault_one_in=0 $./db_stress -progress_reports=0 -test_multi_ops_txns -use_txn -clear_column_family_one_in=0 -column_families=1 -writepercent=0 -delpercent=0 -delrangepercent=0 -customopspercent=60 -readpercent=20 \ -prefixpercent=0 -iterpercent=20 -reopen=0 -ops_per_thread=1000 -ub_a=10000 -ub_c=100 -destroy_db_initially=0 \ -key_spaces_path=/dev/shm/key_spaces_desc -threads=32 -read_fault_one_in=0 ``` Running the following command caught a bug which will be fixed in https://github.com/facebook/rocksdb/issues/9648 . ``` $TEST_TMPDIR=/dev/shm make blackbox_crash_test_with_multiops_wc_txn ``` Pull Request resolved: https://github.com/facebook/rocksdb/pull/9568 Reviewed By: jay-zhuang Differential Revision: D34308154 Pulled By: riversand963 fbshipit-source-id: 99ff1b65c19b46c471d2f2d3b47adcd342a1b9e7
2022-03-17 03:00:04 +01:00
txn_options.use_only_the_last_commit_time_batch_for_recovery =
FLAGS_use_only_the_last_commit_time_batch_for_recovery;
txn_options.lock_timeout = 600000; // 10 min
txn_options.deadlock_detect = true;
*txn = txn_db_->BeginTransaction(write_opts, txn_options);
auto istr = std::to_string(txn_id.fetch_add(1));
Status s = (*txn)->SetName("xid" + istr);
return s;
}
Status StressTest::CommitTxn(Transaction* txn) {
if (!FLAGS_use_txn) {
return Status::InvalidArgument("CommitTxn when FLAGS_use_txn is not set");
}
Status s = txn->Prepare();
if (s.ok()) {
s = txn->Commit();
}
delete txn;
return s;
}
Status StressTest::RollbackTxn(Transaction* txn) {
if (!FLAGS_use_txn) {
return Status::InvalidArgument(
"RollbackTxn when FLAGS_use_txn is not"
" set");
}
Status s = txn->Rollback();
delete txn;
return s;
}
#endif
void StressTest::OperateDb(ThreadState* thread) {
ReadOptions read_opts(FLAGS_verify_checksum, true);
Add rate limiter priority to ReadOptions (#9424) Summary: Users can set the priority for file reads associated with their operation by setting `ReadOptions::rate_limiter_priority` to something other than `Env::IO_TOTAL`. Rate limiting `VerifyChecksum()` and `VerifyFileChecksums()` is the motivation for this PR, so it also includes benchmarks and minor bug fixes to get that working. `RandomAccessFileReader::Read()` already had support for rate limiting compaction reads. I changed that rate limiting to be non-specific to compaction, but rather performed according to the passed in `Env::IOPriority`. Now the compaction read rate limiting is supported by setting `rate_limiter_priority = Env::IO_LOW` on its `ReadOptions`. There is no default value for the new `Env::IOPriority` parameter to `RandomAccessFileReader::Read()`. That means this PR goes through all callers (in some cases multiple layers up the call stack) to find a `ReadOptions` to provide the priority. There are TODOs for cases I believe it would be good to let user control the priority some day (e.g., file footer reads), and no TODO in cases I believe it doesn't matter (e.g., trace file reads). The API doc only lists the missing cases where a file read associated with a provided `ReadOptions` cannot be rate limited. For cases like file ingestion checksum calculation, there is no API to provide `ReadOptions` or `Env::IOPriority`, so I didn't count that as missing. Pull Request resolved: https://github.com/facebook/rocksdb/pull/9424 Test Plan: - new unit tests - new benchmarks on ~50MB database with 1MB/s read rate limit and 100ms refill interval; verified with strace reads are chunked (at 0.1MB per chunk) and spaced roughly 100ms apart. - setup command: `./db_bench -benchmarks=fillrandom,compact -db=/tmp/testdb -target_file_size_base=1048576 -disable_auto_compactions=true -file_checksum=true` - benchmarks command: `strace -ttfe pread64 ./db_bench -benchmarks=verifychecksum,verifyfilechecksums -use_existing_db=true -db=/tmp/testdb -rate_limiter_bytes_per_sec=1048576 -rate_limit_bg_reads=1 -rate_limit_user_ops=true -file_checksum=true` - crash test using IO_USER priority on non-validation reads with https://github.com/facebook/rocksdb/issues/9567 reverted: `python3 tools/db_crashtest.py blackbox --max_key=1000000 --write_buffer_size=524288 --target_file_size_base=524288 --level_compaction_dynamic_level_bytes=true --duration=3600 --rate_limit_bg_reads=true --rate_limit_user_ops=true --rate_limiter_bytes_per_sec=10485760 --interval=10` Reviewed By: hx235 Differential Revision: D33747386 Pulled By: ajkr fbshipit-source-id: a2d985e97912fba8c54763798e04f006ccc56e0c
2022-02-17 08:17:03 +01:00
read_opts.rate_limiter_priority =
FLAGS_rate_limit_user_ops ? Env::IO_USER : Env::IO_TOTAL;
read_opts.async_io = FLAGS_async_io;
read_opts.adaptive_readahead = FLAGS_adaptive_readahead;
WriteOptions write_opts;
Rate-limit automatic WAL flush after each user write (#9607) Summary: **Context:** WAL flush is currently not rate-limited by `Options::rate_limiter`. This PR is to provide rate-limiting to auto WAL flush, the one that automatically happen after each user write operation (i.e, `Options::manual_wal_flush == false`), by adding `WriteOptions::rate_limiter_options`. Note that we are NOT rate-limiting WAL flush that do NOT automatically happen after each user write, such as `Options::manual_wal_flush == true + manual FlushWAL()` (rate-limiting multiple WAL flushes), for the benefits of: - being consistent with [ReadOptions::rate_limiter_priority](https://github.com/facebook/rocksdb/blob/7.0.fb/include/rocksdb/options.h#L515) - being able to turn off some WAL flush's rate-limiting but not all (e.g, turn off specific the WAL flush of a critical user write like a service's heartbeat) `WriteOptions::rate_limiter_options` only accept `Env::IO_USER` and `Env::IO_TOTAL` currently due to an implementation constraint. - The constraint is that we currently queue parallel writes (including WAL writes) based on FIFO policy which does not factor rate limiter priority into this layer's scheduling. If we allow lower priorities such as `Env::IO_HIGH/MID/LOW` and such writes specified with lower priorities occurs before ones specified with higher priorities (even just by a tiny bit in arrival time), the former would have blocked the latter, leading to a "priority inversion" issue and contradictory to what we promise for rate-limiting priority. Therefore we only allow `Env::IO_USER` and `Env::IO_TOTAL` right now before improving that scheduling. A pre-requisite to this feature is to support operation-level rate limiting in `WritableFileWriter`, which is also included in this PR. **Summary:** - Renamed test suite `DBRateLimiterTest to DBRateLimiterOnReadTest` for adding a new test suite - Accept `rate_limiter_priority` in `WritableFileWriter`'s private and public write functions - Passed `WriteOptions::rate_limiter_options` to `WritableFileWriter` in the path of automatic WAL flush. Pull Request resolved: https://github.com/facebook/rocksdb/pull/9607 Test Plan: - Added new unit test to verify existing flush/compaction rate-limiting does not break, since `DBTest, RateLimitingTest` is disabled and current db-level rate-limiting tests focus on read only (e.g, `db_rate_limiter_test`, `DBTest2, RateLimitedCompactionReads`). - Added new unit test `DBRateLimiterOnWriteWALTest, AutoWalFlush` - `strace -ftt -e trace=write ./db_bench -benchmarks=fillseq -db=/dev/shm/testdb -rate_limit_auto_wal_flush=1 -rate_limiter_bytes_per_sec=15 -rate_limiter_refill_period_us=1000000 -write_buffer_size=100000000 -disable_auto_compactions=1 -num=100` - verified that WAL flush(i.e, system-call _write_) were chunked into 15 bytes and each _write_ was roughly 1 second apart - verified the chunking disappeared when `-rate_limit_auto_wal_flush=0` - crash test: `python3 tools/db_crashtest.py blackbox --disable_wal=0 --rate_limit_auto_wal_flush=1 --rate_limiter_bytes_per_sec=10485760 --interval=10` killed as normal **Benchmarked on flush/compaction to ensure no performance regression:** - compaction with rate-limiting (see table 1, avg over 1280-run): pre-change: **915635 micros/op**; post-change: **907350 micros/op (improved by 0.106%)** ``` #!/bin/bash TEST_TMPDIR=/dev/shm/testdb START=1 NUM_DATA_ENTRY=8 N=10 rm -f compact_bmk_output.txt compact_bmk_output_2.txt dont_care_output.txt for i in $(eval echo "{$START..$NUM_DATA_ENTRY}") do NUM_RUN=$(($N*(2**($i-1)))) for j in $(eval echo "{$START..$NUM_RUN}") do ./db_bench --benchmarks=fillrandom -db=$TEST_TMPDIR -disable_auto_compactions=1 -write_buffer_size=6710886 > dont_care_output.txt && ./db_bench --benchmarks=compact -use_existing_db=1 -db=$TEST_TMPDIR -level0_file_num_compaction_trigger=1 -rate_limiter_bytes_per_sec=100000000 | egrep 'compact' done > compact_bmk_output.txt && awk -v NUM_RUN=$NUM_RUN '{sum+=$3;sum_sqrt+=$3^2}END{print sum/NUM_RUN, sqrt(sum_sqrt/NUM_RUN-(sum/NUM_RUN)^2)}' compact_bmk_output.txt >> compact_bmk_output_2.txt done ``` - compaction w/o rate-limiting (see table 2, avg over 640-run): pre-change: **822197 micros/op**; post-change: **823148 micros/op (regressed by 0.12%)** ``` Same as above script, except that -rate_limiter_bytes_per_sec=0 ``` - flush with rate-limiting (see table 3, avg over 320-run, run on the [patch](https://github.com/hx235/rocksdb/commit/ee5c6023a9f6533fab9afdc681568daa21da4953) to augment current db_bench ): pre-change: **745752 micros/op**; post-change: **745331 micros/op (regressed by 0.06 %)** ``` #!/bin/bash TEST_TMPDIR=/dev/shm/testdb START=1 NUM_DATA_ENTRY=8 N=10 rm -f flush_bmk_output.txt flush_bmk_output_2.txt for i in $(eval echo "{$START..$NUM_DATA_ENTRY}") do NUM_RUN=$(($N*(2**($i-1)))) for j in $(eval echo "{$START..$NUM_RUN}") do ./db_bench -db=$TEST_TMPDIR -write_buffer_size=1048576000 -num=1000000 -rate_limiter_bytes_per_sec=100000000 -benchmarks=fillseq,flush | egrep 'flush' done > flush_bmk_output.txt && awk -v NUM_RUN=$NUM_RUN '{sum+=$3;sum_sqrt+=$3^2}END{print sum/NUM_RUN, sqrt(sum_sqrt/NUM_RUN-(sum/NUM_RUN)^2)}' flush_bmk_output.txt >> flush_bmk_output_2.txt done ``` - flush w/o rate-limiting (see table 4, avg over 320-run, run on the [patch](https://github.com/hx235/rocksdb/commit/ee5c6023a9f6533fab9afdc681568daa21da4953) to augment current db_bench): pre-change: **487512 micros/op**, post-change: **485856 micors/ops (improved by 0.34%)** ``` Same as above script, except that -rate_limiter_bytes_per_sec=0 ``` | table 1 - compact with rate-limiting| #-run | (pre-change) avg micros/op | std micros/op | (post-change) avg micros/op | std micros/op | change in avg micros/op (%) -- | -- | -- | -- | -- | -- 10 | 896978 | 16046.9 | 901242 | 15670.9 | 0.475373978 20 | 893718 | 15813 | 886505 | 17544.7 | -0.8070778478 40 | 900426 | 23882.2 | 894958 | 15104.5 | -0.6072681153 80 | 906635 | 21761.5 | 903332 | 23948.3 | -0.3643141948 160 | 898632 | 21098.9 | 907583 | 21145 | 0.9960695813 3.20E+02 | 905252 | 22785.5 | 908106 | 25325.5 | 0.3152713278 6.40E+02 | 905213 | 23598.6 | 906741 | 21370.5 | 0.1688000504 **1.28E+03** | **908316** | **23533.1** | **907350** | **24626.8** | **-0.1063506533** average over #-run | 901896.25 | 21064.9625 | 901977.125 | 20592.025 | 0.008967217682 | table 2 - compact w/o rate-limiting| #-run | (pre-change) avg micros/op | std micros/op | (post-change) avg micros/op | std micros/op | change in avg micros/op (%) -- | -- | -- | -- | -- | -- 10 | 811211 | 26996.7 | 807586 | 28456.4 | -0.4468627768 20 | 815465 | 14803.7 | 814608 | 28719.7 | -0.105093413 40 | 809203 | 26187.1 | 797835 | 25492.1 | -1.404839082 80 | 822088 | 28765.3 | 822192 | 32840.4 | 0.01265071379 160 | 821719 | 36344.7 | 821664 | 29544.9 | -0.006693285661 3.20E+02 | 820921 | 27756.4 | 821403 | 28347.7 | 0.05871454135 **6.40E+02** | **822197** | **28960.6** | **823148** | **30055.1** | **0.1156657103** average over #-run | 8.18E+05 | 2.71E+04 | 8.15E+05 | 2.91E+04 | -0.25 | table 3 - flush with rate-limiting| #-run | (pre-change) avg micros/op | std micros/op | (post-change) avg micros/op | std micros/op | change in avg micros/op (%) -- | -- | -- | -- | -- | -- 10 | 741721 | 11770.8 | 740345 | 5949.76 | -0.1855144994 20 | 735169 | 3561.83 | 743199 | 9755.77 | 1.09226586 40 | 743368 | 8891.03 | 742102 | 8683.22 | -0.1703059588 80 | 742129 | 8148.51 | 743417 | 9631.58| 0.1735547324 160 | 749045 | 9757.21 | 746256 | 9191.86 | -0.3723407806 **3.20E+02** | **745752** | **9819.65** | **745331** | **9840.62** | **-0.0564530836** 6.40E+02 | 749006 | 11080.5 | 748173 | 10578.7 | -0.1112140624 average over #-run | 743741.4286 | 9004.218571 | 744117.5714 | 9090.215714 | 0.05057441238 | table 4 - flush w/o rate-limiting| #-run | (pre-change) avg micros/op | std micros/op | (post-change) avg micros/op | std micros/op | change in avg micros/op (%) -- | -- | -- | -- | -- | -- 10 | 477283 | 24719.6 | 473864 | 12379 | -0.7163464863 20 | 486743 | 20175.2 | 502296 | 23931.3 | 3.195320734 40 | 482846 | 15309.2 | 489820 | 22259.5 | 1.444352858 80 | 491490 | 21883.1 | 490071 | 23085.7 | -0.2887139108 160 | 493347 | 28074.3 | 483609 | 21211.7 | -1.973864238 **3.20E+02** | **487512** | **21401.5** | **485856** | **22195.2** | **-0.3396839462** 6.40E+02 | 490307 | 25418.6 | 485435 | 22405.2 | -0.9936631539 average over #-run | 4.87E+05 | 2.24E+04 | 4.87E+05 | 2.11E+04 | 0.00E+00 Reviewed By: ajkr Differential Revision: D34442441 Pulled By: hx235 fbshipit-source-id: 4790f13e1e5c0a95ae1d1cc93ffcf69dc6e78bdd
2022-03-08 22:19:39 +01:00
if (FLAGS_rate_limit_auto_wal_flush) {
write_opts.rate_limiter_priority = Env::IO_USER;
}
auto shared = thread->shared;
char value[100];
std::string from_db;
if (FLAGS_sync) {
write_opts.sync = true;
}
write_opts.disableWAL = FLAGS_disable_wal;
const int prefix_bound = static_cast<int>(FLAGS_readpercent) +
static_cast<int>(FLAGS_prefixpercent);
const int write_bound = prefix_bound + static_cast<int>(FLAGS_writepercent);
const int del_bound = write_bound + static_cast<int>(FLAGS_delpercent);
const int delrange_bound =
del_bound + static_cast<int>(FLAGS_delrangepercent);
const int iterate_bound =
delrange_bound + static_cast<int>(FLAGS_iterpercent);
const uint64_t ops_per_open = FLAGS_ops_per_thread / (FLAGS_reopen + 1);
#ifndef NDEBUG
if (FLAGS_read_fault_one_in) {
fault_fs_guard->SetThreadLocalReadErrorContext(thread->shared->GetSeed(),
FLAGS_read_fault_one_in);
}
if (FLAGS_write_fault_one_in) {
IOStatus error_msg;
if (FLAGS_injest_error_severity <= 1 || FLAGS_injest_error_severity > 2) {
error_msg = IOStatus::IOError("Retryable IO Error");
error_msg.SetRetryable(true);
} else if (FLAGS_injest_error_severity == 2) {
// Ingest the fatal error
error_msg = IOStatus::IOError("Fatal IO Error");
error_msg.SetDataLoss(true);
}
std::vector<FileType> types = {FileType::kTableFile,
FileType::kDescriptorFile,
FileType::kCurrentFile};
fault_fs_guard->SetRandomWriteError(
thread->shared->GetSeed(), FLAGS_write_fault_one_in, error_msg,
/*inject_for_all_file_types=*/false, types);
}
#endif // NDEBUG
thread->stats.Start();
for (int open_cnt = 0; open_cnt <= FLAGS_reopen; ++open_cnt) {
if (thread->shared->HasVerificationFailedYet() ||
thread->shared->ShouldStopTest()) {
break;
}
if (open_cnt != 0) {
thread->stats.FinishedSingleOp();
MutexLock l(thread->shared->GetMutex());
while (!thread->snapshot_queue.empty()) {
db_->ReleaseSnapshot(thread->snapshot_queue.front().second.snapshot);
delete thread->snapshot_queue.front().second.key_vec;
thread->snapshot_queue.pop();
}
thread->shared->IncVotedReopen();
if (thread->shared->AllVotedReopen()) {
thread->shared->GetStressTest()->Reopen(thread);
thread->shared->GetCondVar()->SignalAll();
} else {
thread->shared->GetCondVar()->Wait();
}
// Commenting this out as we don't want to reset stats on each open.
// thread->stats.Start();
}
for (uint64_t i = 0; i < ops_per_open; i++) {
if (thread->shared->HasVerificationFailedYet()) {
break;
}
// Change Options
if (thread->rand.OneInOpt(FLAGS_set_options_one_in)) {
SetOptions(thread);
}
if (thread->rand.OneInOpt(FLAGS_set_in_place_one_in)) {
options_.inplace_update_support ^= options_.inplace_update_support;
}
if (thread->tid == 0 && FLAGS_verify_db_one_in > 0 &&
thread->rand.OneIn(FLAGS_verify_db_one_in)) {
ContinuouslyVerifyDb(thread);
if (thread->shared->ShouldStopTest()) {
break;
}
}
MaybeClearOneColumnFamily(thread);
if (thread->rand.OneInOpt(FLAGS_sync_wal_one_in)) {
Status s = db_->SyncWAL();
if (!s.ok() && !s.IsNotSupported()) {
fprintf(stderr, "SyncWAL() failed: %s\n", s.ToString().c_str());
}
}
int rand_column_family = thread->rand.Next() % FLAGS_column_families;
ColumnFamilyHandle* column_family = column_families_[rand_column_family];
if (thread->rand.OneInOpt(FLAGS_compact_files_one_in)) {
TestCompactFiles(thread, column_family);
}
int64_t rand_key = GenerateOneKey(thread, i);
std::string keystr = Key(rand_key);
Slice key = keystr;
std::unique_ptr<MutexLock> lock;
if (ShouldAcquireMutexOnKey()) {
lock.reset(new MutexLock(
shared->GetMutexForKey(rand_column_family, rand_key)));
}
if (thread->rand.OneInOpt(FLAGS_compact_range_one_in)) {
TestCompactRange(thread, rand_key, key, column_family);
if (thread->shared->HasVerificationFailedYet()) {
break;
}
}
std::vector<int> rand_column_families =
GenerateColumnFamilies(FLAGS_column_families, rand_column_family);
if (thread->rand.OneInOpt(FLAGS_flush_one_in)) {
Status status = TestFlush(rand_column_families);
if (!status.ok()) {
fprintf(stdout, "Unable to perform Flush(): %s\n",
status.ToString().c_str());
}
}
#ifndef ROCKSDB_LITE
// Verify GetLiveFiles with a 1 in N chance.
if (thread->rand.OneInOpt(FLAGS_get_live_files_one_in) &&
!FLAGS_write_fault_one_in) {
Status status = VerifyGetLiveFiles();
if (!status.ok()) {
VerificationAbort(shared, "VerifyGetLiveFiles status not OK", status);
}
}
// Verify GetSortedWalFiles with a 1 in N chance.
if (thread->rand.OneInOpt(FLAGS_get_sorted_wal_files_one_in)) {
Status status = VerifyGetSortedWalFiles();
if (!status.ok()) {
VerificationAbort(shared, "VerifyGetSortedWalFiles status not OK",
status);
}
}
// Verify GetCurrentWalFile with a 1 in N chance.
if (thread->rand.OneInOpt(FLAGS_get_current_wal_file_one_in)) {
Status status = VerifyGetCurrentWalFile();
if (!status.ok()) {
VerificationAbort(shared, "VerifyGetCurrentWalFile status not OK",
status);
}
}
#endif // !ROCKSDB_LITE
if (thread->rand.OneInOpt(FLAGS_pause_background_one_in)) {
Status status = TestPauseBackground(thread);
if (!status.ok()) {
VerificationAbort(
shared, "Pause/ContinueBackgroundWork status not OK", status);
}
}
#ifndef ROCKSDB_LITE
if (thread->rand.OneInOpt(FLAGS_verify_checksum_one_in)) {
Status status = db_->VerifyChecksum();
if (!status.ok()) {
VerificationAbort(shared, "VerifyChecksum status not OK", status);
}
}
if (thread->rand.OneInOpt(FLAGS_get_property_one_in)) {
TestGetProperty(thread);
}
#endif
std::vector<int64_t> rand_keys = GenerateKeys(rand_key);
if (thread->rand.OneInOpt(FLAGS_ingest_external_file_one_in)) {
TestIngestExternalFile(thread, rand_column_families, rand_keys, lock);
}
if (thread->rand.OneInOpt(FLAGS_backup_one_in)) {
// Beyond a certain DB size threshold, this test becomes heavier than
// it's worth.
uint64_t total_size = 0;
if (FLAGS_backup_max_size > 0) {
std::vector<FileAttributes> files;
db_stress_env->GetChildrenFileAttributes(FLAGS_db, &files);
for (auto& file : files) {
total_size += file.size_bytes;
}
}
if (total_size <= FLAGS_backup_max_size) {
Status s = TestBackupRestore(thread, rand_column_families, rand_keys);
if (!s.ok()) {
VerificationAbort(shared, "Backup/restore gave inconsistent state",
s);
}
}
}
if (thread->rand.OneInOpt(FLAGS_checkpoint_one_in)) {
Status s = TestCheckpoint(thread, rand_column_families, rand_keys);
if (!s.ok()) {
VerificationAbort(shared, "Checkpoint gave inconsistent state", s);
}
}
#ifndef ROCKSDB_LITE
if (thread->rand.OneInOpt(FLAGS_approximate_size_one_in)) {
Status s =
TestApproximateSize(thread, i, rand_column_families, rand_keys);
if (!s.ok()) {
VerificationAbort(shared, "ApproximateSize Failed", s);
}
}
#endif // !ROCKSDB_LITE
if (thread->rand.OneInOpt(FLAGS_acquire_snapshot_one_in)) {
TestAcquireSnapshot(thread, rand_column_family, keystr, i);
}
/*always*/ {
Status s = MaybeReleaseSnapshots(thread, i);
if (!s.ok()) {
VerificationAbort(shared, "Snapshot gave inconsistent state", s);
}
}
// Assign timestamps if necessary.
std::string read_ts_str;
std::string write_ts_str;
Slice read_ts;
Slice write_ts;
if (ShouldAcquireMutexOnKey() && FLAGS_user_timestamp_size > 0) {
read_ts_str = GenerateTimestampForRead();
read_ts = read_ts_str;
read_opts.timestamp = &read_ts;
write_ts_str = NowNanosStr();
write_ts = write_ts_str;
}
int prob_op = thread->rand.Uniform(100);
// Reset this in case we pick something other than a read op. We don't
// want to use a stale value when deciding at the beginning of the loop
// whether to vote to reopen
if (prob_op >= 0 && prob_op < static_cast<int>(FLAGS_readpercent)) {
assert(0 <= prob_op);
// OPERATION read
if (FLAGS_use_multiget) {
// Leave room for one more iteration of the loop with a single key
// batch. This is to ensure that each thread does exactly the same
// number of ops
int multiget_batch_size = static_cast<int>(
std::min(static_cast<uint64_t>(thread->rand.Uniform(64)),
FLAGS_ops_per_thread - i - 1));
// If its the last iteration, ensure that multiget_batch_size is 1
multiget_batch_size = std::max(multiget_batch_size, 1);
rand_keys = GenerateNKeys(thread, multiget_batch_size, i);
TestMultiGet(thread, read_opts, rand_column_families, rand_keys);
i += multiget_batch_size - 1;
} else {
TestGet(thread, read_opts, rand_column_families, rand_keys);
}
} else if (prob_op < prefix_bound) {
assert(static_cast<int>(FLAGS_readpercent) <= prob_op);
// OPERATION prefix scan
// keys are 8 bytes long, prefix size is FLAGS_prefix_size. There are
// (8 - FLAGS_prefix_size) bytes besides the prefix. So there will
// be 2 ^ ((8 - FLAGS_prefix_size) * 8) possible keys with the same
// prefix
TestPrefixScan(thread, read_opts, rand_column_families, rand_keys);
} else if (prob_op < write_bound) {
assert(prefix_bound <= prob_op);
// OPERATION write
TestPut(thread, write_opts, read_opts, rand_column_families, rand_keys,
value, lock);
} else if (prob_op < del_bound) {
assert(write_bound <= prob_op);
// OPERATION delete
TestDelete(thread, write_opts, rand_column_families, rand_keys, lock);
} else if (prob_op < delrange_bound) {
assert(del_bound <= prob_op);
// OPERATION delete range
TestDeleteRange(thread, write_opts, rand_column_families, rand_keys,
lock);
} else if (prob_op < iterate_bound) {
assert(delrange_bound <= prob_op);
// OPERATION iterate
int num_seeks = static_cast<int>(
std::min(static_cast<uint64_t>(thread->rand.Uniform(4)),
FLAGS_ops_per_thread - i - 1));
rand_keys = GenerateNKeys(thread, num_seeks, i);
i += num_seeks - 1;
TestIterate(thread, read_opts, rand_column_families, rand_keys);
} else {
assert(iterate_bound <= prob_op);
TestCustomOperations(thread, rand_column_families);
}
thread->stats.FinishedSingleOp();
#ifndef ROCKSDB_LITE
uint32_t tid = thread->tid;
assert(secondaries_.empty() ||
static_cast<size_t>(tid) < secondaries_.size());
if (thread->rand.OneInOpt(FLAGS_secondary_catch_up_one_in)) {
Status s = secondaries_[tid]->TryCatchUpWithPrimary();
if (!s.ok()) {
VerificationAbort(shared, "Secondary instance failed to catch up", s);
break;
}
}
#endif
}
}
while (!thread->snapshot_queue.empty()) {
db_->ReleaseSnapshot(thread->snapshot_queue.front().second.snapshot);
delete thread->snapshot_queue.front().second.key_vec;
thread->snapshot_queue.pop();
}
thread->stats.Stop();
}
#ifndef ROCKSDB_LITE
// Generated a list of keys that close to boundaries of SST keys.
// If there isn't any SST file in the DB, return empty list.
std::vector<std::string> StressTest::GetWhiteBoxKeys(ThreadState* thread,
DB* db,
ColumnFamilyHandle* cfh,
size_t num_keys) {
ColumnFamilyMetaData cfmd;
db->GetColumnFamilyMetaData(cfh, &cfmd);
std::vector<std::string> boundaries;
for (const LevelMetaData& lmd : cfmd.levels) {
for (const SstFileMetaData& sfmd : lmd.files) {
// If FLAGS_user_timestamp_size > 0, then both smallestkey and largestkey
// have timestamps.
const auto& skey = sfmd.smallestkey;
const auto& lkey = sfmd.largestkey;
assert(skey.size() >= FLAGS_user_timestamp_size);
assert(lkey.size() >= FLAGS_user_timestamp_size);
boundaries.push_back(
skey.substr(0, skey.size() - FLAGS_user_timestamp_size));
boundaries.push_back(
lkey.substr(0, lkey.size() - FLAGS_user_timestamp_size));
}
}
if (boundaries.empty()) {
return {};
}
std::vector<std::string> ret;
for (size_t j = 0; j < num_keys; j++) {
std::string k =
boundaries[thread->rand.Uniform(static_cast<int>(boundaries.size()))];
if (thread->rand.OneIn(3)) {
// Reduce one byte from the string
for (int i = static_cast<int>(k.length()) - 1; i >= 0; i--) {
uint8_t cur = k[i];
if (cur > 0) {
k[i] = static_cast<char>(cur - 1);
break;
} else if (i > 0) {
k[i] = 0xFFu;
}
}
} else if (thread->rand.OneIn(2)) {
// Add one byte to the string
for (int i = static_cast<int>(k.length()) - 1; i >= 0; i--) {
uint8_t cur = k[i];
if (cur < 255) {
k[i] = static_cast<char>(cur + 1);
break;
} else if (i > 0) {
k[i] = 0x00;
}
}
}
ret.push_back(k);
}
return ret;
}
#endif // !ROCKSDB_LITE
// Given a key K, this creates an iterator which scans to K and then
// does a random sequence of Next/Prev operations.
Status StressTest::TestIterate(ThreadState* thread,
const ReadOptions& read_opts,
const std::vector<int>& rand_column_families,
const std::vector<int64_t>& rand_keys) {
Status s;
const Snapshot* snapshot = db_->GetSnapshot();
ReadOptions readoptionscopy = read_opts;
readoptionscopy.snapshot = snapshot;
bool expect_total_order = false;
if (thread->rand.OneIn(16)) {
// When prefix extractor is used, it's useful to cover total order seek.
readoptionscopy.total_order_seek = true;
expect_total_order = true;
} else if (thread->rand.OneIn(4)) {
readoptionscopy.total_order_seek = false;
readoptionscopy.auto_prefix_mode = true;
expect_total_order = true;
} else if (options_.prefix_extractor.get() == nullptr) {
expect_total_order = true;
}
std::string upper_bound_str;
Slice upper_bound;
if (thread->rand.OneIn(16)) {
// in 1/16 chance, set a iterator upper bound
int64_t rand_upper_key = GenerateOneKey(thread, FLAGS_ops_per_thread);
upper_bound_str = Key(rand_upper_key);
upper_bound = Slice(upper_bound_str);
// uppder_bound can be smaller than seek key, but the query itself
// should not crash either.
readoptionscopy.iterate_upper_bound = &upper_bound;
}
std::string lower_bound_str;
Slice lower_bound;
if (thread->rand.OneIn(16)) {
// in 1/16 chance, enable iterator lower bound
int64_t rand_lower_key = GenerateOneKey(thread, FLAGS_ops_per_thread);
lower_bound_str = Key(rand_lower_key);
lower_bound = Slice(lower_bound_str);
// uppder_bound can be smaller than seek key, but the query itself
// should not crash either.
readoptionscopy.iterate_lower_bound = &lower_bound;
}
auto cfh = column_families_[rand_column_families[0]];
std::unique_ptr<Iterator> iter(db_->NewIterator(readoptionscopy, cfh));
std::vector<std::string> key_str;
if (thread->rand.OneIn(16)) {
// Generate keys close to lower or upper bound of SST files.
key_str = GetWhiteBoxKeys(thread, db_, cfh, rand_keys.size());
}
if (key_str.empty()) {
// If key string is not geneerated using white block keys,
// Use randomized key passe in.
for (int64_t rkey : rand_keys) {
key_str.push_back(Key(rkey));
}
}
std::string op_logs;
const size_t kOpLogsLimit = 10000;
for (const std::string& skey : key_str) {
if (op_logs.size() > kOpLogsLimit) {
// Shouldn't take too much memory for the history log. Clear it.
op_logs = "(cleared...)\n";
}
Slice key = skey;
if (readoptionscopy.iterate_upper_bound != nullptr &&
thread->rand.OneIn(2)) {
// 1/2 chance, change the upper bound.
// It is possible that it is changed without first use, but there is no
// problem with that.
int64_t rand_upper_key = GenerateOneKey(thread, FLAGS_ops_per_thread);
upper_bound_str = Key(rand_upper_key);
upper_bound = Slice(upper_bound_str);
} else if (readoptionscopy.iterate_lower_bound != nullptr &&
thread->rand.OneIn(4)) {
// 1/4 chance, change the lower bound.
// It is possible that it is changed without first use, but there is no
// problem with that.
int64_t rand_lower_key = GenerateOneKey(thread, FLAGS_ops_per_thread);
lower_bound_str = Key(rand_lower_key);
lower_bound = Slice(lower_bound_str);
}
// Record some options to op_logs;
op_logs += "total_order_seek: ";
op_logs += (readoptionscopy.total_order_seek ? "1 " : "0 ");
op_logs += "auto_prefix_mode: ";
op_logs += (readoptionscopy.auto_prefix_mode ? "1 " : "0 ");
if (readoptionscopy.iterate_upper_bound != nullptr) {
op_logs += "ub: " + upper_bound.ToString(true) + " ";
}
if (readoptionscopy.iterate_lower_bound != nullptr) {
op_logs += "lb: " + lower_bound.ToString(true) + " ";
}
// Set up an iterator and does the same without bounds and with total
// order seek and compare the results. This is to identify bugs related
// to bounds, prefix extractor or reseeking. Sometimes we are comparing
// iterators with the same set-up, and it doesn't hurt to check them
// to be equal.
Add rate limiter priority to ReadOptions (#9424) Summary: Users can set the priority for file reads associated with their operation by setting `ReadOptions::rate_limiter_priority` to something other than `Env::IO_TOTAL`. Rate limiting `VerifyChecksum()` and `VerifyFileChecksums()` is the motivation for this PR, so it also includes benchmarks and minor bug fixes to get that working. `RandomAccessFileReader::Read()` already had support for rate limiting compaction reads. I changed that rate limiting to be non-specific to compaction, but rather performed according to the passed in `Env::IOPriority`. Now the compaction read rate limiting is supported by setting `rate_limiter_priority = Env::IO_LOW` on its `ReadOptions`. There is no default value for the new `Env::IOPriority` parameter to `RandomAccessFileReader::Read()`. That means this PR goes through all callers (in some cases multiple layers up the call stack) to find a `ReadOptions` to provide the priority. There are TODOs for cases I believe it would be good to let user control the priority some day (e.g., file footer reads), and no TODO in cases I believe it doesn't matter (e.g., trace file reads). The API doc only lists the missing cases where a file read associated with a provided `ReadOptions` cannot be rate limited. For cases like file ingestion checksum calculation, there is no API to provide `ReadOptions` or `Env::IOPriority`, so I didn't count that as missing. Pull Request resolved: https://github.com/facebook/rocksdb/pull/9424 Test Plan: - new unit tests - new benchmarks on ~50MB database with 1MB/s read rate limit and 100ms refill interval; verified with strace reads are chunked (at 0.1MB per chunk) and spaced roughly 100ms apart. - setup command: `./db_bench -benchmarks=fillrandom,compact -db=/tmp/testdb -target_file_size_base=1048576 -disable_auto_compactions=true -file_checksum=true` - benchmarks command: `strace -ttfe pread64 ./db_bench -benchmarks=verifychecksum,verifyfilechecksums -use_existing_db=true -db=/tmp/testdb -rate_limiter_bytes_per_sec=1048576 -rate_limit_bg_reads=1 -rate_limit_user_ops=true -file_checksum=true` - crash test using IO_USER priority on non-validation reads with https://github.com/facebook/rocksdb/issues/9567 reverted: `python3 tools/db_crashtest.py blackbox --max_key=1000000 --write_buffer_size=524288 --target_file_size_base=524288 --level_compaction_dynamic_level_bytes=true --duration=3600 --rate_limit_bg_reads=true --rate_limit_user_ops=true --rate_limiter_bytes_per_sec=10485760 --interval=10` Reviewed By: hx235 Differential Revision: D33747386 Pulled By: ajkr fbshipit-source-id: a2d985e97912fba8c54763798e04f006ccc56e0c
2022-02-17 08:17:03 +01:00
// This `ReadOptions` is for validation purposes. Ignore
// `FLAGS_rate_limit_user_ops` to avoid slowing any validation.
ReadOptions cmp_ro;
cmp_ro.timestamp = readoptionscopy.timestamp;
cmp_ro.snapshot = snapshot;
cmp_ro.total_order_seek = true;
ColumnFamilyHandle* cmp_cfh =
GetControlCfh(thread, rand_column_families[0]);
std::unique_ptr<Iterator> cmp_iter(db_->NewIterator(cmp_ro, cmp_cfh));
bool diverged = false;
bool support_seek_first_or_last = expect_total_order;
LastIterateOp last_op;
if (support_seek_first_or_last && thread->rand.OneIn(100)) {
iter->SeekToFirst();
cmp_iter->SeekToFirst();
last_op = kLastOpSeekToFirst;
op_logs += "STF ";
} else if (support_seek_first_or_last && thread->rand.OneIn(100)) {
iter->SeekToLast();
cmp_iter->SeekToLast();
last_op = kLastOpSeekToLast;
op_logs += "STL ";
} else if (thread->rand.OneIn(8)) {
iter->SeekForPrev(key);
cmp_iter->SeekForPrev(key);
last_op = kLastOpSeekForPrev;
op_logs += "SFP " + key.ToString(true) + " ";
} else {
iter->Seek(key);
cmp_iter->Seek(key);
last_op = kLastOpSeek;
op_logs += "S " + key.ToString(true) + " ";
}
VerifyIterator(thread, cmp_cfh, readoptionscopy, iter.get(), cmp_iter.get(),
last_op, key, op_logs, &diverged);
bool no_reverse =
(FLAGS_memtablerep == "prefix_hash" && !expect_total_order);
for (uint64_t i = 0; i < FLAGS_num_iterations && iter->Valid(); i++) {
if (no_reverse || thread->rand.OneIn(2)) {
iter->Next();
if (!diverged) {
assert(cmp_iter->Valid());
cmp_iter->Next();
}
op_logs += "N";
} else {
iter->Prev();
if (!diverged) {
assert(cmp_iter->Valid());
cmp_iter->Prev();
}
op_logs += "P";
}
last_op = kLastOpNextOrPrev;
VerifyIterator(thread, cmp_cfh, readoptionscopy, iter.get(),
cmp_iter.get(), last_op, key, op_logs, &diverged);
}
if (s.ok()) {
thread->stats.AddIterations(1);
} else {
fprintf(stderr, "TestIterate error: %s\n", s.ToString().c_str());
thread->stats.AddErrors(1);
break;
}
op_logs += "; ";
}
db_->ReleaseSnapshot(snapshot);
return s;
}
#ifndef ROCKSDB_LITE
// Test the return status of GetLiveFiles.
Status StressTest::VerifyGetLiveFiles() const {
std::vector<std::string> live_file;
uint64_t manifest_size = 0;
return db_->GetLiveFiles(live_file, &manifest_size);
}
// Test the return status of GetSortedWalFiles.
Status StressTest::VerifyGetSortedWalFiles() const {
VectorLogPtr log_ptr;
return db_->GetSortedWalFiles(log_ptr);
}
// Test the return status of GetCurrentWalFile.
Status StressTest::VerifyGetCurrentWalFile() const {
std::unique_ptr<LogFile> cur_wal_file;
return db_->GetCurrentWalFile(&cur_wal_file);
}
#endif // !ROCKSDB_LITE
// Compare the two iterator, iter and cmp_iter are in the same position,
// unless iter might be made invalidate or undefined because of
// upper or lower bounds, or prefix extractor.
// Will flag failure if the verification fails.
// diverged = true if the two iterator is already diverged.
// True if verification passed, false if not.
void StressTest::VerifyIterator(ThreadState* thread,
ColumnFamilyHandle* cmp_cfh,
const ReadOptions& ro, Iterator* iter,
Iterator* cmp_iter, LastIterateOp op,
const Slice& seek_key,
const std::string& op_logs, bool* diverged) {
if (*diverged) {
return;
}
if (op == kLastOpSeekToFirst && ro.iterate_lower_bound != nullptr) {
// SeekToFirst() with lower bound is not well defined.
*diverged = true;
return;
} else if (op == kLastOpSeekToLast && ro.iterate_upper_bound != nullptr) {
// SeekToLast() with higher bound is not well defined.
*diverged = true;
return;
} else if (op == kLastOpSeek && ro.iterate_lower_bound != nullptr &&
(options_.comparator->CompareWithoutTimestamp(
*ro.iterate_lower_bound, /*a_has_ts=*/false, seek_key,
/*b_has_ts=*/false) >= 0 ||
(ro.iterate_upper_bound != nullptr &&
options_.comparator->CompareWithoutTimestamp(
*ro.iterate_lower_bound, /*a_has_ts=*/false,
*ro.iterate_upper_bound, /*b_has_ts*/ false) >= 0))) {
// Lower bound behavior is not well defined if it is larger than
// seek key or upper bound. Disable the check for now.
*diverged = true;
return;
} else if (op == kLastOpSeekForPrev && ro.iterate_upper_bound != nullptr &&
(options_.comparator->CompareWithoutTimestamp(
*ro.iterate_upper_bound, /*a_has_ts=*/false, seek_key,
/*b_has_ts=*/false) <= 0 ||
(ro.iterate_lower_bound != nullptr &&
options_.comparator->CompareWithoutTimestamp(
*ro.iterate_lower_bound, /*a_has_ts=*/false,
*ro.iterate_upper_bound, /*b_has_ts=*/false) >= 0))) {
// Uppder bound behavior is not well defined if it is smaller than
// seek key or lower bound. Disable the check for now.
*diverged = true;
return;
}
const SliceTransform* pe = (ro.total_order_seek || ro.auto_prefix_mode)
? nullptr
: options_.prefix_extractor.get();
const Comparator* cmp = options_.comparator;
if (iter->Valid() && !cmp_iter->Valid()) {
if (pe != nullptr) {
if (!pe->InDomain(seek_key)) {
// Prefix seek a non-in-domain key is undefined. Skip checking for
// this scenario.
*diverged = true;
return;
} else if (!pe->InDomain(iter->key())) {
// out of range is iterator key is not in domain anymore.
*diverged = true;
return;
} else if (pe->Transform(iter->key()) != pe->Transform(seek_key)) {
*diverged = true;
return;
}
}
fprintf(stderr,
"Control interator is invalid but iterator has key %s "
"%s\n",
iter->key().ToString(true).c_str(), op_logs.c_str());
*diverged = true;
} else if (cmp_iter->Valid()) {
// Iterator is not valid. It can be legimate if it has already been
// out of upper or lower bound, or filtered out by prefix iterator.
const Slice& total_order_key = cmp_iter->key();
if (pe != nullptr) {
if (!pe->InDomain(seek_key)) {
// Prefix seek a non-in-domain key is undefined. Skip checking for
// this scenario.
*diverged = true;
return;
}
if (!pe->InDomain(total_order_key) ||
pe->Transform(total_order_key) != pe->Transform(seek_key)) {
// If the prefix is exhausted, the only thing needs to check
// is the iterator isn't return a position in prefix.
// Either way, checking can stop from here.
*diverged = true;
if (!iter->Valid() || !pe->InDomain(iter->key()) ||
pe->Transform(iter->key()) != pe->Transform(seek_key)) {
return;
}
fprintf(stderr,
"Iterator stays in prefix but contol doesn't"
" iterator key %s control iterator key %s %s\n",
iter->key().ToString(true).c_str(),
cmp_iter->key().ToString(true).c_str(), op_logs.c_str());
}
}
// Check upper or lower bounds.
if (!*diverged) {
if ((iter->Valid() && iter->key() != cmp_iter->key()) ||
(!iter->Valid() &&
(ro.iterate_upper_bound == nullptr ||
cmp->CompareWithoutTimestamp(total_order_key, /*a_has_ts=*/false,
*ro.iterate_upper_bound,
/*b_has_ts=*/false) < 0) &&
(ro.iterate_lower_bound == nullptr ||
cmp->CompareWithoutTimestamp(total_order_key, /*a_has_ts=*/false,
*ro.iterate_lower_bound,
/*b_has_ts=*/false) > 0))) {
fprintf(stderr,
"Iterator diverged from control iterator which"
" has value %s %s\n",
total_order_key.ToString(true).c_str(), op_logs.c_str());
if (iter->Valid()) {
fprintf(stderr, "iterator has value %s\n",
iter->key().ToString(true).c_str());
} else {
fprintf(stderr, "iterator is not valid\n");
}
*diverged = true;
}
}
}
if (*diverged) {
fprintf(stderr, "Control CF %s\n", cmp_cfh->GetName().c_str());
thread->stats.AddErrors(1);
// Fail fast to preserve the DB state.
thread->shared->SetVerificationFailure();
}
}
#ifdef ROCKSDB_LITE
Status StressTest::TestBackupRestore(
ThreadState* /* thread */,
const std::vector<int>& /* rand_column_families */,
const std::vector<int64_t>& /* rand_keys */) {
assert(false);
fprintf(stderr,
"RocksDB lite does not support "
"TestBackupRestore\n");
std::terminate();
}
Status StressTest::TestCheckpoint(
ThreadState* /* thread */,
const std::vector<int>& /* rand_column_families */,
const std::vector<int64_t>& /* rand_keys */) {
assert(false);
fprintf(stderr,
"RocksDB lite does not support "
"TestCheckpoint\n");
std::terminate();
}
void StressTest::TestCompactFiles(ThreadState* /* thread */,
ColumnFamilyHandle* /* column_family */) {
assert(false);
fprintf(stderr,
"RocksDB lite does not support "
"CompactFiles\n");
std::terminate();
}
#else // ROCKSDB_LITE
Status StressTest::TestBackupRestore(
ThreadState* thread, const std::vector<int>& rand_column_families,
const std::vector<int64_t>& rand_keys) {
std::string backup_dir = FLAGS_db + "/.backup" + ToString(thread->tid);
std::string restore_dir = FLAGS_db + "/.restore" + ToString(thread->tid);
BackupEngineOptions backup_opts(backup_dir);
// For debugging, get info_log from live options
backup_opts.info_log = db_->GetDBOptions().info_log.get();
if (thread->rand.OneIn(10)) {
backup_opts.share_table_files = false;
} else {
backup_opts.share_table_files = true;
if (thread->rand.OneIn(5)) {
backup_opts.share_files_with_checksum = false;
} else {
backup_opts.share_files_with_checksum = true;
if (thread->rand.OneIn(2)) {
// old
Restore file size in backup table file names (and other cleanup) (#7400) Summary: Prior to 6.12, backup files using share_files_with_checksum had the file size encoded in the file name, after the last '\_' and before the last '.'. We considered this an implementation detail subject to change, and indeed removed this information from the file name (with an option to use old behavior) because it was considered ineffective/inefficient for file name uniqueness. However, some downstream RocksDB users were relying on this information since the file size is not explicitly in the backup manifest file. This primary purpose of this change is "retrofitting" the 6.12 release (not yet a public release) to simultaneously support the benefits of the new naming scheme (I/O performance and data correctness at scale) and preserve the file size information, both as default behaviors. With this change, we are essentially making the file size information encoded in the file name an official, though obscure, extension of the backup meta file format. We preserve an option (kLegacyCrc32cAndFileSize) to use the original "legacy" naming scheme, with its caveats, and make it easy to omit the file size information (no kFlagIncludeFileSize), for more compact file names. But note that changing the naming scheme used on an existing db and backup directory can lead to transient space amplification, as some files will be stored under two names in the shared_checksum directory. Because some backups were saved using the original 6.12 naming scheme, we offer two ways of dealing with those files: SST files generated by older 6.12 versions can either use the default naming scheme in effect when the SST files were generated (kFlagMatchInterimNaming, default, no transient space amplification) or can use a new naming scheme (no kFlagMatchInterimNaming, potential space amplification because some already stored files getting a new name). We don't have a natural way to detect which files were generated by previous 6.12 versions, but this change hacks one in by changing DB session ids to now use a more concise encoding, reducing file name length, saving ~dozen bytes from SST files, and making them visually distinct from DB ids so that they are less likely to be mixed up. Two final auxiliary notes: Recognizing that the backup file names have become a de facto part of the backup meta schema, this change makes them easier to parse and extend by putting a distinct marker, 's', before DB session ids embedded in the name. When we extend this to allow custom checksums in the name, they can get their own marker to ensure safe parsing. For backward compatibility, file size does not get a marker but is assumed for `_[0-9]+[.]` Another change from initial 6.12 default behavior is never including file custom checksum in the file name. Looking ahead to 6.13, we do not want the default behavior to cause backup space amplification for someone turning on file custom checksum checking in BackupEngine; we want that to be an easy decision. When implemented, including file custom checksums in backup file names will be a non-default option. Actual file name patterns and priorities, as regexes: kLegacyCrc32cAndFileSize OR pre-6.12 SST file -> [0-9]+_[0-9]+_[0-9]+[.]sst kFlagMatchInterimNaming set (default) AND early 6.12 SST file -> [0-9]+_[0-9a-fA-F-]+[.]sst kUseDbSessionId AND NOT kFlagIncludeFileSize -> [0-9]+_s[0-9A-Z]{20}[.]sst kUseDbSessionId AND kFlagIncludeFileSize (default) -> [0-9]+_s[0-9A-Z]{20}_[0-9]+[.]sst We might add opt-in options for more '\_' separated data in the name, but embedded file size, if present, will always be after last '\_' and before '.sst'. This change was originally applied to version 6.12. (See https://github.com/facebook/rocksdb/issues/7390) Pull Request resolved: https://github.com/facebook/rocksdb/pull/7400 Test Plan: unit tests included. Sync point callbacks are used to mimic previous version SST files. Reviewed By: ajkr Differential Revision: D23759587 Pulled By: pdillinger fbshipit-source-id: f62d8af4e0978de0a34f26288cfbe66049b70025
2020-09-17 19:22:56 +02:00
backup_opts.share_files_with_checksum_naming =
BackupEngineOptions::kLegacyCrc32cAndFileSize;
} else {
// new
backup_opts.share_files_with_checksum_naming =
BackupEngineOptions::kUseDbSessionId;
Restore file size in backup table file names (and other cleanup) (#7400) Summary: Prior to 6.12, backup files using share_files_with_checksum had the file size encoded in the file name, after the last '\_' and before the last '.'. We considered this an implementation detail subject to change, and indeed removed this information from the file name (with an option to use old behavior) because it was considered ineffective/inefficient for file name uniqueness. However, some downstream RocksDB users were relying on this information since the file size is not explicitly in the backup manifest file. This primary purpose of this change is "retrofitting" the 6.12 release (not yet a public release) to simultaneously support the benefits of the new naming scheme (I/O performance and data correctness at scale) and preserve the file size information, both as default behaviors. With this change, we are essentially making the file size information encoded in the file name an official, though obscure, extension of the backup meta file format. We preserve an option (kLegacyCrc32cAndFileSize) to use the original "legacy" naming scheme, with its caveats, and make it easy to omit the file size information (no kFlagIncludeFileSize), for more compact file names. But note that changing the naming scheme used on an existing db and backup directory can lead to transient space amplification, as some files will be stored under two names in the shared_checksum directory. Because some backups were saved using the original 6.12 naming scheme, we offer two ways of dealing with those files: SST files generated by older 6.12 versions can either use the default naming scheme in effect when the SST files were generated (kFlagMatchInterimNaming, default, no transient space amplification) or can use a new naming scheme (no kFlagMatchInterimNaming, potential space amplification because some already stored files getting a new name). We don't have a natural way to detect which files were generated by previous 6.12 versions, but this change hacks one in by changing DB session ids to now use a more concise encoding, reducing file name length, saving ~dozen bytes from SST files, and making them visually distinct from DB ids so that they are less likely to be mixed up. Two final auxiliary notes: Recognizing that the backup file names have become a de facto part of the backup meta schema, this change makes them easier to parse and extend by putting a distinct marker, 's', before DB session ids embedded in the name. When we extend this to allow custom checksums in the name, they can get their own marker to ensure safe parsing. For backward compatibility, file size does not get a marker but is assumed for `_[0-9]+[.]` Another change from initial 6.12 default behavior is never including file custom checksum in the file name. Looking ahead to 6.13, we do not want the default behavior to cause backup space amplification for someone turning on file custom checksum checking in BackupEngine; we want that to be an easy decision. When implemented, including file custom checksums in backup file names will be a non-default option. Actual file name patterns and priorities, as regexes: kLegacyCrc32cAndFileSize OR pre-6.12 SST file -> [0-9]+_[0-9]+_[0-9]+[.]sst kFlagMatchInterimNaming set (default) AND early 6.12 SST file -> [0-9]+_[0-9a-fA-F-]+[.]sst kUseDbSessionId AND NOT kFlagIncludeFileSize -> [0-9]+_s[0-9A-Z]{20}[.]sst kUseDbSessionId AND kFlagIncludeFileSize (default) -> [0-9]+_s[0-9A-Z]{20}_[0-9]+[.]sst We might add opt-in options for more '\_' separated data in the name, but embedded file size, if present, will always be after last '\_' and before '.sst'. This change was originally applied to version 6.12. (See https://github.com/facebook/rocksdb/issues/7390) Pull Request resolved: https://github.com/facebook/rocksdb/pull/7400 Test Plan: unit tests included. Sync point callbacks are used to mimic previous version SST files. Reviewed By: ajkr Differential Revision: D23759587 Pulled By: pdillinger fbshipit-source-id: f62d8af4e0978de0a34f26288cfbe66049b70025
2020-09-17 19:22:56 +02:00
}
if (thread->rand.OneIn(2)) {
backup_opts.share_files_with_checksum_naming =
backup_opts.share_files_with_checksum_naming |
BackupEngineOptions::kFlagIncludeFileSize;
Restore file size in backup table file names (and other cleanup) (#7400) Summary: Prior to 6.12, backup files using share_files_with_checksum had the file size encoded in the file name, after the last '\_' and before the last '.'. We considered this an implementation detail subject to change, and indeed removed this information from the file name (with an option to use old behavior) because it was considered ineffective/inefficient for file name uniqueness. However, some downstream RocksDB users were relying on this information since the file size is not explicitly in the backup manifest file. This primary purpose of this change is "retrofitting" the 6.12 release (not yet a public release) to simultaneously support the benefits of the new naming scheme (I/O performance and data correctness at scale) and preserve the file size information, both as default behaviors. With this change, we are essentially making the file size information encoded in the file name an official, though obscure, extension of the backup meta file format. We preserve an option (kLegacyCrc32cAndFileSize) to use the original "legacy" naming scheme, with its caveats, and make it easy to omit the file size information (no kFlagIncludeFileSize), for more compact file names. But note that changing the naming scheme used on an existing db and backup directory can lead to transient space amplification, as some files will be stored under two names in the shared_checksum directory. Because some backups were saved using the original 6.12 naming scheme, we offer two ways of dealing with those files: SST files generated by older 6.12 versions can either use the default naming scheme in effect when the SST files were generated (kFlagMatchInterimNaming, default, no transient space amplification) or can use a new naming scheme (no kFlagMatchInterimNaming, potential space amplification because some already stored files getting a new name). We don't have a natural way to detect which files were generated by previous 6.12 versions, but this change hacks one in by changing DB session ids to now use a more concise encoding, reducing file name length, saving ~dozen bytes from SST files, and making them visually distinct from DB ids so that they are less likely to be mixed up. Two final auxiliary notes: Recognizing that the backup file names have become a de facto part of the backup meta schema, this change makes them easier to parse and extend by putting a distinct marker, 's', before DB session ids embedded in the name. When we extend this to allow custom checksums in the name, they can get their own marker to ensure safe parsing. For backward compatibility, file size does not get a marker but is assumed for `_[0-9]+[.]` Another change from initial 6.12 default behavior is never including file custom checksum in the file name. Looking ahead to 6.13, we do not want the default behavior to cause backup space amplification for someone turning on file custom checksum checking in BackupEngine; we want that to be an easy decision. When implemented, including file custom checksums in backup file names will be a non-default option. Actual file name patterns and priorities, as regexes: kLegacyCrc32cAndFileSize OR pre-6.12 SST file -> [0-9]+_[0-9]+_[0-9]+[.]sst kFlagMatchInterimNaming set (default) AND early 6.12 SST file -> [0-9]+_[0-9a-fA-F-]+[.]sst kUseDbSessionId AND NOT kFlagIncludeFileSize -> [0-9]+_s[0-9A-Z]{20}[.]sst kUseDbSessionId AND kFlagIncludeFileSize (default) -> [0-9]+_s[0-9A-Z]{20}_[0-9]+[.]sst We might add opt-in options for more '\_' separated data in the name, but embedded file size, if present, will always be after last '\_' and before '.sst'. This change was originally applied to version 6.12. (See https://github.com/facebook/rocksdb/issues/7390) Pull Request resolved: https://github.com/facebook/rocksdb/pull/7400 Test Plan: unit tests included. Sync point callbacks are used to mimic previous version SST files. Reviewed By: ajkr Differential Revision: D23759587 Pulled By: pdillinger fbshipit-source-id: f62d8af4e0978de0a34f26288cfbe66049b70025
2020-09-17 19:22:56 +02:00
}
}
}
New backup meta schema, with file temperatures (#9660) Summary: The primary goal of this change is to add support for backing up and restoring (applying on restore) file temperature metadata, without committing to either the DB manifest or the FS reported "current" temperatures being exclusive "source of truth". To achieve this goal, we need to add temperature information to backup metadata, which requires updated backup meta schema. Fortunately I prepared for this in https://github.com/facebook/rocksdb/issues/8069, which began forward compatibility in version 6.19.0 for this kind of schema update. (Previously, backup meta schema was not extensible! Making this schema update public will allow some other "nice to have" features like taking backups with hard links, and avoiding crc32c checksum computation when another checksum is already available.) While schema version 2 is newly public, the default schema version is still 1. Until we change the default, users will need to set to 2 to enable features like temperature data backup+restore. New metadata like temperature information will be ignored with a warning in versions before this change and since 6.19.0. The metadata is considered ignorable because a functioning DB can be restored without it. Some detail: * Some renaming because "future schema" is now just public schema 2. * Initialize some atomics in TestFs (linter reported) * Add temperature hint support to SstFileDumper (used by BackupEngine) Pull Request resolved: https://github.com/facebook/rocksdb/pull/9660 Test Plan: related unit test majorly updated for the new functionality, including some shared testing support for tracking temperatures in a FS. Some other tests and testing hooks into production code also updated for making the backup meta schema change public. Reviewed By: ajkr Differential Revision: D34686968 Pulled By: pdillinger fbshipit-source-id: 3ac1fa3e67ee97ca8a5103d79cc87d872c1d862a
2022-03-18 19:06:17 +01:00
if (thread->rand.OneIn(2)) {
backup_opts.schema_version = 1;
} else {
backup_opts.schema_version = 2;
}
BackupEngine* backup_engine = nullptr;
std::string from = "a backup/restore operation";
Status s = BackupEngine::Open(db_stress_env, backup_opts, &backup_engine);
if (!s.ok()) {
from = "BackupEngine::Open";
}
if (s.ok()) {
New backup meta schema, with file temperatures (#9660) Summary: The primary goal of this change is to add support for backing up and restoring (applying on restore) file temperature metadata, without committing to either the DB manifest or the FS reported "current" temperatures being exclusive "source of truth". To achieve this goal, we need to add temperature information to backup metadata, which requires updated backup meta schema. Fortunately I prepared for this in https://github.com/facebook/rocksdb/issues/8069, which began forward compatibility in version 6.19.0 for this kind of schema update. (Previously, backup meta schema was not extensible! Making this schema update public will allow some other "nice to have" features like taking backups with hard links, and avoiding crc32c checksum computation when another checksum is already available.) While schema version 2 is newly public, the default schema version is still 1. Until we change the default, users will need to set to 2 to enable features like temperature data backup+restore. New metadata like temperature information will be ignored with a warning in versions before this change and since 6.19.0. The metadata is considered ignorable because a functioning DB can be restored without it. Some detail: * Some renaming because "future schema" is now just public schema 2. * Initialize some atomics in TestFs (linter reported) * Add temperature hint support to SstFileDumper (used by BackupEngine) Pull Request resolved: https://github.com/facebook/rocksdb/pull/9660 Test Plan: related unit test majorly updated for the new functionality, including some shared testing support for tracking temperatures in a FS. Some other tests and testing hooks into production code also updated for making the backup meta schema change public. Reviewed By: ajkr Differential Revision: D34686968 Pulled By: pdillinger fbshipit-source-id: 3ac1fa3e67ee97ca8a5103d79cc87d872c1d862a
2022-03-18 19:06:17 +01:00
if (backup_opts.schema_version >= 2 && thread->rand.OneIn(2)) {
TEST_BackupMetaSchemaOptions test_opts;
Begin forward compatibility for new backup meta schema (#8069) Summary: This does not add any new public APIs or published functionality, but adds the ability to read and use (and in tests, write) backups with a new meta file schema, based on the old schema but not forward-compatible (before this change). The new schema enables some capabilities not in the old: * Explicit versioning, so that users get clean error messages the next time we want to break forward compatibility. * Ignoring unrecognized fields (with warning), so that new non-critical features can be added without breaking forward compatibility. * Rejecting future "non-ignorable" fields, so that new features critical to some use-cases could potentially be added outside of linear schema versions, with broken forward compatibility. * Fields at the end of the meta file, such as for checksum of the meta file's contents (up to that point) * New optional 'size' field for each file, which is checked when present * Optionally omitting 'crc32' field, so that we aren't required to have a crc32c checksum for files to take a backup. (E.g. to support backup via hard links and to better support file custom checksums.) Because we do not have a JSON parser and to share code, the new schema is simply derived from the old schema. BackupEngine code is updated to allow missing checksums in some places, and to make that easier, `has_checksum` and `verify_checksum_after_work` are eliminated. Empty `checksum_hex` indicates checksum is unknown. I'm not too afraid of regressing on data integrity, because (a) we have pretty good test coverage of corruption detection in backups, and (b) we are increasingly relying on the DB itself for data integrity rather than it being an exclusive feature of backups. Pull Request resolved: https://github.com/facebook/rocksdb/pull/8069 Test Plan: new unit tests, added to crash test (some local run with boosted backup probability) Reviewed By: ajkr Differential Revision: D27139824 Pulled By: pdillinger fbshipit-source-id: 9e0e4decfb42bb84783d64d2d246456d97e8e8c5
2021-03-20 04:14:38 +01:00
test_opts.crc32c_checksums = thread->rand.OneIn(2) == 0;
test_opts.file_sizes = thread->rand.OneIn(2) == 0;
New backup meta schema, with file temperatures (#9660) Summary: The primary goal of this change is to add support for backing up and restoring (applying on restore) file temperature metadata, without committing to either the DB manifest or the FS reported "current" temperatures being exclusive "source of truth". To achieve this goal, we need to add temperature information to backup metadata, which requires updated backup meta schema. Fortunately I prepared for this in https://github.com/facebook/rocksdb/issues/8069, which began forward compatibility in version 6.19.0 for this kind of schema update. (Previously, backup meta schema was not extensible! Making this schema update public will allow some other "nice to have" features like taking backups with hard links, and avoiding crc32c checksum computation when another checksum is already available.) While schema version 2 is newly public, the default schema version is still 1. Until we change the default, users will need to set to 2 to enable features like temperature data backup+restore. New metadata like temperature information will be ignored with a warning in versions before this change and since 6.19.0. The metadata is considered ignorable because a functioning DB can be restored without it. Some detail: * Some renaming because "future schema" is now just public schema 2. * Initialize some atomics in TestFs (linter reported) * Add temperature hint support to SstFileDumper (used by BackupEngine) Pull Request resolved: https://github.com/facebook/rocksdb/pull/9660 Test Plan: related unit test majorly updated for the new functionality, including some shared testing support for tracking temperatures in a FS. Some other tests and testing hooks into production code also updated for making the backup meta schema change public. Reviewed By: ajkr Differential Revision: D34686968 Pulled By: pdillinger fbshipit-source-id: 3ac1fa3e67ee97ca8a5103d79cc87d872c1d862a
2022-03-18 19:06:17 +01:00
TEST_SetBackupMetaSchemaOptions(backup_engine, test_opts);
Begin forward compatibility for new backup meta schema (#8069) Summary: This does not add any new public APIs or published functionality, but adds the ability to read and use (and in tests, write) backups with a new meta file schema, based on the old schema but not forward-compatible (before this change). The new schema enables some capabilities not in the old: * Explicit versioning, so that users get clean error messages the next time we want to break forward compatibility. * Ignoring unrecognized fields (with warning), so that new non-critical features can be added without breaking forward compatibility. * Rejecting future "non-ignorable" fields, so that new features critical to some use-cases could potentially be added outside of linear schema versions, with broken forward compatibility. * Fields at the end of the meta file, such as for checksum of the meta file's contents (up to that point) * New optional 'size' field for each file, which is checked when present * Optionally omitting 'crc32' field, so that we aren't required to have a crc32c checksum for files to take a backup. (E.g. to support backup via hard links and to better support file custom checksums.) Because we do not have a JSON parser and to share code, the new schema is simply derived from the old schema. BackupEngine code is updated to allow missing checksums in some places, and to make that easier, `has_checksum` and `verify_checksum_after_work` are eliminated. Empty `checksum_hex` indicates checksum is unknown. I'm not too afraid of regressing on data integrity, because (a) we have pretty good test coverage of corruption detection in backups, and (b) we are increasingly relying on the DB itself for data integrity rather than it being an exclusive feature of backups. Pull Request resolved: https://github.com/facebook/rocksdb/pull/8069 Test Plan: new unit tests, added to crash test (some local run with boosted backup probability) Reviewed By: ajkr Differential Revision: D27139824 Pulled By: pdillinger fbshipit-source-id: 9e0e4decfb42bb84783d64d2d246456d97e8e8c5
2021-03-20 04:14:38 +01:00
}
CreateBackupOptions create_opts;
if (FLAGS_disable_wal) {
// The verification can only work when latest value of `key` is backed up,
// which requires flushing in case of WAL disabled.
//
// Note this triggers a flush with a key lock held. Meanwhile, operations
// like flush/compaction may attempt to grab key locks like in
// `DbStressCompactionFilter`. The philosophy around preventing deadlock
// is the background operation key lock acquisition only tries but does
// not wait for the lock. So here in the foreground it is OK to hold the
// lock and wait on a background operation (flush).
create_opts.flush_before_backup = true;
}
s = backup_engine->CreateNewBackup(create_opts, db_);
if (!s.ok()) {
from = "BackupEngine::CreateNewBackup";
}
}
if (s.ok()) {
delete backup_engine;
backup_engine = nullptr;
s = BackupEngine::Open(db_stress_env, backup_opts, &backup_engine);
if (!s.ok()) {
from = "BackupEngine::Open (again)";
}
}
std::vector<BackupInfo> backup_info;
Make backups openable as read-only DBs (#8142) Summary: A current limitation of backups is that you don't know the exact database state of when the backup was taken. With this new feature, you can at least inspect the backup's DB state without restoring it by opening it as a read-only DB. Rather than add something like OpenAsReadOnlyDB to the BackupEngine API, which would inhibit opening stackable DB implementations read-only (if/when their APIs support it), we instead provide a DB name and Env that can be used to open as a read-only DB. Possible follow-up work: * Add a version of GetBackupInfo for a single backup. * Let CreateNewBackup return the BackupID of the newly-created backup. Implementation details: Refactored ChrootFileSystem to split off new base class RemapFileSystem, which allows more general remapping of files. We use this base class to implement BackupEngineImpl::RemapSharedFileSystem. To minimize API impact, I decided to just add these fields `name_for_open` and `env_for_open` to those set by GetBackupInfo when include_file_details=true. Creating the RemapSharedFileSystem adds a bit to the memory consumption, perhaps unnecessarily in some cases, but this has been mitigated by (a) only initialize the RemapSharedFileSystem lazily when GetBackupInfo with include_file_details=true is called, and (b) using the existing `shared_ptr<FileInfo>` objects to hold most of the mapping data. To enhance API safety, RemapSharedFileSystem is wrapped by new ReadOnlyFileSystem which rejects any attempts to write. This uncovered a couple of places in which DB::OpenForReadOnly would write to the filesystem, so I fixed these. Added a release note because this affects logging. Additional minor refactoring in backupable_db.cc to support the new functionality. Pull Request resolved: https://github.com/facebook/rocksdb/pull/8142 Test Plan: new test (run with ASAN and UBSAN), added to stress test and ran it for a while with amplified backup_one_in Reviewed By: ajkr Differential Revision: D27535408 Pulled By: pdillinger fbshipit-source-id: 04666d310aa0261ef6b2385c43ca793ce1dfd148
2021-04-06 23:36:45 +02:00
// If inplace_not_restore, we verify the backup by opening it as a
// read-only DB. If !inplace_not_restore, we restore it to a temporary
// directory for verification.
bool inplace_not_restore = thread->rand.OneIn(3);
if (s.ok()) {
Make backups openable as read-only DBs (#8142) Summary: A current limitation of backups is that you don't know the exact database state of when the backup was taken. With this new feature, you can at least inspect the backup's DB state without restoring it by opening it as a read-only DB. Rather than add something like OpenAsReadOnlyDB to the BackupEngine API, which would inhibit opening stackable DB implementations read-only (if/when their APIs support it), we instead provide a DB name and Env that can be used to open as a read-only DB. Possible follow-up work: * Add a version of GetBackupInfo for a single backup. * Let CreateNewBackup return the BackupID of the newly-created backup. Implementation details: Refactored ChrootFileSystem to split off new base class RemapFileSystem, which allows more general remapping of files. We use this base class to implement BackupEngineImpl::RemapSharedFileSystem. To minimize API impact, I decided to just add these fields `name_for_open` and `env_for_open` to those set by GetBackupInfo when include_file_details=true. Creating the RemapSharedFileSystem adds a bit to the memory consumption, perhaps unnecessarily in some cases, but this has been mitigated by (a) only initialize the RemapSharedFileSystem lazily when GetBackupInfo with include_file_details=true is called, and (b) using the existing `shared_ptr<FileInfo>` objects to hold most of the mapping data. To enhance API safety, RemapSharedFileSystem is wrapped by new ReadOnlyFileSystem which rejects any attempts to write. This uncovered a couple of places in which DB::OpenForReadOnly would write to the filesystem, so I fixed these. Added a release note because this affects logging. Additional minor refactoring in backupable_db.cc to support the new functionality. Pull Request resolved: https://github.com/facebook/rocksdb/pull/8142 Test Plan: new test (run with ASAN and UBSAN), added to stress test and ran it for a while with amplified backup_one_in Reviewed By: ajkr Differential Revision: D27535408 Pulled By: pdillinger fbshipit-source-id: 04666d310aa0261ef6b2385c43ca793ce1dfd148
2021-04-06 23:36:45 +02:00
backup_engine->GetBackupInfo(&backup_info,
/*include_file_details*/ inplace_not_restore);
if (backup_info.empty()) {
s = Status::NotFound("no backups found");
from = "BackupEngine::GetBackupInfo";
}
}
if (s.ok() && thread->rand.OneIn(2)) {
s = backup_engine->VerifyBackup(
backup_info.front().backup_id,
thread->rand.OneIn(2) /* verify_with_checksum */);
if (!s.ok()) {
from = "BackupEngine::VerifyBackup";
}
}
const bool allow_persistent = thread->tid == 0; // not too many
bool from_latest = false;
Make backups openable as read-only DBs (#8142) Summary: A current limitation of backups is that you don't know the exact database state of when the backup was taken. With this new feature, you can at least inspect the backup's DB state without restoring it by opening it as a read-only DB. Rather than add something like OpenAsReadOnlyDB to the BackupEngine API, which would inhibit opening stackable DB implementations read-only (if/when their APIs support it), we instead provide a DB name and Env that can be used to open as a read-only DB. Possible follow-up work: * Add a version of GetBackupInfo for a single backup. * Let CreateNewBackup return the BackupID of the newly-created backup. Implementation details: Refactored ChrootFileSystem to split off new base class RemapFileSystem, which allows more general remapping of files. We use this base class to implement BackupEngineImpl::RemapSharedFileSystem. To minimize API impact, I decided to just add these fields `name_for_open` and `env_for_open` to those set by GetBackupInfo when include_file_details=true. Creating the RemapSharedFileSystem adds a bit to the memory consumption, perhaps unnecessarily in some cases, but this has been mitigated by (a) only initialize the RemapSharedFileSystem lazily when GetBackupInfo with include_file_details=true is called, and (b) using the existing `shared_ptr<FileInfo>` objects to hold most of the mapping data. To enhance API safety, RemapSharedFileSystem is wrapped by new ReadOnlyFileSystem which rejects any attempts to write. This uncovered a couple of places in which DB::OpenForReadOnly would write to the filesystem, so I fixed these. Added a release note because this affects logging. Additional minor refactoring in backupable_db.cc to support the new functionality. Pull Request resolved: https://github.com/facebook/rocksdb/pull/8142 Test Plan: new test (run with ASAN and UBSAN), added to stress test and ran it for a while with amplified backup_one_in Reviewed By: ajkr Differential Revision: D27535408 Pulled By: pdillinger fbshipit-source-id: 04666d310aa0261ef6b2385c43ca793ce1dfd148
2021-04-06 23:36:45 +02:00
int count = static_cast<int>(backup_info.size());
if (s.ok() && !inplace_not_restore) {
if (count > 1) {
s = backup_engine->RestoreDBFromBackup(
RestoreOptions(), backup_info[thread->rand.Uniform(count)].backup_id,
restore_dir /* db_dir */, restore_dir /* wal_dir */);
if (!s.ok()) {
from = "BackupEngine::RestoreDBFromBackup";
}
} else {
from_latest = true;
s = backup_engine->RestoreDBFromLatestBackup(RestoreOptions(),
restore_dir /* db_dir */,
restore_dir /* wal_dir */);
if (!s.ok()) {
from = "BackupEngine::RestoreDBFromLatestBackup";
}
}
}
Make backups openable as read-only DBs (#8142) Summary: A current limitation of backups is that you don't know the exact database state of when the backup was taken. With this new feature, you can at least inspect the backup's DB state without restoring it by opening it as a read-only DB. Rather than add something like OpenAsReadOnlyDB to the BackupEngine API, which would inhibit opening stackable DB implementations read-only (if/when their APIs support it), we instead provide a DB name and Env that can be used to open as a read-only DB. Possible follow-up work: * Add a version of GetBackupInfo for a single backup. * Let CreateNewBackup return the BackupID of the newly-created backup. Implementation details: Refactored ChrootFileSystem to split off new base class RemapFileSystem, which allows more general remapping of files. We use this base class to implement BackupEngineImpl::RemapSharedFileSystem. To minimize API impact, I decided to just add these fields `name_for_open` and `env_for_open` to those set by GetBackupInfo when include_file_details=true. Creating the RemapSharedFileSystem adds a bit to the memory consumption, perhaps unnecessarily in some cases, but this has been mitigated by (a) only initialize the RemapSharedFileSystem lazily when GetBackupInfo with include_file_details=true is called, and (b) using the existing `shared_ptr<FileInfo>` objects to hold most of the mapping data. To enhance API safety, RemapSharedFileSystem is wrapped by new ReadOnlyFileSystem which rejects any attempts to write. This uncovered a couple of places in which DB::OpenForReadOnly would write to the filesystem, so I fixed these. Added a release note because this affects logging. Additional minor refactoring in backupable_db.cc to support the new functionality. Pull Request resolved: https://github.com/facebook/rocksdb/pull/8142 Test Plan: new test (run with ASAN and UBSAN), added to stress test and ran it for a while with amplified backup_one_in Reviewed By: ajkr Differential Revision: D27535408 Pulled By: pdillinger fbshipit-source-id: 04666d310aa0261ef6b2385c43ca793ce1dfd148
2021-04-06 23:36:45 +02:00
if (s.ok() && !inplace_not_restore) {
// Purge early if restoring, to ensure the restored directory doesn't
// have some secret dependency on the backup directory.
uint32_t to_keep = 0;
if (allow_persistent) {
// allow one thread to keep up to 2 backups
to_keep = thread->rand.Uniform(3);
}
s = backup_engine->PurgeOldBackups(to_keep);
if (!s.ok()) {
from = "BackupEngine::PurgeOldBackups";
}
}
DB* restored_db = nullptr;
std::vector<ColumnFamilyHandle*> restored_cf_handles;
// Not yet implemented: opening restored BlobDB or TransactionDB
if (s.ok() && !FLAGS_use_txn && !FLAGS_use_blob_db) {
Options restore_options(options_);
restore_options.listeners.clear();
// Avoid dangling/shared file descriptors, for reliable destroy
restore_options.sst_file_manager = nullptr;
std::vector<ColumnFamilyDescriptor> cf_descriptors;
// TODO(ajkr): `column_family_names_` is not safe to access here when
// `clear_column_family_one_in != 0`. But we can't easily switch to
// `ListColumnFamilies` to get names because it won't necessarily give
// the same order as `column_family_names_`.
assert(FLAGS_clear_column_family_one_in == 0);
for (auto name : column_family_names_) {
cf_descriptors.emplace_back(name, ColumnFamilyOptions(restore_options));
}
Make backups openable as read-only DBs (#8142) Summary: A current limitation of backups is that you don't know the exact database state of when the backup was taken. With this new feature, you can at least inspect the backup's DB state without restoring it by opening it as a read-only DB. Rather than add something like OpenAsReadOnlyDB to the BackupEngine API, which would inhibit opening stackable DB implementations read-only (if/when their APIs support it), we instead provide a DB name and Env that can be used to open as a read-only DB. Possible follow-up work: * Add a version of GetBackupInfo for a single backup. * Let CreateNewBackup return the BackupID of the newly-created backup. Implementation details: Refactored ChrootFileSystem to split off new base class RemapFileSystem, which allows more general remapping of files. We use this base class to implement BackupEngineImpl::RemapSharedFileSystem. To minimize API impact, I decided to just add these fields `name_for_open` and `env_for_open` to those set by GetBackupInfo when include_file_details=true. Creating the RemapSharedFileSystem adds a bit to the memory consumption, perhaps unnecessarily in some cases, but this has been mitigated by (a) only initialize the RemapSharedFileSystem lazily when GetBackupInfo with include_file_details=true is called, and (b) using the existing `shared_ptr<FileInfo>` objects to hold most of the mapping data. To enhance API safety, RemapSharedFileSystem is wrapped by new ReadOnlyFileSystem which rejects any attempts to write. This uncovered a couple of places in which DB::OpenForReadOnly would write to the filesystem, so I fixed these. Added a release note because this affects logging. Additional minor refactoring in backupable_db.cc to support the new functionality. Pull Request resolved: https://github.com/facebook/rocksdb/pull/8142 Test Plan: new test (run with ASAN and UBSAN), added to stress test and ran it for a while with amplified backup_one_in Reviewed By: ajkr Differential Revision: D27535408 Pulled By: pdillinger fbshipit-source-id: 04666d310aa0261ef6b2385c43ca793ce1dfd148
2021-04-06 23:36:45 +02:00
if (inplace_not_restore) {
BackupInfo& info = backup_info[thread->rand.Uniform(count)];
restore_options.env = info.env_for_open.get();
s = DB::OpenForReadOnly(DBOptions(restore_options), info.name_for_open,
cf_descriptors, &restored_cf_handles,
&restored_db);
if (!s.ok()) {
from = "DB::OpenForReadOnly in backup/restore";
}
} else {
s = DB::Open(DBOptions(restore_options), restore_dir, cf_descriptors,
&restored_cf_handles, &restored_db);
if (!s.ok()) {
from = "DB::Open in backup/restore";
}
}
}
// Note the column families chosen by `rand_column_families` cannot be
// dropped while the locks for `rand_keys` are held. So we should not have
// to worry about accessing those column families throughout this function.
//
// For simplicity, currently only verifies existence/non-existence of a
// single key
for (size_t i = 0; restored_db && s.ok() && i < rand_column_families.size();
++i) {
std::string key_str = Key(rand_keys[0]);
Slice key = key_str;
std::string restored_value;
Add rate limiter priority to ReadOptions (#9424) Summary: Users can set the priority for file reads associated with their operation by setting `ReadOptions::rate_limiter_priority` to something other than `Env::IO_TOTAL`. Rate limiting `VerifyChecksum()` and `VerifyFileChecksums()` is the motivation for this PR, so it also includes benchmarks and minor bug fixes to get that working. `RandomAccessFileReader::Read()` already had support for rate limiting compaction reads. I changed that rate limiting to be non-specific to compaction, but rather performed according to the passed in `Env::IOPriority`. Now the compaction read rate limiting is supported by setting `rate_limiter_priority = Env::IO_LOW` on its `ReadOptions`. There is no default value for the new `Env::IOPriority` parameter to `RandomAccessFileReader::Read()`. That means this PR goes through all callers (in some cases multiple layers up the call stack) to find a `ReadOptions` to provide the priority. There are TODOs for cases I believe it would be good to let user control the priority some day (e.g., file footer reads), and no TODO in cases I believe it doesn't matter (e.g., trace file reads). The API doc only lists the missing cases where a file read associated with a provided `ReadOptions` cannot be rate limited. For cases like file ingestion checksum calculation, there is no API to provide `ReadOptions` or `Env::IOPriority`, so I didn't count that as missing. Pull Request resolved: https://github.com/facebook/rocksdb/pull/9424 Test Plan: - new unit tests - new benchmarks on ~50MB database with 1MB/s read rate limit and 100ms refill interval; verified with strace reads are chunked (at 0.1MB per chunk) and spaced roughly 100ms apart. - setup command: `./db_bench -benchmarks=fillrandom,compact -db=/tmp/testdb -target_file_size_base=1048576 -disable_auto_compactions=true -file_checksum=true` - benchmarks command: `strace -ttfe pread64 ./db_bench -benchmarks=verifychecksum,verifyfilechecksums -use_existing_db=true -db=/tmp/testdb -rate_limiter_bytes_per_sec=1048576 -rate_limit_bg_reads=1 -rate_limit_user_ops=true -file_checksum=true` - crash test using IO_USER priority on non-validation reads with https://github.com/facebook/rocksdb/issues/9567 reverted: `python3 tools/db_crashtest.py blackbox --max_key=1000000 --write_buffer_size=524288 --target_file_size_base=524288 --level_compaction_dynamic_level_bytes=true --duration=3600 --rate_limit_bg_reads=true --rate_limit_user_ops=true --rate_limiter_bytes_per_sec=10485760 --interval=10` Reviewed By: hx235 Differential Revision: D33747386 Pulled By: ajkr fbshipit-source-id: a2d985e97912fba8c54763798e04f006ccc56e0c
2022-02-17 08:17:03 +01:00
// This `ReadOptions` is for validation purposes. Ignore
// `FLAGS_rate_limit_user_ops` to avoid slowing any validation.
ReadOptions read_opts;
std::string ts_str;
Slice ts;
if (FLAGS_user_timestamp_size > 0) {
ts_str = GenerateTimestampForRead();
ts = ts_str;
read_opts.timestamp = &ts;
}
Status get_status = restored_db->Get(
read_opts, restored_cf_handles[rand_column_families[i]], key,
&restored_value);
bool exists = thread->shared->Exists(rand_column_families[i], rand_keys[0]);
if (get_status.ok()) {
if (!exists && from_latest && ShouldAcquireMutexOnKey()) {
s = Status::Corruption("key exists in restore but not in original db");
}
} else if (get_status.IsNotFound()) {
if (exists && from_latest && ShouldAcquireMutexOnKey()) {
s = Status::Corruption("key exists in original db but not in restore");
}
} else {
s = get_status;
if (!s.ok()) {
from = "DB::Get in backup/restore";
}
}
}
if (restored_db != nullptr) {
for (auto* cf_handle : restored_cf_handles) {
restored_db->DestroyColumnFamilyHandle(cf_handle);
}
delete restored_db;
restored_db = nullptr;
}
Make backups openable as read-only DBs (#8142) Summary: A current limitation of backups is that you don't know the exact database state of when the backup was taken. With this new feature, you can at least inspect the backup's DB state without restoring it by opening it as a read-only DB. Rather than add something like OpenAsReadOnlyDB to the BackupEngine API, which would inhibit opening stackable DB implementations read-only (if/when their APIs support it), we instead provide a DB name and Env that can be used to open as a read-only DB. Possible follow-up work: * Add a version of GetBackupInfo for a single backup. * Let CreateNewBackup return the BackupID of the newly-created backup. Implementation details: Refactored ChrootFileSystem to split off new base class RemapFileSystem, which allows more general remapping of files. We use this base class to implement BackupEngineImpl::RemapSharedFileSystem. To minimize API impact, I decided to just add these fields `name_for_open` and `env_for_open` to those set by GetBackupInfo when include_file_details=true. Creating the RemapSharedFileSystem adds a bit to the memory consumption, perhaps unnecessarily in some cases, but this has been mitigated by (a) only initialize the RemapSharedFileSystem lazily when GetBackupInfo with include_file_details=true is called, and (b) using the existing `shared_ptr<FileInfo>` objects to hold most of the mapping data. To enhance API safety, RemapSharedFileSystem is wrapped by new ReadOnlyFileSystem which rejects any attempts to write. This uncovered a couple of places in which DB::OpenForReadOnly would write to the filesystem, so I fixed these. Added a release note because this affects logging. Additional minor refactoring in backupable_db.cc to support the new functionality. Pull Request resolved: https://github.com/facebook/rocksdb/pull/8142 Test Plan: new test (run with ASAN and UBSAN), added to stress test and ran it for a while with amplified backup_one_in Reviewed By: ajkr Differential Revision: D27535408 Pulled By: pdillinger fbshipit-source-id: 04666d310aa0261ef6b2385c43ca793ce1dfd148
2021-04-06 23:36:45 +02:00
if (s.ok() && inplace_not_restore) {
// Purge late if inplace open read-only
uint32_t to_keep = 0;
if (allow_persistent) {
// allow one thread to keep up to 2 backups
to_keep = thread->rand.Uniform(3);
}
s = backup_engine->PurgeOldBackups(to_keep);
if (!s.ok()) {
from = "BackupEngine::PurgeOldBackups";
}
}
if (backup_engine != nullptr) {
delete backup_engine;
backup_engine = nullptr;
}
if (s.ok()) {
// Preserve directories on failure, or allowed persistent backup
if (!allow_persistent) {
s = DestroyDir(db_stress_env, backup_dir);
if (!s.ok()) {
from = "Destroy backup dir";
}
}
}
if (s.ok()) {
s = DestroyDir(db_stress_env, restore_dir);
if (!s.ok()) {
from = "Destroy restore dir";
}
}
if (!s.ok()) {
fprintf(stderr, "Failure in %s with: %s\n", from.c_str(),
s.ToString().c_str());
}
return s;
}
Status StressTest::TestApproximateSize(
ThreadState* thread, uint64_t iteration,
const std::vector<int>& rand_column_families,
const std::vector<int64_t>& rand_keys) {
// rand_keys likely only has one key. Just use the first one.
assert(!rand_keys.empty());
assert(!rand_column_families.empty());
int64_t key1 = rand_keys[0];
int64_t key2;
if (thread->rand.OneIn(2)) {
// Two totally random keys. This tends to cover large ranges.
key2 = GenerateOneKey(thread, iteration);
if (key2 < key1) {
std::swap(key1, key2);
}
} else {
// Unless users pass a very large FLAGS_max_key, it we should not worry
// about overflow. It is for testing, so we skip the overflow checking
// for simplicity.
key2 = key1 + static_cast<int64_t>(thread->rand.Uniform(1000));
}
std::string key1_str = Key(key1);
std::string key2_str = Key(key2);
Range range{Slice(key1_str), Slice(key2_str)};
SizeApproximationOptions sao;
sao.include_memtables = thread->rand.OneIn(2);
if (sao.include_memtables) {
sao.include_files = thread->rand.OneIn(2);
}
if (thread->rand.OneIn(2)) {
if (thread->rand.OneIn(2)) {
sao.files_size_error_margin = 0.0;
} else {
sao.files_size_error_margin =
static_cast<double>(thread->rand.Uniform(3));
}
}
uint64_t result;
return db_->GetApproximateSizes(
sao, column_families_[rand_column_families[0]], &range, 1, &result);
}
Status StressTest::TestCheckpoint(ThreadState* thread,
const std::vector<int>& rand_column_families,
const std::vector<int64_t>& rand_keys) {
std::string checkpoint_dir =
FLAGS_db + "/.checkpoint" + ToString(thread->tid);
Options tmp_opts(options_);
tmp_opts.listeners.clear();
tmp_opts.env = db_stress_env;
DestroyDB(checkpoint_dir, tmp_opts);
if (db_stress_env->FileExists(checkpoint_dir).ok()) {
// If the directory might still exist, try to delete the files one by one.
// Likely a trash file is still there.
Status my_s = DestroyDir(db_stress_env, checkpoint_dir);
if (!my_s.ok()) {
fprintf(stderr, "Fail to destory directory before checkpoint: %s",
my_s.ToString().c_str());
}
}
Checkpoint* checkpoint = nullptr;
Status s = Checkpoint::Create(db_, &checkpoint);
if (s.ok()) {
s = checkpoint->CreateCheckpoint(checkpoint_dir);
if (!s.ok()) {
fprintf(stderr, "Fail to create checkpoint to %s\n",
checkpoint_dir.c_str());
std::vector<std::string> files;
Status my_s = db_stress_env->GetChildren(checkpoint_dir, &files);
if (my_s.ok()) {
for (const auto& f : files) {
fprintf(stderr, " %s\n", f.c_str());
}
} else {
fprintf(stderr, "Fail to get files under the directory to %s\n",
my_s.ToString().c_str());
}
}
}
delete checkpoint;
checkpoint = nullptr;
std::vector<ColumnFamilyHandle*> cf_handles;
DB* checkpoint_db = nullptr;
if (s.ok()) {
Options options(options_);
options.listeners.clear();
// Avoid race condition in trash handling after delete checkpoint_db
options.sst_file_manager.reset();
std::vector<ColumnFamilyDescriptor> cf_descs;
// TODO(ajkr): `column_family_names_` is not safe to access here when
// `clear_column_family_one_in != 0`. But we can't easily switch to
// `ListColumnFamilies` to get names because it won't necessarily give
// the same order as `column_family_names_`.
assert(FLAGS_clear_column_family_one_in == 0);
if (FLAGS_clear_column_family_one_in == 0) {
for (const auto& name : column_family_names_) {
cf_descs.emplace_back(name, ColumnFamilyOptions(options));
}
s = DB::OpenForReadOnly(DBOptions(options), checkpoint_dir, cf_descs,
&cf_handles, &checkpoint_db);
}
}
if (checkpoint_db != nullptr) {
// Note the column families chosen by `rand_column_families` cannot be
// dropped while the locks for `rand_keys` are held. So we should not have
// to worry about accessing those column families throughout this function.
for (size_t i = 0; s.ok() && i < rand_column_families.size(); ++i) {
std::string key_str = Key(rand_keys[0]);
Slice key = key_str;
std::string value;
Status get_status = checkpoint_db->Get(
ReadOptions(), cf_handles[rand_column_families[i]], key, &value);
bool exists =
thread->shared->Exists(rand_column_families[i], rand_keys[0]);
if (get_status.ok()) {
if (!exists && ShouldAcquireMutexOnKey()) {
s = Status::Corruption(
"key exists in checkpoint but not in original db");
}
} else if (get_status.IsNotFound()) {
if (exists && ShouldAcquireMutexOnKey()) {
s = Status::Corruption(
"key exists in original db but not in checkpoint");
}
} else {
s = get_status;
}
}
for (auto cfh : cf_handles) {
delete cfh;
}
cf_handles.clear();
delete checkpoint_db;
checkpoint_db = nullptr;
}
if (!s.ok()) {
fprintf(stderr, "A checkpoint operation failed with: %s\n",
s.ToString().c_str());
} else {
DestroyDB(checkpoint_dir, tmp_opts);
}
return s;
}
void StressTest::TestGetProperty(ThreadState* thread) const {
std::unordered_set<std::string> levelPropertyNames = {
DB::Properties::kAggregatedTablePropertiesAtLevel,
DB::Properties::kCompressionRatioAtLevelPrefix,
DB::Properties::kNumFilesAtLevelPrefix,
};
std::unordered_set<std::string> unknownPropertyNames = {
DB::Properties::kEstimateOldestKeyTime,
DB::Properties::kOptionsStatistics,
DB::Properties::
kLiveSstFilesSizeAtTemperature, // similar to levelPropertyNames, it
// requires a number suffix
};
unknownPropertyNames.insert(levelPropertyNames.begin(),
levelPropertyNames.end());
std::string prop;
for (const auto& ppt_name_and_info : InternalStats::ppt_name_to_info) {
bool res = db_->GetProperty(ppt_name_and_info.first, &prop);
if (unknownPropertyNames.find(ppt_name_and_info.first) ==
unknownPropertyNames.end()) {
if (!res) {
fprintf(stderr, "Failed to get DB property: %s\n",
ppt_name_and_info.first.c_str());
thread->shared->SetVerificationFailure();
}
if (ppt_name_and_info.second.handle_int != nullptr) {
uint64_t prop_int;
if (!db_->GetIntProperty(ppt_name_and_info.first, &prop_int)) {
fprintf(stderr, "Failed to get Int property: %s\n",
ppt_name_and_info.first.c_str());
thread->shared->SetVerificationFailure();
}
}
if (ppt_name_and_info.second.handle_map != nullptr) {
std::map<std::string, std::string> prop_map;
if (!db_->GetMapProperty(ppt_name_and_info.first, &prop_map)) {
fprintf(stderr, "Failed to get Map property: %s\n",
ppt_name_and_info.first.c_str());
thread->shared->SetVerificationFailure();
}
}
}
}
ROCKSDB_NAMESPACE::ColumnFamilyMetaData cf_meta_data;
db_->GetColumnFamilyMetaData(&cf_meta_data);
int level_size = static_cast<int>(cf_meta_data.levels.size());
for (int level = 0; level < level_size; level++) {
for (const auto& ppt_name : levelPropertyNames) {
bool res = db_->GetProperty(ppt_name + std::to_string(level), &prop);
if (!res) {
fprintf(stderr, "Failed to get DB property: %s\n",
(ppt_name + std::to_string(level)).c_str());
thread->shared->SetVerificationFailure();
}
}
}
// Test for an invalid property name
if (thread->rand.OneIn(100)) {
if (db_->GetProperty("rocksdb.invalid_property_name", &prop)) {
fprintf(stderr, "Failed to return false for invalid property name\n");
thread->shared->SetVerificationFailure();
}
}
}
void StressTest::TestCompactFiles(ThreadState* thread,
ColumnFamilyHandle* column_family) {
ROCKSDB_NAMESPACE::ColumnFamilyMetaData cf_meta_data;
db_->GetColumnFamilyMetaData(column_family, &cf_meta_data);
if (cf_meta_data.levels.empty()) {
return;
}
// Randomly compact up to three consecutive files from a level
const int kMaxRetry = 3;
for (int attempt = 0; attempt < kMaxRetry; ++attempt) {
size_t random_level =
thread->rand.Uniform(static_cast<int>(cf_meta_data.levels.size()));
const auto& files = cf_meta_data.levels[random_level].files;
if (files.size() > 0) {
size_t random_file_index =
thread->rand.Uniform(static_cast<int>(files.size()));
if (files[random_file_index].being_compacted) {
// Retry as the selected file is currently being compacted
continue;
}
std::vector<std::string> input_files;
input_files.push_back(files[random_file_index].name);
if (random_file_index > 0 &&
!files[random_file_index - 1].being_compacted) {
input_files.push_back(files[random_file_index - 1].name);
}
if (random_file_index + 1 < files.size() &&
!files[random_file_index + 1].being_compacted) {
input_files.push_back(files[random_file_index + 1].name);
}
size_t output_level =
std::min(random_level + 1, cf_meta_data.levels.size() - 1);
auto s = db_->CompactFiles(CompactionOptions(), column_family,
input_files, static_cast<int>(output_level));
if (!s.ok()) {
fprintf(stdout, "Unable to perform CompactFiles(): %s\n",
s.ToString().c_str());
thread->stats.AddNumCompactFilesFailed(1);
} else {
thread->stats.AddNumCompactFilesSucceed(1);
}
break;
}
}
}
#endif // ROCKSDB_LITE
Status StressTest::TestFlush(const std::vector<int>& rand_column_families) {
FlushOptions flush_opts;
if (FLAGS_atomic_flush) {
return db_->Flush(flush_opts, column_families_);
}
std::vector<ColumnFamilyHandle*> cfhs;
std::for_each(rand_column_families.begin(), rand_column_families.end(),
[this, &cfhs](int k) { cfhs.push_back(column_families_[k]); });
return db_->Flush(flush_opts, cfhs);
}
Status StressTest::TestPauseBackground(ThreadState* thread) {
Status status = db_->PauseBackgroundWork();
if (!status.ok()) {
return status;
}
// To avoid stalling/deadlocking ourself in this thread, just
// sleep here during pause and let other threads do db operations.
// Sleep up to ~16 seconds (2**24 microseconds), but very skewed
// toward short pause. (1 chance in 25 of pausing >= 1s;
// 1 chance in 625 of pausing full 16s.)
int pwr2_micros =
std::min(thread->rand.Uniform(25), thread->rand.Uniform(25));
clock_->SleepForMicroseconds(1 << pwr2_micros);
return db_->ContinueBackgroundWork();
}
void StressTest::TestAcquireSnapshot(ThreadState* thread,
int rand_column_family,
const std::string& keystr, uint64_t i) {
Slice key = keystr;
ColumnFamilyHandle* column_family = column_families_[rand_column_family];
Add rate limiter priority to ReadOptions (#9424) Summary: Users can set the priority for file reads associated with their operation by setting `ReadOptions::rate_limiter_priority` to something other than `Env::IO_TOTAL`. Rate limiting `VerifyChecksum()` and `VerifyFileChecksums()` is the motivation for this PR, so it also includes benchmarks and minor bug fixes to get that working. `RandomAccessFileReader::Read()` already had support for rate limiting compaction reads. I changed that rate limiting to be non-specific to compaction, but rather performed according to the passed in `Env::IOPriority`. Now the compaction read rate limiting is supported by setting `rate_limiter_priority = Env::IO_LOW` on its `ReadOptions`. There is no default value for the new `Env::IOPriority` parameter to `RandomAccessFileReader::Read()`. That means this PR goes through all callers (in some cases multiple layers up the call stack) to find a `ReadOptions` to provide the priority. There are TODOs for cases I believe it would be good to let user control the priority some day (e.g., file footer reads), and no TODO in cases I believe it doesn't matter (e.g., trace file reads). The API doc only lists the missing cases where a file read associated with a provided `ReadOptions` cannot be rate limited. For cases like file ingestion checksum calculation, there is no API to provide `ReadOptions` or `Env::IOPriority`, so I didn't count that as missing. Pull Request resolved: https://github.com/facebook/rocksdb/pull/9424 Test Plan: - new unit tests - new benchmarks on ~50MB database with 1MB/s read rate limit and 100ms refill interval; verified with strace reads are chunked (at 0.1MB per chunk) and spaced roughly 100ms apart. - setup command: `./db_bench -benchmarks=fillrandom,compact -db=/tmp/testdb -target_file_size_base=1048576 -disable_auto_compactions=true -file_checksum=true` - benchmarks command: `strace -ttfe pread64 ./db_bench -benchmarks=verifychecksum,verifyfilechecksums -use_existing_db=true -db=/tmp/testdb -rate_limiter_bytes_per_sec=1048576 -rate_limit_bg_reads=1 -rate_limit_user_ops=true -file_checksum=true` - crash test using IO_USER priority on non-validation reads with https://github.com/facebook/rocksdb/issues/9567 reverted: `python3 tools/db_crashtest.py blackbox --max_key=1000000 --write_buffer_size=524288 --target_file_size_base=524288 --level_compaction_dynamic_level_bytes=true --duration=3600 --rate_limit_bg_reads=true --rate_limit_user_ops=true --rate_limiter_bytes_per_sec=10485760 --interval=10` Reviewed By: hx235 Differential Revision: D33747386 Pulled By: ajkr fbshipit-source-id: a2d985e97912fba8c54763798e04f006ccc56e0c
2022-02-17 08:17:03 +01:00
// This `ReadOptions` is for validation purposes. Ignore
// `FLAGS_rate_limit_user_ops` to avoid slowing any validation.
ReadOptions ropt;
#ifndef ROCKSDB_LITE
auto db_impl = static_cast_with_check<DBImpl>(db_->GetRootDB());
const bool ww_snapshot = thread->rand.OneIn(10);
const Snapshot* snapshot =
ww_snapshot ? db_impl->GetSnapshotForWriteConflictBoundary()
: db_->GetSnapshot();
#else
const Snapshot* snapshot = db_->GetSnapshot();
#endif // !ROCKSDB_LITE
ropt.snapshot = snapshot;
// Ideally, we want snapshot taking and timestamp generation to be atomic
// here, so that the snapshot corresponds to the timestamp. However, it is
// not possible with current GetSnapshot() API.
std::string ts_str;
Slice ts;
if (FLAGS_user_timestamp_size > 0) {
ts_str = GenerateTimestampForRead();
ts = ts_str;
ropt.timestamp = &ts;
}
std::string value_at;
// When taking a snapshot, we also read a key from that snapshot. We
// will later read the same key before releasing the snapshot and
// verify that the results are the same.
auto status_at = db_->Get(ropt, column_family, key, &value_at);
std::vector<bool>* key_vec = nullptr;
if (FLAGS_compare_full_db_state_snapshot && (thread->tid == 0)) {
key_vec = new std::vector<bool>(FLAGS_max_key);
// When `prefix_extractor` is set, seeking to beginning and scanning
// across prefixes are only supported with `total_order_seek` set.
ropt.total_order_seek = true;
std::unique_ptr<Iterator> iterator(db_->NewIterator(ropt));
for (iterator->SeekToFirst(); iterator->Valid(); iterator->Next()) {
uint64_t key_val;
if (GetIntVal(iterator->key().ToString(), &key_val)) {
(*key_vec)[key_val] = true;
}
}
}
ThreadState::SnapshotState snap_state = {snapshot,
rand_column_family,
column_family->GetName(),
keystr,
status_at,
value_at,
key_vec,
ts_str};
uint64_t hold_for = FLAGS_snapshot_hold_ops;
if (FLAGS_long_running_snapshots) {
// Hold 10% of snapshots for 10x more
if (thread->rand.OneIn(10)) {
assert(hold_for < port::kMaxInt64 / 10);
hold_for *= 10;
// Hold 1% of snapshots for 100x more
if (thread->rand.OneIn(10)) {
assert(hold_for < port::kMaxInt64 / 10);
hold_for *= 10;
}
}
}
uint64_t release_at = std::min(FLAGS_ops_per_thread - 1, i + hold_for);
thread->snapshot_queue.emplace(release_at, snap_state);
}
Status StressTest::MaybeReleaseSnapshots(ThreadState* thread, uint64_t i) {
while (!thread->snapshot_queue.empty() &&
i >= thread->snapshot_queue.front().first) {
auto snap_state = thread->snapshot_queue.front().second;
assert(snap_state.snapshot);
// Note: this is unsafe as the cf might be dropped concurrently. But
// it is ok since unclean cf drop is cunnrently not supported by write
// prepared transactions.
Status s = AssertSame(db_, column_families_[snap_state.cf_at], snap_state);
db_->ReleaseSnapshot(snap_state.snapshot);
delete snap_state.key_vec;
thread->snapshot_queue.pop();
if (!s.ok()) {
return s;
}
}
return Status::OK();
}
void StressTest::TestCompactRange(ThreadState* thread, int64_t rand_key,
const Slice& start_key,
ColumnFamilyHandle* column_family) {
int64_t end_key_num;
if (port::kMaxInt64 - rand_key < FLAGS_compact_range_width) {
end_key_num = port::kMaxInt64;
} else {
end_key_num = FLAGS_compact_range_width + rand_key;
}
std::string end_key_buf = Key(end_key_num);
Slice end_key(end_key_buf);
CompactRangeOptions cro;
cro.exclusive_manual_compaction = static_cast<bool>(thread->rand.Next() % 2);
cro.change_level = static_cast<bool>(thread->rand.Next() % 2);
std::vector<BottommostLevelCompaction> bottom_level_styles = {
BottommostLevelCompaction::kSkip,
BottommostLevelCompaction::kIfHaveCompactionFilter,
BottommostLevelCompaction::kForce,
BottommostLevelCompaction::kForceOptimized};
cro.bottommost_level_compaction =
bottom_level_styles[thread->rand.Next() %
static_cast<uint32_t>(bottom_level_styles.size())];
cro.allow_write_stall = static_cast<bool>(thread->rand.Next() % 2);
cro.max_subcompactions = static_cast<uint32_t>(thread->rand.Next() % 4);
const Snapshot* pre_snapshot = nullptr;
uint32_t pre_hash = 0;
if (thread->rand.OneIn(2)) {
// Do some validation by declaring a snapshot and compare the data before
// and after the compaction
pre_snapshot = db_->GetSnapshot();
pre_hash =
GetRangeHash(thread, pre_snapshot, column_family, start_key, end_key);
}
Status status = db_->CompactRange(cro, column_family, &start_key, &end_key);
if (!status.ok()) {
fprintf(stdout, "Unable to perform CompactRange(): %s\n",
status.ToString().c_str());
}
if (pre_snapshot != nullptr) {
uint32_t post_hash =
GetRangeHash(thread, pre_snapshot, column_family, start_key, end_key);
if (pre_hash != post_hash) {
fprintf(stderr,
"Data hash different before and after compact range "
"start_key %s end_key %s\n",
start_key.ToString(true).c_str(), end_key.ToString(true).c_str());
thread->stats.AddErrors(1);
// Fail fast to preserve the DB state.
thread->shared->SetVerificationFailure();
}
db_->ReleaseSnapshot(pre_snapshot);
}
}
uint32_t StressTest::GetRangeHash(ThreadState* thread, const Snapshot* snapshot,
ColumnFamilyHandle* column_family,
const Slice& start_key,
const Slice& end_key) {
const std::string kCrcCalculatorSepearator = ";";
uint32_t crc = 0;
Add rate limiter priority to ReadOptions (#9424) Summary: Users can set the priority for file reads associated with their operation by setting `ReadOptions::rate_limiter_priority` to something other than `Env::IO_TOTAL`. Rate limiting `VerifyChecksum()` and `VerifyFileChecksums()` is the motivation for this PR, so it also includes benchmarks and minor bug fixes to get that working. `RandomAccessFileReader::Read()` already had support for rate limiting compaction reads. I changed that rate limiting to be non-specific to compaction, but rather performed according to the passed in `Env::IOPriority`. Now the compaction read rate limiting is supported by setting `rate_limiter_priority = Env::IO_LOW` on its `ReadOptions`. There is no default value for the new `Env::IOPriority` parameter to `RandomAccessFileReader::Read()`. That means this PR goes through all callers (in some cases multiple layers up the call stack) to find a `ReadOptions` to provide the priority. There are TODOs for cases I believe it would be good to let user control the priority some day (e.g., file footer reads), and no TODO in cases I believe it doesn't matter (e.g., trace file reads). The API doc only lists the missing cases where a file read associated with a provided `ReadOptions` cannot be rate limited. For cases like file ingestion checksum calculation, there is no API to provide `ReadOptions` or `Env::IOPriority`, so I didn't count that as missing. Pull Request resolved: https://github.com/facebook/rocksdb/pull/9424 Test Plan: - new unit tests - new benchmarks on ~50MB database with 1MB/s read rate limit and 100ms refill interval; verified with strace reads are chunked (at 0.1MB per chunk) and spaced roughly 100ms apart. - setup command: `./db_bench -benchmarks=fillrandom,compact -db=/tmp/testdb -target_file_size_base=1048576 -disable_auto_compactions=true -file_checksum=true` - benchmarks command: `strace -ttfe pread64 ./db_bench -benchmarks=verifychecksum,verifyfilechecksums -use_existing_db=true -db=/tmp/testdb -rate_limiter_bytes_per_sec=1048576 -rate_limit_bg_reads=1 -rate_limit_user_ops=true -file_checksum=true` - crash test using IO_USER priority on non-validation reads with https://github.com/facebook/rocksdb/issues/9567 reverted: `python3 tools/db_crashtest.py blackbox --max_key=1000000 --write_buffer_size=524288 --target_file_size_base=524288 --level_compaction_dynamic_level_bytes=true --duration=3600 --rate_limit_bg_reads=true --rate_limit_user_ops=true --rate_limiter_bytes_per_sec=10485760 --interval=10` Reviewed By: hx235 Differential Revision: D33747386 Pulled By: ajkr fbshipit-source-id: a2d985e97912fba8c54763798e04f006ccc56e0c
2022-02-17 08:17:03 +01:00
// This `ReadOptions` is for validation purposes. Ignore
// `FLAGS_rate_limit_user_ops` to avoid slowing any validation.
ReadOptions ro;
ro.snapshot = snapshot;
ro.total_order_seek = true;
std::string ts_str;
Slice ts;
if (FLAGS_user_timestamp_size > 0) {
ts_str = GenerateTimestampForRead();
ts = ts_str;
ro.timestamp = &ts;
}
std::unique_ptr<Iterator> it(db_->NewIterator(ro, column_family));
for (it->Seek(start_key);
it->Valid() && options_.comparator->Compare(it->key(), end_key) <= 0;
it->Next()) {
crc = crc32c::Extend(crc, it->key().data(), it->key().size());
crc = crc32c::Extend(crc, kCrcCalculatorSepearator.data(), 1);
crc = crc32c::Extend(crc, it->value().data(), it->value().size());
crc = crc32c::Extend(crc, kCrcCalculatorSepearator.data(), 1);
}
if (!it->status().ok()) {
fprintf(stderr, "Iterator non-OK when calculating range CRC: %s\n",
it->status().ToString().c_str());
thread->stats.AddErrors(1);
// Fail fast to preserve the DB state.
thread->shared->SetVerificationFailure();
}
return crc;
}
void StressTest::PrintEnv() const {
fprintf(stdout, "RocksDB version : %d.%d\n", kMajorVersion,
kMinorVersion);
fprintf(stdout, "Format version : %d\n", FLAGS_format_version);
fprintf(stdout, "TransactionDB : %s\n",
FLAGS_use_txn ? "true" : "false");
Improve stress test for transactions (#9568) Summary: Test only, no change to functionality. Extremely low risk of library regression. Update test key generation by maintaining existing and non-existing keys. Update db_crashtest.py to drive multiops_txn stress test for both write-committed and write-prepared. Add a make target 'blackbox_crash_test_with_multiops_txn'. Running the following commands caught the bug exposed in https://github.com/facebook/rocksdb/issues/9571. ``` $rm -rf /tmp/rocksdbtest/* $./db_stress -progress_reports=0 -test_multi_ops_txns -use_txn -clear_column_family_one_in=0 \ -column_families=1 -writepercent=0 -delpercent=0 -delrangepercent=0 -customopspercent=60 \ -readpercent=20 -prefixpercent=0 -iterpercent=20 -reopen=0 -ops_per_thread=1000 -ub_a=10000 \ -ub_c=100 -destroy_db_initially=0 -key_spaces_path=/dev/shm/key_spaces_desc -threads=32 -read_fault_one_in=0 $./db_stress -progress_reports=0 -test_multi_ops_txns -use_txn -clear_column_family_one_in=0 -column_families=1 -writepercent=0 -delpercent=0 -delrangepercent=0 -customopspercent=60 -readpercent=20 \ -prefixpercent=0 -iterpercent=20 -reopen=0 -ops_per_thread=1000 -ub_a=10000 -ub_c=100 -destroy_db_initially=0 \ -key_spaces_path=/dev/shm/key_spaces_desc -threads=32 -read_fault_one_in=0 ``` Running the following command caught a bug which will be fixed in https://github.com/facebook/rocksdb/issues/9648 . ``` $TEST_TMPDIR=/dev/shm make blackbox_crash_test_with_multiops_wc_txn ``` Pull Request resolved: https://github.com/facebook/rocksdb/pull/9568 Reviewed By: jay-zhuang Differential Revision: D34308154 Pulled By: riversand963 fbshipit-source-id: 99ff1b65c19b46c471d2f2d3b47adcd342a1b9e7
2022-03-17 03:00:04 +01:00
if (FLAGS_use_txn) {
#ifndef ROCKSDB_LITE
fprintf(stdout, "Two write queues: : %s\n",
FLAGS_two_write_queues ? "true" : "false");
fprintf(stdout, "Write policy : %d\n",
static_cast<int>(FLAGS_txn_write_policy));
if (static_cast<uint64_t>(TxnDBWritePolicy::WRITE_PREPARED) ==
FLAGS_txn_write_policy ||
static_cast<uint64_t>(TxnDBWritePolicy::WRITE_UNPREPARED) ==
FLAGS_txn_write_policy) {
fprintf(stdout, "Snapshot cache bits : %d\n",
static_cast<int>(FLAGS_wp_snapshot_cache_bits));
fprintf(stdout, "Commit cache bits : %d\n",
static_cast<int>(FLAGS_wp_commit_cache_bits));
}
fprintf(stdout, "last cwb for recovery : %s\n",
FLAGS_use_only_the_last_commit_time_batch_for_recovery ? "true"
: "false");
#endif // !ROCKSDB_LITE
}
#ifndef ROCKSDB_LITE
fprintf(stdout, "Stacked BlobDB : %s\n",
FLAGS_use_blob_db ? "true" : "false");
#endif // !ROCKSDB_LITE
fprintf(stdout, "Read only mode : %s\n",
FLAGS_read_only ? "true" : "false");
fprintf(stdout, "Atomic flush : %s\n",
FLAGS_atomic_flush ? "true" : "false");
fprintf(stdout, "Column families : %d\n", FLAGS_column_families);
if (!FLAGS_test_batches_snapshots) {
fprintf(stdout, "Clear CFs one in : %d\n",
FLAGS_clear_column_family_one_in);
}
fprintf(stdout, "Number of threads : %d\n", FLAGS_threads);
fprintf(stdout, "Ops per thread : %lu\n",
(unsigned long)FLAGS_ops_per_thread);
std::string ttl_state("unused");
if (FLAGS_ttl > 0) {
ttl_state = ToString(FLAGS_ttl);
}
fprintf(stdout, "Time to live(sec) : %s\n", ttl_state.c_str());
fprintf(stdout, "Read percentage : %d%%\n", FLAGS_readpercent);
fprintf(stdout, "Prefix percentage : %d%%\n", FLAGS_prefixpercent);
fprintf(stdout, "Write percentage : %d%%\n", FLAGS_writepercent);
fprintf(stdout, "Delete percentage : %d%%\n", FLAGS_delpercent);
fprintf(stdout, "Delete range percentage : %d%%\n", FLAGS_delrangepercent);
fprintf(stdout, "No overwrite percentage : %d%%\n",
FLAGS_nooverwritepercent);
fprintf(stdout, "Iterate percentage : %d%%\n", FLAGS_iterpercent);
fprintf(stdout, "Custom ops percentage : %d%%\n", FLAGS_customopspercent);
fprintf(stdout, "DB-write-buffer-size : %" PRIu64 "\n",
FLAGS_db_write_buffer_size);
fprintf(stdout, "Write-buffer-size : %d\n", FLAGS_write_buffer_size);
fprintf(stdout, "Iterations : %lu\n",
(unsigned long)FLAGS_num_iterations);
fprintf(stdout, "Max key : %lu\n",
(unsigned long)FLAGS_max_key);
fprintf(stdout, "Ratio #ops/#keys : %f\n",
(1.0 * FLAGS_ops_per_thread * FLAGS_threads) / FLAGS_max_key);
fprintf(stdout, "Num times DB reopens : %d\n", FLAGS_reopen);
fprintf(stdout, "Batches/snapshots : %d\n",
FLAGS_test_batches_snapshots);
fprintf(stdout, "Do update in place : %d\n", FLAGS_in_place_update);
fprintf(stdout, "Num keys per lock : %d\n",
1 << FLAGS_log2_keys_per_lock);
std::string compression = CompressionTypeToString(compression_type_e);
fprintf(stdout, "Compression : %s\n", compression.c_str());
std::string bottommost_compression =
CompressionTypeToString(bottommost_compression_type_e);
fprintf(stdout, "Bottommost Compression : %s\n",
bottommost_compression.c_str());
std::string checksum = ChecksumTypeToString(checksum_type_e);
fprintf(stdout, "Checksum type : %s\n", checksum.c_str());
fprintf(stdout, "File checksum impl : %s\n",
FLAGS_file_checksum_impl.c_str());
fprintf(stdout, "Bloom bits / key : %s\n",
FormatDoubleParam(FLAGS_bloom_bits).c_str());
fprintf(stdout, "Max subcompactions : %" PRIu64 "\n",
FLAGS_subcompactions);
fprintf(stdout, "Use MultiGet : %s\n",
FLAGS_use_multiget ? "true" : "false");
const char* memtablerep = "";
switch (FLAGS_rep_factory) {
case kSkipList:
memtablerep = "skip_list";
break;
case kHashSkipList:
memtablerep = "prefix_hash";
break;
case kVectorRep:
memtablerep = "vector";
break;
}
fprintf(stdout, "Memtablerep : %s\n", memtablerep);
#ifndef NDEBUG
KillPoint* kp = KillPoint::GetInstance();
fprintf(stdout, "Test kill odd : %d\n", kp->rocksdb_kill_odds);
if (!kp->rocksdb_kill_exclude_prefixes.empty()) {
fprintf(stdout, "Skipping kill points prefixes:\n");
for (auto& p : kp->rocksdb_kill_exclude_prefixes) {
fprintf(stdout, " %s\n", p.c_str());
}
}
#endif
fprintf(stdout, "Periodic Compaction Secs : %" PRIu64 "\n",
FLAGS_periodic_compaction_seconds);
fprintf(stdout, "Compaction TTL : %" PRIu64 "\n",
FLAGS_compaction_ttl);
fprintf(stdout, "Background Purge : %d\n",
static_cast<int>(FLAGS_avoid_unnecessary_blocking_io));
fprintf(stdout, "Write DB ID to manifest : %d\n",
static_cast<int>(FLAGS_write_dbid_to_manifest));
fprintf(stdout, "Max Write Batch Group Size: %" PRIu64 "\n",
FLAGS_max_write_batch_group_size_bytes);
fprintf(stdout, "Use dynamic level : %d\n",
static_cast<int>(FLAGS_level_compaction_dynamic_level_bytes));
fprintf(stdout, "Read fault one in : %d\n", FLAGS_read_fault_one_in);
fprintf(stdout, "Write fault one in : %d\n", FLAGS_write_fault_one_in);
fprintf(stdout, "Open metadata write fault one in:\n");
fprintf(stdout, " %d\n",
FLAGS_open_metadata_write_fault_one_in);
fprintf(stdout, "Sync fault injection : %d\n", FLAGS_sync_fault_injection);
fprintf(stdout, "Best efforts recovery : %d\n",
static_cast<int>(FLAGS_best_efforts_recovery));
fprintf(stdout, "Fail if OPTIONS file error: %d\n",
static_cast<int>(FLAGS_fail_if_options_file_error));
fprintf(stdout, "User timestamp size bytes : %d\n",
static_cast<int>(FLAGS_user_timestamp_size));
fprintf(stdout, "------------------------------------------------\n");
}
void StressTest::Open() {
assert(db_ == nullptr);
#ifndef ROCKSDB_LITE
assert(txn_db_ == nullptr);
#endif
if (FLAGS_options_file.empty()) {
BlockBasedTableOptions block_based_options;
block_based_options.block_cache = cache_;
block_based_options.cache_index_and_filter_blocks =
FLAGS_cache_index_and_filter_blocks;
block_based_options.metadata_cache_options.top_level_index_pinning =
static_cast<PinningTier>(FLAGS_top_level_index_pinning);
block_based_options.metadata_cache_options.partition_pinning =
static_cast<PinningTier>(FLAGS_partition_pinning);
block_based_options.metadata_cache_options.unpartitioned_pinning =
static_cast<PinningTier>(FLAGS_unpartitioned_pinning);
block_based_options.block_cache_compressed = compressed_cache_;
block_based_options.checksum = checksum_type_e;
block_based_options.block_size = FLAGS_block_size;
Account memory of big memory users in BlockBasedTable in global memory limit (#9748) Summary: **Context:** Through heap profiling, we discovered that `BlockBasedTableReader` objects can accumulate and lead to high memory usage (e.g, `max_open_file = -1`). These memories are currently not saved, not tracked, not constrained and not cache evict-able. As a first step to improve this, similar to https://github.com/facebook/rocksdb/pull/8428, this PR is to track an estimate of `BlockBasedTableReader` object's memory in block cache and fail future creation if the memory usage exceeds the available space of cache at the time of creation. **Summary:** - Approximate big memory users (`BlockBasedTable::Rep` and `TableProperties` )' memory usage in addition to the existing estimated ones (filter block/index block/un-compression dictionary) - Charge all of these memory usages to block cache on `BlockBasedTable::Open()` and release them on `~BlockBasedTable()` as there is no memory usage fluctuation of concern in between - Refactor on CacheReservationManager (and its call-sites) to add concurrent support for BlockBasedTable used in this PR. Pull Request resolved: https://github.com/facebook/rocksdb/pull/9748 Test Plan: - New unit tests - db bench: `OpenDb` : **-0.52% in ms** - Setup `./db_bench -benchmarks=fillseq -db=/dev/shm/testdb -disable_auto_compactions=1 -write_buffer_size=1048576` - Repeated run with pre-change w/o feature and post-change with feature, benchmark `OpenDb`: `./db_bench -benchmarks=readrandom -use_existing_db=1 -db=/dev/shm/testdb -reserve_table_reader_memory=true (remove this when running w/o feature) -file_opening_threads=3 -open_files=-1 -report_open_timing=true| egrep 'OpenDb:'` #-run | (feature-off) avg milliseconds | std milliseconds | (feature-on) avg milliseconds | std milliseconds | change (%) -- | -- | -- | -- | -- | -- 10 | 11.4018 | 5.95173 | 9.47788 | 1.57538 | -16.87382694 20 | 9.23746 | 0.841053 | 9.32377 | 1.14074 | 0.9343477536 40 | 9.0876 | 0.671129 | 9.35053 | 1.11713 | 2.893283155 80 | 9.72514 | 2.28459 | 9.52013 | 1.0894 | -2.108041632 160 | 9.74677 | 0.991234 | 9.84743 | 1.73396 | 1.032752389 320 | 10.7297 | 5.11555 | 10.547 | 1.97692 | **-1.70275031** 640 | 11.7092 | 2.36565 | 11.7869 | 2.69377 | **0.6635807741** - db bench on write with cost to cache in WriteBufferManager (just in case this PR's CRM refactoring accidentally slows down anything in WBM) : `fillseq` : **+0.54% in micros/op** `./db_bench -benchmarks=fillseq -db=/dev/shm/testdb -disable_auto_compactions=1 -cost_write_buffer_to_cache=true -write_buffer_size=10000000000 | egrep 'fillseq'` #-run | (pre-PR) avg micros/op | std micros/op | (post-PR) avg micros/op | std micros/op | change (%) -- | -- | -- | -- | -- | -- 10 | 6.15 | 0.260187 | 6.289 | 0.371192 | 2.260162602 20 | 7.28025 | 0.465402 | 7.37255 | 0.451256 | 1.267813605 40 | 7.06312 | 0.490654 | 7.13803 | 0.478676 | **1.060579461** 80 | 7.14035 | 0.972831 | 7.14196 | 0.92971 | **0.02254791432** - filter bench: `bloom filter`: **-0.78% in ms/key** - ` ./filter_bench -impl=2 -quick -reserve_table_builder_memory=true | grep 'Build avg'` #-run | (pre-PR) avg ns/key | std ns/key | (post-PR) ns/key | std ns/key | change (%) -- | -- | -- | -- | -- | -- 10 | 26.4369 | 0.442182 | 26.3273 | 0.422919 | **-0.4145720565** 20 | 26.4451 | 0.592787 | 26.1419 | 0.62451 | **-1.1465262** - Crash test `python3 tools/db_crashtest.py blackbox --reserve_table_reader_memory=1 --cache_size=1` killed as normal Reviewed By: ajkr Differential Revision: D35136549 Pulled By: hx235 fbshipit-source-id: 146978858d0f900f43f4eb09bfd3e83195e3be28
2022-04-06 19:33:00 +02:00
block_based_options.reserve_table_reader_memory =
FLAGS_reserve_table_reader_memory;
block_based_options.format_version =
static_cast<uint32_t>(FLAGS_format_version);
block_based_options.index_block_restart_interval =
static_cast<int32_t>(FLAGS_index_block_restart_interval);
block_based_options.filter_policy = filter_policy_;
block_based_options.partition_filters = FLAGS_partition_filters;
Minimize memory internal fragmentation for Bloom filters (#6427) Summary: New experimental option BBTO::optimize_filters_for_memory builds filters that maximize their use of "usable size" from malloc_usable_size, which is also used to compute block cache charges. Rather than always "rounding up," we track state in the BloomFilterPolicy object to mix essentially "rounding down" and "rounding up" so that the average FP rate of all generated filters is the same as without the option. (YMMV as heavily accessed filters might be unluckily lower accuracy.) Thus, the option near-minimizes what the block cache considers as "memory used" for a given target Bloom filter false positive rate and Bloom filter implementation. There are no forward or backward compatibility issues with this change, though it only works on the format_version=5 Bloom filter. With Jemalloc, we see about 10% reduction in memory footprint (and block cache charge) for Bloom filters, but 1-2% increase in storage footprint, due to encoding efficiency losses (FP rate is non-linear with bits/key). Why not weighted random round up/down rather than state tracking? By only requiring malloc_usable_size, we don't actually know what the next larger and next smaller usable sizes for the allocator are. We pick a requested size, accept and use whatever usable size it has, and use the difference to inform our next choice. This allows us to narrow in on the right balance without tracking/predicting usable sizes. Why not weight history of generated filter false positive rates by number of keys? This could lead to excess skew in small filters after generating a large filter. Results from filter_bench with jemalloc (irrelevant details omitted): (normal keys/filter, but high variance) $ ./filter_bench -quick -impl=2 -average_keys_per_filter=30000 -vary_key_count_ratio=0.9 Build avg ns/key: 29.6278 Number of filters: 5516 Total size (MB): 200.046 Reported total allocated memory (MB): 220.597 Reported internal fragmentation: 10.2732% Bits/key stored: 10.0097 Average FP rate %: 0.965228 $ ./filter_bench -quick -impl=2 -average_keys_per_filter=30000 -vary_key_count_ratio=0.9 -optimize_filters_for_memory Build avg ns/key: 30.5104 Number of filters: 5464 Total size (MB): 200.015 Reported total allocated memory (MB): 200.322 Reported internal fragmentation: 0.153709% Bits/key stored: 10.1011 Average FP rate %: 0.966313 (very few keys / filter, optimization not as effective due to ~59 byte internal fragmentation in blocked Bloom filter representation) $ ./filter_bench -quick -impl=2 -average_keys_per_filter=1000 -vary_key_count_ratio=0.9 Build avg ns/key: 29.5649 Number of filters: 162950 Total size (MB): 200.001 Reported total allocated memory (MB): 224.624 Reported internal fragmentation: 12.3117% Bits/key stored: 10.2951 Average FP rate %: 0.821534 $ ./filter_bench -quick -impl=2 -average_keys_per_filter=1000 -vary_key_count_ratio=0.9 -optimize_filters_for_memory Build avg ns/key: 31.8057 Number of filters: 159849 Total size (MB): 200 Reported total allocated memory (MB): 208.846 Reported internal fragmentation: 4.42297% Bits/key stored: 10.4948 Average FP rate %: 0.811006 (high keys/filter) $ ./filter_bench -quick -impl=2 -average_keys_per_filter=1000000 -vary_key_count_ratio=0.9 Build avg ns/key: 29.7017 Number of filters: 164 Total size (MB): 200.352 Reported total allocated memory (MB): 221.5 Reported internal fragmentation: 10.5552% Bits/key stored: 10.0003 Average FP rate %: 0.969358 $ ./filter_bench -quick -impl=2 -average_keys_per_filter=1000000 -vary_key_count_ratio=0.9 -optimize_filters_for_memory Build avg ns/key: 30.7131 Number of filters: 160 Total size (MB): 200.928 Reported total allocated memory (MB): 200.938 Reported internal fragmentation: 0.00448054% Bits/key stored: 10.1852 Average FP rate %: 0.963387 And from db_bench (block cache) with jemalloc: $ ./db_bench -db=/dev/shm/dbbench.no_optimize -benchmarks=fillrandom -format_version=5 -value_size=90 -bloom_bits=10 -num=2000000 -threads=8 -compaction_style=2 -fifo_compaction_max_table_files_size_mb=10000 -fifo_compaction_allow_compaction=false $ ./db_bench -db=/dev/shm/dbbench -benchmarks=fillrandom -format_version=5 -value_size=90 -bloom_bits=10 -num=2000000 -threads=8 -optimize_filters_for_memory -compaction_style=2 -fifo_compaction_max_table_files_size_mb=10000 -fifo_compaction_allow_compaction=false $ (for FILE in /dev/shm/dbbench.no_optimize/*.sst; do ./sst_dump --file=$FILE --show_properties | grep 'filter block' ; done) | awk '{ t += $4; } END { print t; }' 17063835 $ (for FILE in /dev/shm/dbbench/*.sst; do ./sst_dump --file=$FILE --show_properties | grep 'filter block' ; done) | awk '{ t += $4; } END { print t; }' 17430747 $ #^ 2.1% additional filter storage $ ./db_bench -db=/dev/shm/dbbench.no_optimize -use_existing_db -benchmarks=readrandom,stats -statistics -bloom_bits=10 -num=2000000 -compaction_style=2 -fifo_compaction_max_table_files_size_mb=10000 -fifo_compaction_allow_compaction=false -duration=10 -cache_index_and_filter_blocks -cache_size=1000000000 rocksdb.block.cache.index.add COUNT : 33 rocksdb.block.cache.index.bytes.insert COUNT : 8440400 rocksdb.block.cache.filter.add COUNT : 33 rocksdb.block.cache.filter.bytes.insert COUNT : 21087528 rocksdb.bloom.filter.useful COUNT : 4963889 rocksdb.bloom.filter.full.positive COUNT : 1214081 rocksdb.bloom.filter.full.true.positive COUNT : 1161999 $ #^ 1.04 % observed FP rate $ ./db_bench -db=/dev/shm/dbbench -use_existing_db -benchmarks=readrandom,stats -statistics -bloom_bits=10 -num=2000000 -compaction_style=2 -fifo_compaction_max_table_files_size_mb=10000 -fifo_compaction_allow_compaction=false -optimize_filters_for_memory -duration=10 -cache_index_and_filter_blocks -cache_size=1000000000 rocksdb.block.cache.index.add COUNT : 33 rocksdb.block.cache.index.bytes.insert COUNT : 8448592 rocksdb.block.cache.filter.add COUNT : 33 rocksdb.block.cache.filter.bytes.insert COUNT : 18220328 rocksdb.bloom.filter.useful COUNT : 5360933 rocksdb.bloom.filter.full.positive COUNT : 1321315 rocksdb.bloom.filter.full.true.positive COUNT : 1262999 $ #^ 1.08 % observed FP rate, 13.6% less memory usage for filters (Due to specific key density, this example tends to generate filters that are "worse than average" for internal fragmentation. "Better than average" cases can show little or no improvement.) Pull Request resolved: https://github.com/facebook/rocksdb/pull/6427 Test Plan: unit test added, 'make check' with gcc, clang and valgrind Reviewed By: siying Differential Revision: D22124374 Pulled By: pdillinger fbshipit-source-id: f3e3aa152f9043ddf4fae25799e76341d0d8714e
2020-06-22 22:30:57 +02:00
block_based_options.optimize_filters_for_memory =
FLAGS_optimize_filters_for_memory;
Detect (new) Bloom/Ribbon Filter construction corruption (#9342) Summary: Note: rebase on and merge after https://github.com/facebook/rocksdb/pull/9349, https://github.com/facebook/rocksdb/pull/9345, (optional) https://github.com/facebook/rocksdb/pull/9393 **Context:** (Quoted from pdillinger) Layers of information during new Bloom/Ribbon Filter construction in building block-based tables includes the following: a) set of keys to add to filter b) set of hashes to add to filter (64-bit hash applied to each key) c) set of Bloom indices to set in filter, with duplicates d) set of Bloom indices to set in filter, deduplicated e) final filter and its checksum This PR aims to detect corruption (e.g, unexpected hardware/software corruption on data structures residing in the memory for a long time) from b) to e) and leave a) as future works for application level. - b)'s corruption is detected by verifying the xor checksum of the hash entries calculated as the entries accumulate before being added to the filter. (i.e, `XXPH3FilterBitsBuilder::MaybeVerifyHashEntriesChecksum()`) - c) - e)'s corruption is detected by verifying the hash entries indeed exists in the constructed filter by re-querying these hash entries in the filter (i.e, `FilterBitsBuilder::MaybePostVerify()`) after computing the block checksum (except for PartitionFilter, which is done right after each `FilterBitsBuilder::Finish` for impl simplicity - see code comment for more). For this stage of detection, we assume hash entries are not corrupted after checking on b) since the time interval from b) to c) is relatively short IMO. Option to enable this feature of detection is `BlockBasedTableOptions::detect_filter_construct_corruption` which is false by default. **Summary:** - Implemented new functions `XXPH3FilterBitsBuilder::MaybeVerifyHashEntriesChecksum()` and `FilterBitsBuilder::MaybePostVerify()` - Ensured hash entries, final filter and banding and their [cache reservation ](https://github.com/facebook/rocksdb/issues/9073) are released properly despite corruption - See [Filter.construction.artifacts.release.point.pdf ](https://github.com/facebook/rocksdb/files/7923487/Design.Filter.construction.artifacts.release.point.pdf) for high-level design - Bundled and refactored hash entries's related artifact in XXPH3FilterBitsBuilder into `HashEntriesInfo` for better control on lifetime of these artifact during `SwapEntires`, `ResetEntries` - Ensured RocksDB block-based table builder calls `FilterBitsBuilder::MaybePostVerify()` after constructing the filter by `FilterBitsBuilder::Finish()` - When encountering such filter construction corruption, stop writing the filter content to files and mark such a block-based table building non-ok by storing the corruption status in the builder. Pull Request resolved: https://github.com/facebook/rocksdb/pull/9342 Test Plan: - Added new unit test `DBFilterConstructionCorruptionTestWithParam.DetectCorruption` - Included this new feature in `DBFilterConstructionReserveMemoryTestWithParam.ReserveMemory` as this feature heavily touch ReserveMemory's impl - For fallback case, I run `./filter_bench -impl=3 -detect_filter_construct_corruption=true -reserve_table_builder_memory=true -strict_capacity_limit=true -quick -runs 10 | grep 'Build avg'` to make sure nothing break. - Added to `filter_bench`: increased filter construction time by **30%**, mostly by `MaybePostVerify()` - FastLocalBloom - Before change: `./filter_bench -impl=2 -quick -runs 10 | grep 'Build avg'`: **28.86643s** - After change: - `./filter_bench -impl=2 -detect_filter_construct_corruption=false -quick -runs 10 | grep 'Build avg'` (expect a tiny increase due to MaybePostVerify is always called regardless): **27.6644s (-4% perf improvement might be due to now we don't drop bloom hash entry in `AddAllEntries` along iteration but in bulk later, same with the bypassing-MaybePostVerify case below)** - `./filter_bench -impl=2 -detect_filter_construct_corruption=true -quick -runs 10 | grep 'Build avg'` (expect acceptable increase): **34.41159s (+20%)** - `./filter_bench -impl=2 -detect_filter_construct_corruption=true -quick -runs 10 | grep 'Build avg'` (by-passing MaybePostVerify, expect minor increase): **27.13431s (-6%)** - Standard128Ribbon - Before change: `./filter_bench -impl=3 -quick -runs 10 | grep 'Build avg'`: **122.5384s** - After change: - `./filter_bench -impl=3 -detect_filter_construct_corruption=false -quick -runs 10 | grep 'Build avg'` (expect a tiny increase due to MaybePostVerify is always called regardless - verified by removing MaybePostVerify under this case and found only +-1ns difference): **124.3588s (+2%)** - `./filter_bench -impl=3 -detect_filter_construct_corruption=true -quick -runs 10 | grep 'Build avg'`(expect acceptable increase): **159.4946s (+30%)** - `./filter_bench -impl=3 -detect_filter_construct_corruption=true -quick -runs 10 | grep 'Build avg'`(by-passing MaybePostVerify, expect minor increase) : **125.258s (+2%)** - Added to `db_stress`: `make crash_test`, `./db_stress --detect_filter_construct_corruption=true` - Manually smoke-tested: manually corrupted the filter construction in some db level tests with basic PUT and background flush. As expected, the error did get returned to users in subsequent PUT and Flush status. Reviewed By: pdillinger Differential Revision: D33746928 Pulled By: hx235 fbshipit-source-id: cb056426be5a7debc1cd16f23bc250f36a08ca57
2022-02-02 02:41:20 +01:00
block_based_options.detect_filter_construct_corruption =
FLAGS_detect_filter_construct_corruption;
block_based_options.index_type =
static_cast<BlockBasedTableOptions::IndexType>(FLAGS_index_type);
block_based_options.prepopulate_block_cache =
static_cast<BlockBasedTableOptions::PrepopulateBlockCache>(
FLAGS_prepopulate_block_cache);
options_.table_factory.reset(
NewBlockBasedTableFactory(block_based_options));
options_.db_write_buffer_size = FLAGS_db_write_buffer_size;
options_.write_buffer_size = FLAGS_write_buffer_size;
options_.max_write_buffer_number = FLAGS_max_write_buffer_number;
options_.min_write_buffer_number_to_merge =
FLAGS_min_write_buffer_number_to_merge;
options_.max_write_buffer_number_to_maintain =
FLAGS_max_write_buffer_number_to_maintain;
options_.max_write_buffer_size_to_maintain =
FLAGS_max_write_buffer_size_to_maintain;
options_.memtable_prefix_bloom_size_ratio =
FLAGS_memtable_prefix_bloom_size_ratio;
options_.memtable_whole_key_filtering = FLAGS_memtable_whole_key_filtering;
Improve stress test for transactions (#9568) Summary: Test only, no change to functionality. Extremely low risk of library regression. Update test key generation by maintaining existing and non-existing keys. Update db_crashtest.py to drive multiops_txn stress test for both write-committed and write-prepared. Add a make target 'blackbox_crash_test_with_multiops_txn'. Running the following commands caught the bug exposed in https://github.com/facebook/rocksdb/issues/9571. ``` $rm -rf /tmp/rocksdbtest/* $./db_stress -progress_reports=0 -test_multi_ops_txns -use_txn -clear_column_family_one_in=0 \ -column_families=1 -writepercent=0 -delpercent=0 -delrangepercent=0 -customopspercent=60 \ -readpercent=20 -prefixpercent=0 -iterpercent=20 -reopen=0 -ops_per_thread=1000 -ub_a=10000 \ -ub_c=100 -destroy_db_initially=0 -key_spaces_path=/dev/shm/key_spaces_desc -threads=32 -read_fault_one_in=0 $./db_stress -progress_reports=0 -test_multi_ops_txns -use_txn -clear_column_family_one_in=0 -column_families=1 -writepercent=0 -delpercent=0 -delrangepercent=0 -customopspercent=60 -readpercent=20 \ -prefixpercent=0 -iterpercent=20 -reopen=0 -ops_per_thread=1000 -ub_a=10000 -ub_c=100 -destroy_db_initially=0 \ -key_spaces_path=/dev/shm/key_spaces_desc -threads=32 -read_fault_one_in=0 ``` Running the following command caught a bug which will be fixed in https://github.com/facebook/rocksdb/issues/9648 . ``` $TEST_TMPDIR=/dev/shm make blackbox_crash_test_with_multiops_wc_txn ``` Pull Request resolved: https://github.com/facebook/rocksdb/pull/9568 Reviewed By: jay-zhuang Differential Revision: D34308154 Pulled By: riversand963 fbshipit-source-id: 99ff1b65c19b46c471d2f2d3b47adcd342a1b9e7
2022-03-17 03:00:04 +01:00
options_.disable_auto_compactions = FLAGS_disable_auto_compactions;
options_.max_background_compactions = FLAGS_max_background_compactions;
options_.max_background_flushes = FLAGS_max_background_flushes;
options_.compaction_style =
static_cast<ROCKSDB_NAMESPACE::CompactionStyle>(FLAGS_compaction_style);
if (FLAGS_prefix_size >= 0) {
options_.prefix_extractor.reset(
NewFixedPrefixTransform(FLAGS_prefix_size));
}
options_.max_open_files = FLAGS_open_files;
options_.statistics = dbstats;
options_.env = db_stress_env;
options_.use_fsync = FLAGS_use_fsync;
options_.compaction_readahead_size = FLAGS_compaction_readahead_size;
options_.allow_mmap_reads = FLAGS_mmap_read;
options_.allow_mmap_writes = FLAGS_mmap_write;
options_.use_direct_reads = FLAGS_use_direct_reads;
options_.use_direct_io_for_flush_and_compaction =
FLAGS_use_direct_io_for_flush_and_compaction;
options_.recycle_log_file_num =
static_cast<size_t>(FLAGS_recycle_log_file_num);
options_.target_file_size_base = FLAGS_target_file_size_base;
options_.target_file_size_multiplier = FLAGS_target_file_size_multiplier;
options_.max_bytes_for_level_base = FLAGS_max_bytes_for_level_base;
options_.max_bytes_for_level_multiplier =
FLAGS_max_bytes_for_level_multiplier;
options_.level0_stop_writes_trigger = FLAGS_level0_stop_writes_trigger;
options_.level0_slowdown_writes_trigger =
FLAGS_level0_slowdown_writes_trigger;
options_.level0_file_num_compaction_trigger =
FLAGS_level0_file_num_compaction_trigger;
options_.compression = compression_type_e;
options_.bottommost_compression = bottommost_compression_type_e;
options_.compression_opts.max_dict_bytes = FLAGS_compression_max_dict_bytes;
options_.compression_opts.zstd_max_train_bytes =
FLAGS_compression_zstd_max_train_bytes;
options_.compression_opts.parallel_threads =
FLAGS_compression_parallel_threads;
Limit buffering for collecting samples for compression dictionary (#7970) Summary: For dictionary compression, we need to collect some representative samples of the data to be compressed, which we use to either generate or train (when `CompressionOptions::zstd_max_train_bytes > 0`) a dictionary. Previously, the strategy was to buffer all the data blocks during flush, and up to the target file size during compaction. That strategy allowed us to randomly pick samples from as wide a range as possible that'd be guaranteed to land in a single output file. However, some users try to make huge files in memory-constrained environments, where this strategy can cause OOM. This PR introduces an option, `CompressionOptions::max_dict_buffer_bytes`, that limits how much data blocks are buffered before we switch to unbuffered mode (which means creating the per-SST dictionary, writing out the buffered data, and compressing/writing new blocks as soon as they are built). It is not strict as we currently buffer more than just data blocks -- also keys are buffered. But it does make a step towards giving users predictable memory usage. Related changes include: - Changed sampling for dictionary compression to select unique data blocks when there is limited availability of data blocks - Made use of `BlockBuilder::SwapAndReset()` to save an allocation+memcpy when buffering data blocks for building a dictionary - Changed `ParseBoolean()` to accept an input containing characters after the boolean. This is necessary since, with this PR, a value for `CompressionOptions::enabled` is no longer necessarily the final component in the `CompressionOptions` string. Pull Request resolved: https://github.com/facebook/rocksdb/pull/7970 Test Plan: - updated `CompressionOptions` unit tests to verify limit is respected (to the extent expected in the current implementation) in various scenarios of flush/compaction to bottommost/non-bottommost level - looked at jemalloc heap profiles right before and after switching to unbuffered mode during flush/compaction. Verified memory usage in buffering is proportional to the limit set. Reviewed By: pdillinger Differential Revision: D26467994 Pulled By: ajkr fbshipit-source-id: 3da4ef9fba59974e4ef40e40c01611002c861465
2021-02-19 23:06:59 +01:00
options_.compression_opts.max_dict_buffer_bytes =
FLAGS_compression_max_dict_buffer_bytes;
options_.create_if_missing = true;
options_.max_manifest_file_size = FLAGS_max_manifest_file_size;
options_.inplace_update_support = FLAGS_in_place_update;
options_.max_subcompactions = static_cast<uint32_t>(FLAGS_subcompactions);
options_.allow_concurrent_memtable_write =
FLAGS_allow_concurrent_memtable_write;
Memtable sampling for mempurge heuristic. (#8628) Summary: Changes the API of the MemPurge process: the `bool experimental_allow_mempurge` and `experimental_mempurge_policy` flags have been replaced by a `double experimental_mempurge_threshold` option. This change of API reflects another major change introduced in this PR: the MemPurgeDecider() function now works by sampling the memtables being flushed to estimate the overall amount of useful payload (payload minus the garbage), and then compare this useful payload estimate with the `double experimental_mempurge_threshold` value. Therefore, when the value of this flag is `0.0` (default value), mempurge is simply deactivated. On the other hand, a value of `DBL_MAX` would be equivalent to always going through a mempurge regardless of the garbage ratio estimate. At the moment, a `double experimental_mempurge_threshold` value else than 0.0 or `DBL_MAX` is opnly supported`with the `SkipList` memtable representation. Regarding the sampling, this PR includes the introduction of a `MemTable::UniqueRandomSample` function that collects (approximately) random entries from the memtable by using the new `SkipList::Iterator::RandomSeek()` under the hood, or by iterating through each memtable entry, depending on the target sample size and the total number of entries. The unit tests have been readapted to support this new API. Pull Request resolved: https://github.com/facebook/rocksdb/pull/8628 Reviewed By: pdillinger Differential Revision: D30149315 Pulled By: bjlemaire fbshipit-source-id: 1feef5390c95db6f4480ab4434716533d3947f27
2021-08-11 03:07:48 +02:00
options_.experimental_mempurge_threshold =
FLAGS_experimental_mempurge_threshold;
options_.periodic_compaction_seconds = FLAGS_periodic_compaction_seconds;
options_.ttl = FLAGS_compaction_ttl;
options_.enable_pipelined_write = FLAGS_enable_pipelined_write;
options_.enable_write_thread_adaptive_yield =
FLAGS_enable_write_thread_adaptive_yield;
options_.compaction_options_universal.size_ratio =
FLAGS_universal_size_ratio;
options_.compaction_options_universal.min_merge_width =
FLAGS_universal_min_merge_width;
options_.compaction_options_universal.max_merge_width =
FLAGS_universal_max_merge_width;
options_.compaction_options_universal.max_size_amplification_percent =
FLAGS_universal_max_size_amplification_percent;
options_.atomic_flush = FLAGS_atomic_flush;
options_.avoid_unnecessary_blocking_io =
FLAGS_avoid_unnecessary_blocking_io;
options_.write_dbid_to_manifest = FLAGS_write_dbid_to_manifest;
options_.avoid_flush_during_recovery = FLAGS_avoid_flush_during_recovery;
options_.max_write_batch_group_size_bytes =
FLAGS_max_write_batch_group_size_bytes;
options_.level_compaction_dynamic_level_bytes =
FLAGS_level_compaction_dynamic_level_bytes;
options_.file_checksum_gen_factory =
GetFileChecksumImpl(FLAGS_file_checksum_impl);
options_.track_and_verify_wals_in_manifest = true;
// Integrated BlobDB
options_.enable_blob_files = FLAGS_enable_blob_files;
options_.min_blob_size = FLAGS_min_blob_size;
options_.blob_file_size = FLAGS_blob_file_size;
options_.blob_compression_type =
StringToCompressionType(FLAGS_blob_compression_type.c_str());
options_.enable_blob_garbage_collection =
FLAGS_enable_blob_garbage_collection;
options_.blob_garbage_collection_age_cutoff =
FLAGS_blob_garbage_collection_age_cutoff;
Make it possible to force the garbage collection of the oldest blob files (#8994) Summary: The current BlobDB garbage collection logic works by relocating the valid blobs from the oldest blob files as they are encountered during compaction, and cleaning up blob files once they contain nothing but garbage. However, with sufficiently skewed workloads, it is theoretically possible to end up in a situation when few or no compactions get scheduled for the SST files that contain references to the oldest blob files, which can lead to increased space amp due to the lack of GC. In order to efficiently handle such workloads, the patch adds a new BlobDB configuration option called `blob_garbage_collection_force_threshold`, which signals to BlobDB to schedule targeted compactions for the SST files that keep alive the oldest batch of blob files if the overall ratio of garbage in the given blob files meets the threshold *and* all the given blob files are eligible for GC based on `blob_garbage_collection_age_cutoff`. (For example, if the new option is set to 0.9, targeted compactions will get scheduled if the sum of garbage bytes meets or exceeds 90% of the sum of total bytes in the oldest blob files, assuming all affected blob files are below the age-based cutoff.) The net result of these targeted compactions is that the valid blobs in the oldest blob files are relocated and the oldest blob files themselves cleaned up (since *all* SST files that rely on them get compacted away). These targeted compactions are similar to periodic compactions in the sense that they force certain SST files that otherwise would not get picked up to undergo compaction and also in the sense that instead of merging files from multiple levels, they target a single file. (Note: such compactions might still include neighboring files from the same level due to the need of having a "clean cut" boundary but they never include any files from any other level.) This functionality is currently only supported with the leveled compaction style and is inactive by default (since the default value is set to 1.0, i.e. 100%). Pull Request resolved: https://github.com/facebook/rocksdb/pull/8994 Test Plan: Ran `make check` and tested using `db_bench` and the stress/crash tests. Reviewed By: riversand963 Differential Revision: D31489850 Pulled By: ltamasi fbshipit-source-id: 44057d511726a0e2a03c5d9313d7511b3f0c4eab
2021-10-12 03:00:44 +02:00
options_.blob_garbage_collection_force_threshold =
FLAGS_blob_garbage_collection_force_threshold;
options_.blob_compaction_readahead_size =
FLAGS_blob_compaction_readahead_size;
} else {
#ifdef ROCKSDB_LITE
fprintf(stderr, "--options_file not supported in lite mode\n");
exit(1);
#else
DBOptions db_options;
std::vector<ColumnFamilyDescriptor> cf_descriptors;
Status s = LoadOptionsFromFile(FLAGS_options_file, db_stress_env,
&db_options, &cf_descriptors);
db_options.env = new DbStressEnvWrapper(db_stress_env);
if (!s.ok()) {
fprintf(stderr, "Unable to load options file %s --- %s\n",
FLAGS_options_file.c_str(), s.ToString().c_str());
exit(1);
}
options_ = Options(db_options, cf_descriptors[0].options);
#endif // ROCKSDB_LITE
}
if (FLAGS_rate_limiter_bytes_per_sec > 0) {
options_.rate_limiter.reset(NewGenericRateLimiter(
FLAGS_rate_limiter_bytes_per_sec, 1000 /* refill_period_us */,
10 /* fairness */,
FLAGS_rate_limit_bg_reads ? RateLimiter::Mode::kReadsOnly
: RateLimiter::Mode::kWritesOnly));
}
if (FLAGS_sst_file_manager_bytes_per_sec > 0 ||
FLAGS_sst_file_manager_bytes_per_truncate > 0) {
Status status;
options_.sst_file_manager.reset(NewSstFileManager(
db_stress_env, options_.info_log, "" /* trash_dir */,
static_cast<int64_t>(FLAGS_sst_file_manager_bytes_per_sec),
true /* delete_existing_trash */, &status,
0.25 /* max_trash_db_ratio */,
FLAGS_sst_file_manager_bytes_per_truncate));
if (!status.ok()) {
fprintf(stderr, "SstFileManager creation failed: %s\n",
status.ToString().c_str());
exit(1);
}
}
if (FLAGS_prefix_size == 0 && FLAGS_rep_factory == kHashSkipList) {
fprintf(stderr,
"prefeix_size cannot be zero if memtablerep == prefix_hash\n");
exit(1);
}
if (FLAGS_prefix_size != 0 && FLAGS_rep_factory != kHashSkipList) {
fprintf(stderr,
"WARNING: prefix_size is non-zero but "
"memtablerep != prefix_hash\n");
}
switch (FLAGS_rep_factory) {
case kSkipList:
// no need to do anything
break;
#ifndef ROCKSDB_LITE
case kHashSkipList:
options_.memtable_factory.reset(NewHashSkipListRepFactory(10000));
break;
case kVectorRep:
options_.memtable_factory.reset(new VectorRepFactory());
break;
#else
default:
fprintf(stderr,
"RocksdbLite only supports skip list mem table. Skip "
"--rep_factory\n");
#endif // ROCKSDB_LITE
}
if (FLAGS_use_full_merge_v1) {
options_.merge_operator = MergeOperators::CreateDeprecatedPutOperator();
} else {
options_.merge_operator = MergeOperators::CreatePutOperator();
}
if (FLAGS_enable_compaction_filter) {
options_.compaction_filter_factory =
std::make_shared<DbStressCompactionFilterFactory>();
}
options_.table_properties_collector_factories.emplace_back(
std::make_shared<DbStressTablePropertiesCollectorFactory>());
options_.best_efforts_recovery = FLAGS_best_efforts_recovery;
options_.paranoid_file_checks = FLAGS_paranoid_file_checks;
options_.fail_if_options_file_error = FLAGS_fail_if_options_file_error;
if ((options_.enable_blob_files || options_.enable_blob_garbage_collection ||
FLAGS_allow_setting_blob_options_dynamically) &&
FLAGS_best_efforts_recovery) {
fprintf(stderr,
"Integrated BlobDB is currently incompatible with best-effort "
"recovery\n");
exit(1);
}
fprintf(stdout,
"Integrated BlobDB: blob files enabled %d, min blob size %" PRIu64
", blob file size %" PRIu64
", blob compression type %s, blob GC enabled %d, cutoff %f, force "
"threshold %f, blob compaction readahead size %" PRIu64 "\n",
options_.enable_blob_files, options_.min_blob_size,
options_.blob_file_size,
CompressionTypeToString(options_.blob_compression_type).c_str(),
options_.enable_blob_garbage_collection,
options_.blob_garbage_collection_age_cutoff,
options_.blob_garbage_collection_force_threshold,
options_.blob_compaction_readahead_size);
fprintf(stdout, "DB path: [%s]\n", FLAGS_db.c_str());
Status s;
if (FLAGS_user_timestamp_size > 0) {
CheckAndSetOptionsForUserTimestamp();
}
if (FLAGS_ttl == -1) {
std::vector<std::string> existing_column_families;
s = DB::ListColumnFamilies(DBOptions(options_), FLAGS_db,
&existing_column_families); // ignore errors
if (!s.ok()) {
// DB doesn't exist
assert(existing_column_families.empty());
assert(column_family_names_.empty());
column_family_names_.push_back(kDefaultColumnFamilyName);
} else if (column_family_names_.empty()) {
// this is the first call to the function Open()
column_family_names_ = existing_column_families;
} else {
// this is a reopen. just assert that existing column_family_names are
// equivalent to what we remember
auto sorted_cfn = column_family_names_;
std::sort(sorted_cfn.begin(), sorted_cfn.end());
std::sort(existing_column_families.begin(),
existing_column_families.end());
if (sorted_cfn != existing_column_families) {
fprintf(stderr, "Expected column families differ from the existing:\n");
fprintf(stderr, "Expected: {");
for (auto cf : sorted_cfn) {
fprintf(stderr, "%s ", cf.c_str());
}
fprintf(stderr, "}\n");
fprintf(stderr, "Existing: {");
for (auto cf : existing_column_families) {
fprintf(stderr, "%s ", cf.c_str());
}
fprintf(stderr, "}\n");
}
assert(sorted_cfn == existing_column_families);
}
std::vector<ColumnFamilyDescriptor> cf_descriptors;
for (auto name : column_family_names_) {
if (name != kDefaultColumnFamilyName) {
new_column_family_name_ =
std::max(new_column_family_name_.load(), std::stoi(name) + 1);
}
cf_descriptors.emplace_back(name, ColumnFamilyOptions(options_));
}
while (cf_descriptors.size() < (size_t)FLAGS_column_families) {
std::string name = ToString(new_column_family_name_.load());
new_column_family_name_++;
cf_descriptors.emplace_back(name, ColumnFamilyOptions(options_));
column_family_names_.push_back(name);
}
options_.listeners.clear();
Experimental support for SST unique IDs (#8990) Summary: * New public header unique_id.h and function GetUniqueIdFromTableProperties which computes a universally unique identifier based on table properties of table files from recent RocksDB versions. * Generation of DB session IDs is refactored so that they are guaranteed unique in the lifetime of a process running RocksDB. (SemiStructuredUniqueIdGen, new test included.) Along with file numbers, this enables SST unique IDs to be guaranteed unique among SSTs generated in a single process, and "better than random" between processes. See https://github.com/pdillinger/unique_id * In addition to public API producing 'external' unique IDs, there is a function for producing 'internal' unique IDs, with functions for converting between the two. In short, the external ID is "safe" for things people might do with it, and the internal ID enables more "power user" features for the future. Specifically, the external ID goes through a hashing layer so that any subset of bits in the external ID can be used as a hash of the full ID, while also preserving uniqueness guarantees in the first 128 bits (bijective both on first 128 bits and on full 192 bits). Intended follow-up: * Use the internal unique IDs in cache keys. (Avoid conflicts with https://github.com/facebook/rocksdb/issues/8912) (The file offset can be XORed into the third 64-bit value of the unique ID.) * Publish the external unique IDs in FileStorageInfo (https://github.com/facebook/rocksdb/issues/8968) Pull Request resolved: https://github.com/facebook/rocksdb/pull/8990 Test Plan: Unit tests added, and checking of unique ids in stress test. NOTE in stress test we do not generate nearly enough files to thoroughly stress uniqueness, but the test trims off pieces of the ID to check for uniqueness so that we can infer (with some assumptions) stronger properties in the aggregate. Reviewed By: zhichao-cao, mrambacher Differential Revision: D31582865 Pulled By: pdillinger fbshipit-source-id: 1f620c4c86af9abe2a8d177b9ccf2ad2b9f48243
2021-10-19 08:28:28 +02:00
#ifndef ROCKSDB_LITE
options_.listeners.emplace_back(new DbStressListener(
FLAGS_db, options_.db_paths, cf_descriptors, db_stress_listener_env));
Experimental support for SST unique IDs (#8990) Summary: * New public header unique_id.h and function GetUniqueIdFromTableProperties which computes a universally unique identifier based on table properties of table files from recent RocksDB versions. * Generation of DB session IDs is refactored so that they are guaranteed unique in the lifetime of a process running RocksDB. (SemiStructuredUniqueIdGen, new test included.) Along with file numbers, this enables SST unique IDs to be guaranteed unique among SSTs generated in a single process, and "better than random" between processes. See https://github.com/pdillinger/unique_id * In addition to public API producing 'external' unique IDs, there is a function for producing 'internal' unique IDs, with functions for converting between the two. In short, the external ID is "safe" for things people might do with it, and the internal ID enables more "power user" features for the future. Specifically, the external ID goes through a hashing layer so that any subset of bits in the external ID can be used as a hash of the full ID, while also preserving uniqueness guarantees in the first 128 bits (bijective both on first 128 bits and on full 192 bits). Intended follow-up: * Use the internal unique IDs in cache keys. (Avoid conflicts with https://github.com/facebook/rocksdb/issues/8912) (The file offset can be XORed into the third 64-bit value of the unique ID.) * Publish the external unique IDs in FileStorageInfo (https://github.com/facebook/rocksdb/issues/8968) Pull Request resolved: https://github.com/facebook/rocksdb/pull/8990 Test Plan: Unit tests added, and checking of unique ids in stress test. NOTE in stress test we do not generate nearly enough files to thoroughly stress uniqueness, but the test trims off pieces of the ID to check for uniqueness so that we can infer (with some assumptions) stronger properties in the aggregate. Reviewed By: zhichao-cao, mrambacher Differential Revision: D31582865 Pulled By: pdillinger fbshipit-source-id: 1f620c4c86af9abe2a8d177b9ccf2ad2b9f48243
2021-10-19 08:28:28 +02:00
#endif // !ROCKSDB_LITE
options_.create_missing_column_families = true;
if (!FLAGS_use_txn) {
#ifndef NDEBUG
// Determine whether we need to ingest file metadata write failures
// during DB reopen. If it does, enable it.
// Only ingest metadata error if it is reopening, as initial open
// failure doesn't need to be handled.
// TODO cover transaction DB is not covered in this fault test too.
bool ingest_meta_error = false;
bool ingest_write_error = false;
bool ingest_read_error = false;
if ((FLAGS_open_metadata_write_fault_one_in ||
FLAGS_open_write_fault_one_in || FLAGS_open_read_fault_one_in) &&
fault_fs_guard
->FileExists(FLAGS_db + "/CURRENT", IOOptions(), nullptr)
.ok()) {
if (!FLAGS_sync) {
// When DB Stress is not sync mode, we expect all WAL writes to
// WAL is durable. Buffering unsynced writes will cause false
// positive in crash tests. Before we figure out a way to
// solve it, skip WAL from failure injection.
fault_fs_guard->SetSkipDirectWritableTypes({kWalFile});
}
ingest_meta_error = FLAGS_open_metadata_write_fault_one_in;
ingest_write_error = FLAGS_open_write_fault_one_in;
ingest_read_error = FLAGS_open_read_fault_one_in;
if (ingest_meta_error) {
fault_fs_guard->EnableMetadataWriteErrorInjection();
fault_fs_guard->SetRandomMetadataWriteError(
FLAGS_open_metadata_write_fault_one_in);
}
if (ingest_write_error) {
fault_fs_guard->SetFilesystemDirectWritable(false);
fault_fs_guard->EnableWriteErrorInjection();
fault_fs_guard->SetRandomWriteError(
static_cast<uint32_t>(FLAGS_seed), FLAGS_open_write_fault_one_in,
IOStatus::IOError("Injected Open Error"),
/*inject_for_all_file_types=*/true, /*types=*/{});
}
if (ingest_read_error) {
fault_fs_guard->SetRandomReadError(FLAGS_open_read_fault_one_in);
}
}
while (true) {
#endif // NDEBUG
#ifndef ROCKSDB_LITE
// StackableDB-based BlobDB
if (FLAGS_use_blob_db) {
blob_db::BlobDBOptions blob_db_options;
blob_db_options.min_blob_size = FLAGS_blob_db_min_blob_size;
blob_db_options.bytes_per_sync = FLAGS_blob_db_bytes_per_sync;
blob_db_options.blob_file_size = FLAGS_blob_db_file_size;
blob_db_options.enable_garbage_collection = FLAGS_blob_db_enable_gc;
blob_db_options.garbage_collection_cutoff = FLAGS_blob_db_gc_cutoff;
blob_db::BlobDB* blob_db = nullptr;
s = blob_db::BlobDB::Open(options_, blob_db_options, FLAGS_db,
cf_descriptors, &column_families_,
&blob_db);
if (s.ok()) {
db_ = blob_db;
}
} else
#endif // !ROCKSDB_LITE
{
if (db_preload_finished_.load() && FLAGS_read_only) {
s = DB::OpenForReadOnly(DBOptions(options_), FLAGS_db,
cf_descriptors, &column_families_, &db_);
} else {
s = DB::Open(DBOptions(options_), FLAGS_db, cf_descriptors,
&column_families_, &db_);
}
}
#ifndef NDEBUG
if (ingest_meta_error || ingest_write_error || ingest_read_error) {
fault_fs_guard->SetFilesystemDirectWritable(true);
fault_fs_guard->DisableMetadataWriteErrorInjection();
fault_fs_guard->DisableWriteErrorInjection();
fault_fs_guard->SetSkipDirectWritableTypes({});
fault_fs_guard->SetRandomReadError(0);
if (s.ok()) {
// Ingested errors might happen in background compactions. We
// wait for all compactions to finish to make sure DB is in
// clean state before executing queries.
s = static_cast_with_check<DBImpl>(db_->GetRootDB())
->TEST_WaitForCompact(true);
if (!s.ok()) {
for (auto cf : column_families_) {
delete cf;
}
column_families_.clear();
delete db_;
db_ = nullptr;
}
}
if (!s.ok()) {
// After failure to opening a DB due to IO error, retry should
// successfully open the DB with correct data if no IO error shows
// up.
ingest_meta_error = false;
ingest_write_error = false;
ingest_read_error = false;
Random rand(static_cast<uint32_t>(FLAGS_seed));
if (rand.OneIn(2)) {
fault_fs_guard->DeleteFilesCreatedAfterLastDirSync(IOOptions(),
nullptr);
}
if (rand.OneIn(3)) {
fault_fs_guard->DropUnsyncedFileData();
} else if (rand.OneIn(2)) {
fault_fs_guard->DropRandomUnsyncedFileData(&rand);
}
continue;
}
}
break;
}
#endif // NDEBUG
} else {
#ifndef ROCKSDB_LITE
TransactionDBOptions txn_db_options;
assert(FLAGS_txn_write_policy <= TxnDBWritePolicy::WRITE_UNPREPARED);
txn_db_options.write_policy =
static_cast<TxnDBWritePolicy>(FLAGS_txn_write_policy);
if (FLAGS_unordered_write) {
assert(txn_db_options.write_policy == TxnDBWritePolicy::WRITE_PREPARED);
options_.unordered_write = true;
options_.two_write_queues = true;
txn_db_options.skip_concurrency_control = true;
Improve stress test for transactions (#9568) Summary: Test only, no change to functionality. Extremely low risk of library regression. Update test key generation by maintaining existing and non-existing keys. Update db_crashtest.py to drive multiops_txn stress test for both write-committed and write-prepared. Add a make target 'blackbox_crash_test_with_multiops_txn'. Running the following commands caught the bug exposed in https://github.com/facebook/rocksdb/issues/9571. ``` $rm -rf /tmp/rocksdbtest/* $./db_stress -progress_reports=0 -test_multi_ops_txns -use_txn -clear_column_family_one_in=0 \ -column_families=1 -writepercent=0 -delpercent=0 -delrangepercent=0 -customopspercent=60 \ -readpercent=20 -prefixpercent=0 -iterpercent=20 -reopen=0 -ops_per_thread=1000 -ub_a=10000 \ -ub_c=100 -destroy_db_initially=0 -key_spaces_path=/dev/shm/key_spaces_desc -threads=32 -read_fault_one_in=0 $./db_stress -progress_reports=0 -test_multi_ops_txns -use_txn -clear_column_family_one_in=0 -column_families=1 -writepercent=0 -delpercent=0 -delrangepercent=0 -customopspercent=60 -readpercent=20 \ -prefixpercent=0 -iterpercent=20 -reopen=0 -ops_per_thread=1000 -ub_a=10000 -ub_c=100 -destroy_db_initially=0 \ -key_spaces_path=/dev/shm/key_spaces_desc -threads=32 -read_fault_one_in=0 ``` Running the following command caught a bug which will be fixed in https://github.com/facebook/rocksdb/issues/9648 . ``` $TEST_TMPDIR=/dev/shm make blackbox_crash_test_with_multiops_wc_txn ``` Pull Request resolved: https://github.com/facebook/rocksdb/pull/9568 Reviewed By: jay-zhuang Differential Revision: D34308154 Pulled By: riversand963 fbshipit-source-id: 99ff1b65c19b46c471d2f2d3b47adcd342a1b9e7
2022-03-17 03:00:04 +01:00
} else {
options_.two_write_queues = FLAGS_two_write_queues;
}
Improve stress test for transactions (#9568) Summary: Test only, no change to functionality. Extremely low risk of library regression. Update test key generation by maintaining existing and non-existing keys. Update db_crashtest.py to drive multiops_txn stress test for both write-committed and write-prepared. Add a make target 'blackbox_crash_test_with_multiops_txn'. Running the following commands caught the bug exposed in https://github.com/facebook/rocksdb/issues/9571. ``` $rm -rf /tmp/rocksdbtest/* $./db_stress -progress_reports=0 -test_multi_ops_txns -use_txn -clear_column_family_one_in=0 \ -column_families=1 -writepercent=0 -delpercent=0 -delrangepercent=0 -customopspercent=60 \ -readpercent=20 -prefixpercent=0 -iterpercent=20 -reopen=0 -ops_per_thread=1000 -ub_a=10000 \ -ub_c=100 -destroy_db_initially=0 -key_spaces_path=/dev/shm/key_spaces_desc -threads=32 -read_fault_one_in=0 $./db_stress -progress_reports=0 -test_multi_ops_txns -use_txn -clear_column_family_one_in=0 -column_families=1 -writepercent=0 -delpercent=0 -delrangepercent=0 -customopspercent=60 -readpercent=20 \ -prefixpercent=0 -iterpercent=20 -reopen=0 -ops_per_thread=1000 -ub_a=10000 -ub_c=100 -destroy_db_initially=0 \ -key_spaces_path=/dev/shm/key_spaces_desc -threads=32 -read_fault_one_in=0 ``` Running the following command caught a bug which will be fixed in https://github.com/facebook/rocksdb/issues/9648 . ``` $TEST_TMPDIR=/dev/shm make blackbox_crash_test_with_multiops_wc_txn ``` Pull Request resolved: https://github.com/facebook/rocksdb/pull/9568 Reviewed By: jay-zhuang Differential Revision: D34308154 Pulled By: riversand963 fbshipit-source-id: 99ff1b65c19b46c471d2f2d3b47adcd342a1b9e7
2022-03-17 03:00:04 +01:00
txn_db_options.wp_snapshot_cache_bits =
static_cast<size_t>(FLAGS_wp_snapshot_cache_bits);
txn_db_options.wp_commit_cache_bits =
static_cast<size_t>(FLAGS_wp_commit_cache_bits);
s = TransactionDB::Open(options_, txn_db_options, FLAGS_db,
cf_descriptors, &column_families_, &txn_db_);
if (!s.ok()) {
fprintf(stderr, "Error in opening the TransactionDB [%s]\n",
s.ToString().c_str());
fflush(stderr);
}
assert(s.ok());
db_ = txn_db_;
// after a crash, rollback to commit recovered transactions
std::vector<Transaction*> trans;
txn_db_->GetAllPreparedTransactions(&trans);
Random rand(static_cast<uint32_t>(FLAGS_seed));
for (auto txn : trans) {
if (rand.OneIn(2)) {
s = txn->Commit();
assert(s.ok());
} else {
s = txn->Rollback();
assert(s.ok());
}
delete txn;
}
trans.clear();
txn_db_->GetAllPreparedTransactions(&trans);
assert(trans.size() == 0);
#endif
}
assert(!s.ok() || column_families_.size() ==
static_cast<size_t>(FLAGS_column_families));
if (s.ok() && FLAGS_test_secondary) {
#ifndef ROCKSDB_LITE
secondaries_.resize(FLAGS_threads);
std::fill(secondaries_.begin(), secondaries_.end(), nullptr);
secondary_cfh_lists_.clear();
secondary_cfh_lists_.resize(FLAGS_threads);
Options tmp_opts;
// TODO(yanqin) support max_open_files != -1 for secondary instance.
tmp_opts.max_open_files = -1;
tmp_opts.statistics = dbstats_secondaries;
tmp_opts.env = db_stress_env;
for (size_t i = 0; i != static_cast<size_t>(FLAGS_threads); ++i) {
const std::string secondary_path =
FLAGS_secondaries_base + "/" + std::to_string(i);
s = DB::OpenAsSecondary(tmp_opts, FLAGS_db, secondary_path,
cf_descriptors, &secondary_cfh_lists_[i],
&secondaries_[i]);
if (!s.ok()) {
break;
}
}
#else
fprintf(stderr, "Secondary is not supported in RocksDBLite\n");
exit(1);
#endif
}
Improve stress test for transactions (#9568) Summary: Test only, no change to functionality. Extremely low risk of library regression. Update test key generation by maintaining existing and non-existing keys. Update db_crashtest.py to drive multiops_txn stress test for both write-committed and write-prepared. Add a make target 'blackbox_crash_test_with_multiops_txn'. Running the following commands caught the bug exposed in https://github.com/facebook/rocksdb/issues/9571. ``` $rm -rf /tmp/rocksdbtest/* $./db_stress -progress_reports=0 -test_multi_ops_txns -use_txn -clear_column_family_one_in=0 \ -column_families=1 -writepercent=0 -delpercent=0 -delrangepercent=0 -customopspercent=60 \ -readpercent=20 -prefixpercent=0 -iterpercent=20 -reopen=0 -ops_per_thread=1000 -ub_a=10000 \ -ub_c=100 -destroy_db_initially=0 -key_spaces_path=/dev/shm/key_spaces_desc -threads=32 -read_fault_one_in=0 $./db_stress -progress_reports=0 -test_multi_ops_txns -use_txn -clear_column_family_one_in=0 -column_families=1 -writepercent=0 -delpercent=0 -delrangepercent=0 -customopspercent=60 -readpercent=20 \ -prefixpercent=0 -iterpercent=20 -reopen=0 -ops_per_thread=1000 -ub_a=10000 -ub_c=100 -destroy_db_initially=0 \ -key_spaces_path=/dev/shm/key_spaces_desc -threads=32 -read_fault_one_in=0 ``` Running the following command caught a bug which will be fixed in https://github.com/facebook/rocksdb/issues/9648 . ``` $TEST_TMPDIR=/dev/shm make blackbox_crash_test_with_multiops_wc_txn ``` Pull Request resolved: https://github.com/facebook/rocksdb/pull/9568 Reviewed By: jay-zhuang Differential Revision: D34308154 Pulled By: riversand963 fbshipit-source-id: 99ff1b65c19b46c471d2f2d3b47adcd342a1b9e7
2022-03-17 03:00:04 +01:00
// Secondary instance does not support write-prepared/write-unprepared
// transactions, thus just disable secondary instance if we use
// transaction.
if (s.ok() && FLAGS_continuous_verification_interval > 0 &&
!FLAGS_use_txn && !cmp_db_) {
Options tmp_opts;
// TODO(yanqin) support max_open_files != -1 for secondary instance.
tmp_opts.max_open_files = -1;
tmp_opts.env = db_stress_env;
std::string secondary_path = FLAGS_secondaries_base + "/cmp_database";
s = DB::OpenAsSecondary(tmp_opts, FLAGS_db, secondary_path,
cf_descriptors, &cmp_cfhs_, &cmp_db_);
assert(!s.ok() ||
cmp_cfhs_.size() == static_cast<size_t>(FLAGS_column_families));
}
} else {
#ifndef ROCKSDB_LITE
DBWithTTL* db_with_ttl;
s = DBWithTTL::Open(options_, FLAGS_db, &db_with_ttl, FLAGS_ttl);
db_ = db_with_ttl;
if (FLAGS_test_secondary) {
secondaries_.resize(FLAGS_threads);
std::fill(secondaries_.begin(), secondaries_.end(), nullptr);
Options tmp_opts;
tmp_opts.env = options_.env;
// TODO(yanqin) support max_open_files != -1 for secondary instance.
tmp_opts.max_open_files = -1;
for (size_t i = 0; i != static_cast<size_t>(FLAGS_threads); ++i) {
const std::string secondary_path =
FLAGS_secondaries_base + "/" + std::to_string(i);
s = DB::OpenAsSecondary(tmp_opts, FLAGS_db, secondary_path,
&secondaries_[i]);
if (!s.ok()) {
break;
}
}
}
#else
fprintf(stderr, "TTL is not supported in RocksDBLite\n");
exit(1);
#endif
}
if (!s.ok()) {
fprintf(stderr, "open error: %s\n", s.ToString().c_str());
exit(1);
}
}
void StressTest::Reopen(ThreadState* thread) {
#ifndef ROCKSDB_LITE
// BG jobs in WritePrepared must be canceled first because i) they can access
// the db via a callbac ii) they hold on to a snapshot and the upcoming
// ::Close would complain about it.
const bool write_prepared = FLAGS_use_txn && FLAGS_txn_write_policy != 0;
bool bg_canceled __attribute__((unused)) = false;
if (write_prepared || thread->rand.OneIn(2)) {
const bool wait =
write_prepared || static_cast<bool>(thread->rand.OneIn(2));
CancelAllBackgroundWork(db_, wait);
bg_canceled = wait;
}
assert(!write_prepared || bg_canceled);
#else
(void) thread;
#endif
for (auto cf : column_families_) {
delete cf;
}
column_families_.clear();
#ifndef ROCKSDB_LITE
if (thread->rand.OneIn(2)) {
Status s = db_->Close();
if (!s.ok()) {
fprintf(stderr, "Non-ok close status: %s\n", s.ToString().c_str());
fflush(stderr);
}
assert(s.ok());
}
#endif
delete db_;
db_ = nullptr;
#ifndef ROCKSDB_LITE
txn_db_ = nullptr;
#endif
assert(secondaries_.size() == secondary_cfh_lists_.size());
size_t n = secondaries_.size();
for (size_t i = 0; i != n; ++i) {
for (auto* cf : secondary_cfh_lists_[i]) {
delete cf;
}
secondary_cfh_lists_[i].clear();
delete secondaries_[i];
}
secondaries_.clear();
num_times_reopened_++;
auto now = clock_->NowMicros();
fprintf(stdout, "%s Reopening database for the %dth time\n",
clock_->TimeToString(now / 1000000).c_str(), num_times_reopened_);
Open();
if ((FLAGS_sync_fault_injection || FLAGS_disable_wal) && IsStateTracked()) {
Status s = thread->shared->SaveAtAndAfter(db_);
if (!s.ok()) {
fprintf(stderr, "Error enabling history tracing: %s\n",
s.ToString().c_str());
exit(1);
}
}
}
void StressTest::CheckAndSetOptionsForUserTimestamp() {
assert(FLAGS_user_timestamp_size > 0);
const Comparator* const cmp = test::BytewiseComparatorWithU64TsWrapper();
assert(cmp);
if (FLAGS_user_timestamp_size != cmp->timestamp_size()) {
fprintf(stderr,
"Only -user_timestamp_size=%d is supported in stress test.\n",
static_cast<int>(cmp->timestamp_size()));
exit(1);
}
if (FLAGS_use_merge || FLAGS_use_full_merge_v1) {
fprintf(stderr, "Merge does not support timestamp yet.\n");
exit(1);
}
if (FLAGS_delrangepercent > 0) {
fprintf(stderr, "DeleteRange does not support timestamp yet.\n");
exit(1);
}
if (FLAGS_use_txn) {
fprintf(stderr, "TransactionDB does not support timestamp yet.\n");
exit(1);
}
if (FLAGS_read_only) {
fprintf(stderr, "When opened as read-only, timestamp not supported.\n");
exit(1);
}
if (FLAGS_test_secondary || FLAGS_secondary_catch_up_one_in > 0 ||
FLAGS_continuous_verification_interval > 0) {
fprintf(stderr, "Secondary instance does not support timestamp.\n");
exit(1);
}
if (FLAGS_checkpoint_one_in > 0) {
fprintf(stderr,
"-checkpoint_one_in=%d requires "
"DBImplReadOnly, which is not supported with timestamp\n",
FLAGS_checkpoint_one_in);
exit(1);
}
#ifndef ROCKSDB_LITE
if (FLAGS_enable_blob_files || FLAGS_use_blob_db) {
fprintf(stderr, "BlobDB not supported with timestamp.\n");
exit(1);
}
#endif // !ROCKSDB_LITE
if (FLAGS_enable_compaction_filter) {
fprintf(stderr, "CompactionFilter not supported with timestamp.\n");
exit(1);
}
if (FLAGS_test_cf_consistency || FLAGS_test_batches_snapshots) {
fprintf(stderr,
"Due to per-key ts-seq ordering constraint, only the (default) "
"non-batched test is supported with timestamp.\n");
exit(1);
}
if (FLAGS_ingest_external_file_one_in > 0) {
fprintf(stderr, "Bulk loading may not support timestamp yet.\n");
exit(1);
}
options_.comparator = cmp;
}
} // namespace ROCKSDB_NAMESPACE
#endif // GFLAGS