andreacavalli/rocksdb
0
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity
52aab66424
rocksdb/options/options.cc
Sagar Vemuri ce3b2912d2 Level Compaction with TTL 
Summary:
Level Compaction with TTL.

As of today, a file could exist in the LSM tree without going through the compaction process for a really long time if there are no updates to the data in the file's key range. For example, in certain use cases, the keys are not actually "deleted"; instead they are just set to empty values. There might not be any more writes to this "deleted" key range, and if so, such data could remain in the LSM for a really long time resulting in wasted space.

Introducing a TTL could solve this problem. Files (and, in turn, data) older than TTL will be scheduled for compaction when there is no other background work. This will make the data go through the regular compaction process and get rid of old unwanted data.
This also has the (good) side-effect of all the data in the non-bottommost level being newer than ttl, and all data in the bottommost level older than ttl. It could lead to more writes while reducing space.

This functionality can be controlled by the newly introduced column family option -- ttl.

TODO for later:
- Make ttl mutable
- Extend TTL to Universal compaction as well? (TTL is already supported in FIFO)
- Maybe deprecate CompactionOptionsFIFO.ttl in favor of this new ttl option.
Closes https://github.com/facebook/rocksdb/pull/3591

Differential Revision: D7275442

Pulled By: sagar0

fbshipit-source-id: dcba484717341200d419b0953dafcdf9eb2f0267
2018-04-13 04:51:17 -07:00

557 lines
23 KiB
C++
Raw Blame 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.
#include "rocksdb/options.h"
#ifndef __STDC_FORMAT_MACROS
#define __STDC_FORMAT_MACROS
#endif
#include <inttypes.h>
#include <limits>
#include "monitoring/statistics.h"
#include "options/db_options.h"
#include "options/options_helper.h"
#include "rocksdb/cache.h"
#include "rocksdb/compaction_filter.h"
#include "rocksdb/comparator.h"
#include "rocksdb/env.h"
#include "rocksdb/memtablerep.h"
#include "rocksdb/merge_operator.h"
#include "rocksdb/slice.h"
#include "rocksdb/slice_transform.h"
#include "rocksdb/sst_file_manager.h"
#include "rocksdb/table.h"
#include "rocksdb/table_properties.h"
#include "rocksdb/wal_filter.h"
#include "table/block_based_table_factory.h"
#include "util/compression.h"
namespace rocksdb {
AdvancedColumnFamilyOptions::AdvancedColumnFamilyOptions() {
assert(memtable_factory.get() != nullptr);
}
AdvancedColumnFamilyOptions::AdvancedColumnFamilyOptions(const Options& options)
: max_write_buffer_number(options.max_write_buffer_number),
min_write_buffer_number_to_merge(
options.min_write_buffer_number_to_merge),
max_write_buffer_number_to_maintain(
options.max_write_buffer_number_to_maintain),
inplace_update_support(options.inplace_update_support),
inplace_update_num_locks(options.inplace_update_num_locks),
inplace_callback(options.inplace_callback),
memtable_prefix_bloom_size_ratio(
options.memtable_prefix_bloom_size_ratio),
memtable_huge_page_size(options.memtable_huge_page_size),
memtable_insert_with_hint_prefix_extractor(
options.memtable_insert_with_hint_prefix_extractor),
bloom_locality(options.bloom_locality),
arena_block_size(options.arena_block_size),
compression_per_level(options.compression_per_level),
num_levels(options.num_levels),
level0_slowdown_writes_trigger(options.level0_slowdown_writes_trigger),
level0_stop_writes_trigger(options.level0_stop_writes_trigger),
target_file_size_base(options.target_file_size_base),
target_file_size_multiplier(options.target_file_size_multiplier),
level_compaction_dynamic_level_bytes(
options.level_compaction_dynamic_level_bytes),
max_bytes_for_level_multiplier(options.max_bytes_for_level_multiplier),
max_bytes_for_level_multiplier_additional(
options.max_bytes_for_level_multiplier_additional),
max_compaction_bytes(options.max_compaction_bytes),
soft_pending_compaction_bytes_limit(
options.soft_pending_compaction_bytes_limit),
hard_pending_compaction_bytes_limit(
options.hard_pending_compaction_bytes_limit),
compaction_style(options.compaction_style),
compaction_pri(options.compaction_pri),
compaction_options_universal(options.compaction_options_universal),
compaction_options_fifo(options.compaction_options_fifo),
max_sequential_skip_in_iterations(
options.max_sequential_skip_in_iterations),
memtable_factory(options.memtable_factory),
table_properties_collector_factories(
options.table_properties_collector_factories),
max_successive_merges(options.max_successive_merges),
optimize_filters_for_hits(options.optimize_filters_for_hits),
paranoid_file_checks(options.paranoid_file_checks),
force_consistency_checks(options.force_consistency_checks),
report_bg_io_stats(options.report_bg_io_stats),
ttl(options.ttl) {
assert(memtable_factory.get() != nullptr);
if (max_bytes_for_level_multiplier_additional.size() <
static_cast<unsigned int>(num_levels)) {
max_bytes_for_level_multiplier_additional.resize(num_levels, 1);
}
}
ColumnFamilyOptions::ColumnFamilyOptions()
: compression(Snappy_Supported() ? kSnappyCompression : kNoCompression),
table_factory(
std::shared_ptr<TableFactory>(new BlockBasedTableFactory())) {}
ColumnFamilyOptions::ColumnFamilyOptions(const Options& options)
: ColumnFamilyOptions(*static_cast<const ColumnFamilyOptions*>(&options)) {}
DBOptions::DBOptions() {}
DBOptions::DBOptions(const Options& options)
: DBOptions(*static_cast<const DBOptions*>(&options)) {}
void DBOptions::Dump(Logger* log) const {
ImmutableDBOptions(*this).Dump(log);
MutableDBOptions(*this).Dump(log);
} // DBOptions::Dump
void ColumnFamilyOptions::Dump(Logger* log) const {
ROCKS_LOG_HEADER(log, " Options.comparator: %s",
comparator->Name());
ROCKS_LOG_HEADER(log, " Options.merge_operator: %s",
merge_operator ? merge_operator->Name() : "None");
ROCKS_LOG_HEADER(log, " Options.compaction_filter: %s",
compaction_filter ? compaction_filter->Name() : "None");
ROCKS_LOG_HEADER(
log, " Options.compaction_filter_factory: %s",
compaction_filter_factory ? compaction_filter_factory->Name() : "None");
ROCKS_LOG_HEADER(log, " Options.memtable_factory: %s",
memtable_factory->Name());
ROCKS_LOG_HEADER(log, " Options.table_factory: %s",
table_factory->Name());
ROCKS_LOG_HEADER(log, " table_factory options: %s",
table_factory->GetPrintableTableOptions().c_str());
ROCKS_LOG_HEADER(log, " Options.write_buffer_size: %" ROCKSDB_PRIszt,
write_buffer_size);
ROCKS_LOG_HEADER(log, " Options.max_write_buffer_number: %d",
max_write_buffer_number);
if (!compression_per_level.empty()) {
for (unsigned int i = 0; i < compression_per_level.size(); i++) {
ROCKS_LOG_HEADER(
log, " Options.compression[%d]: %s", i,
CompressionTypeToString(compression_per_level[i]).c_str());
}
} else {
ROCKS_LOG_HEADER(log, " Options.compression: %s",
CompressionTypeToString(compression).c_str());
}
ROCKS_LOG_HEADER(
log, " Options.bottommost_compression: %s",
bottommost_compression == kDisableCompressionOption
? "Disabled"
: CompressionTypeToString(bottommost_compression).c_str());
ROCKS_LOG_HEADER(
log, " Options.prefix_extractor: %s",
prefix_extractor == nullptr ? "nullptr" : prefix_extractor->Name());
ROCKS_LOG_HEADER(log,
" Options.memtable_insert_with_hint_prefix_extractor: %s",
memtable_insert_with_hint_prefix_extractor == nullptr
? "nullptr"
: memtable_insert_with_hint_prefix_extractor->Name());
ROCKS_LOG_HEADER(log, " Options.num_levels: %d", num_levels);
ROCKS_LOG_HEADER(log, " Options.min_write_buffer_number_to_merge: %d",
min_write_buffer_number_to_merge);
ROCKS_LOG_HEADER(log, " Options.max_write_buffer_number_to_maintain: %d",
max_write_buffer_number_to_maintain);
ROCKS_LOG_HEADER(log, " Options.compression_opts.window_bits: %d",
compression_opts.window_bits);
ROCKS_LOG_HEADER(log, " Options.compression_opts.level: %d",
compression_opts.level);
ROCKS_LOG_HEADER(log, " Options.compression_opts.strategy: %d",
compression_opts.strategy);
ROCKS_LOG_HEADER(
log,
" Options.compression_opts.max_dict_bytes: %" ROCKSDB_PRIszt,
compression_opts.max_dict_bytes);
ROCKS_LOG_HEADER(log,
" Options.compression_opts.zstd_max_train_bytes: "
"%" ROCKSDB_PRIszt,
compression_opts.zstd_max_train_bytes);
ROCKS_LOG_HEADER(log, " Options.level0_file_num_compaction_trigger: %d",
level0_file_num_compaction_trigger);
ROCKS_LOG_HEADER(log, " Options.level0_slowdown_writes_trigger: %d",
level0_slowdown_writes_trigger);
ROCKS_LOG_HEADER(log, " Options.level0_stop_writes_trigger: %d",
level0_stop_writes_trigger);
ROCKS_LOG_HEADER(
log, " Options.target_file_size_base: %" PRIu64,
target_file_size_base);
ROCKS_LOG_HEADER(log, " Options.target_file_size_multiplier: %d",
target_file_size_multiplier);
ROCKS_LOG_HEADER(
log, " Options.max_bytes_for_level_base: %" PRIu64,
max_bytes_for_level_base);
ROCKS_LOG_HEADER(log, "Options.level_compaction_dynamic_level_bytes: %d",
level_compaction_dynamic_level_bytes);
ROCKS_LOG_HEADER(log, " Options.max_bytes_for_level_multiplier: %f",
max_bytes_for_level_multiplier);
for (size_t i = 0; i < max_bytes_for_level_multiplier_additional.size();
i++) {
ROCKS_LOG_HEADER(
log, "Options.max_bytes_for_level_multiplier_addtl[%" ROCKSDB_PRIszt
"]: %d",
i, max_bytes_for_level_multiplier_additional[i]);
}
ROCKS_LOG_HEADER(
log, " Options.max_sequential_skip_in_iterations: %" PRIu64,
max_sequential_skip_in_iterations);
ROCKS_LOG_HEADER(
log, " Options.max_compaction_bytes: %" PRIu64,
max_compaction_bytes);
ROCKS_LOG_HEADER(
log,
" Options.arena_block_size: %" ROCKSDB_PRIszt,
arena_block_size);
ROCKS_LOG_HEADER(log,
" Options.soft_pending_compaction_bytes_limit: %" PRIu64,
soft_pending_compaction_bytes_limit);
ROCKS_LOG_HEADER(log,
" Options.hard_pending_compaction_bytes_limit: %" PRIu64,
hard_pending_compaction_bytes_limit);
ROCKS_LOG_HEADER(log, " Options.rate_limit_delay_max_milliseconds: %u",
rate_limit_delay_max_milliseconds);
ROCKS_LOG_HEADER(log, " Options.disable_auto_compactions: %d",
disable_auto_compactions);
const auto& it_compaction_style =
compaction_style_to_string.find(compaction_style);
std::string str_compaction_style;
if (it_compaction_style == compaction_style_to_string.end()) {
assert(false);
str_compaction_style = "unknown_" + std::to_string(compaction_style);
} else {
str_compaction_style = it_compaction_style->second;
}
ROCKS_LOG_HEADER(log,
" Options.compaction_style: %s",
str_compaction_style.c_str());
const auto& it_compaction_pri =
compaction_pri_to_string.find(compaction_pri);
std::string str_compaction_pri;
if (it_compaction_pri == compaction_pri_to_string.end()) {
assert(false);
str_compaction_pri = "unknown_" + std::to_string(compaction_pri);
} else {
str_compaction_pri = it_compaction_pri->second;
}
ROCKS_LOG_HEADER(log,
" Options.compaction_pri: %s",
str_compaction_pri.c_str());
ROCKS_LOG_HEADER(log,
"Options.compaction_options_universal.size_ratio: %u",
compaction_options_universal.size_ratio);
ROCKS_LOG_HEADER(log,
"Options.compaction_options_universal.min_merge_width: %u",
compaction_options_universal.min_merge_width);
ROCKS_LOG_HEADER(log,
"Options.compaction_options_universal.max_merge_width: %u",
compaction_options_universal.max_merge_width);
ROCKS_LOG_HEADER(
log,
"Options.compaction_options_universal."
"max_size_amplification_percent: %u",
compaction_options_universal.max_size_amplification_percent);
ROCKS_LOG_HEADER(
log,
"Options.compaction_options_universal.compression_size_percent: %d",
compaction_options_universal.compression_size_percent);
const auto& it_compaction_stop_style = compaction_stop_style_to_string.find(
compaction_options_universal.stop_style);
std::string str_compaction_stop_style;
if (it_compaction_stop_style == compaction_stop_style_to_string.end()) {
assert(false);
str_compaction_stop_style =
"unknown_" + std::to_string(compaction_options_universal.stop_style);
} else {
str_compaction_stop_style = it_compaction_stop_style->second;
}
ROCKS_LOG_HEADER(log,
"Options.compaction_options_universal.stop_style: %s",
str_compaction_stop_style.c_str());
ROCKS_LOG_HEADER(
log, "Options.compaction_options_fifo.max_table_files_size: %" PRIu64,
compaction_options_fifo.max_table_files_size);
ROCKS_LOG_HEADER(log,
"Options.compaction_options_fifo.allow_compaction: %d",
compaction_options_fifo.allow_compaction);
ROCKS_LOG_HEADER(log, "Options.compaction_options_fifo.ttl: %" PRIu64,
compaction_options_fifo.ttl);
std::string collector_names;
for (const auto& collector_factory : table_properties_collector_factories) {
collector_names.append(collector_factory->Name());
collector_names.append("; ");
}
ROCKS_LOG_HEADER(
log, " Options.table_properties_collectors: %s",
collector_names.c_str());
ROCKS_LOG_HEADER(log,
" Options.inplace_update_support: %d",
inplace_update_support);
ROCKS_LOG_HEADER(
log,
" Options.inplace_update_num_locks: %" ROCKSDB_PRIszt,
inplace_update_num_locks);
// TODO: easier config for bloom (maybe based on avg key/value size)
ROCKS_LOG_HEADER(
log, " Options.memtable_prefix_bloom_size_ratio: %f",
memtable_prefix_bloom_size_ratio);
ROCKS_LOG_HEADER(log, " Options.memtable_huge_page_size: %" ROCKSDB_PRIszt,
memtable_huge_page_size);
ROCKS_LOG_HEADER(log,
" Options.bloom_locality: %d",
bloom_locality);
ROCKS_LOG_HEADER(
log,
" Options.max_successive_merges: %" ROCKSDB_PRIszt,
max_successive_merges);
ROCKS_LOG_HEADER(log,
" Options.optimize_filters_for_hits: %d",
optimize_filters_for_hits);
ROCKS_LOG_HEADER(log, " Options.paranoid_file_checks: %d",
paranoid_file_checks);
ROCKS_LOG_HEADER(log, " Options.force_consistency_checks: %d",
force_consistency_checks);
ROCKS_LOG_HEADER(log, " Options.report_bg_io_stats: %d",
report_bg_io_stats);
ROCKS_LOG_HEADER(log, " Options.ttl: %d", ttl);
} // ColumnFamilyOptions::Dump
void Options::Dump(Logger* log) const {
DBOptions::Dump(log);
ColumnFamilyOptions::Dump(log);
} // Options::Dump
void Options::DumpCFOptions(Logger* log) const {
ColumnFamilyOptions::Dump(log);
} // Options::DumpCFOptions
//
// The goal of this method is to create a configuration that
// allows an application to write all files into L0 and
// then do a single compaction to output all files into L1.
Options*
Options::PrepareForBulkLoad()
{
// never slowdown ingest.
level0_file_num_compaction_trigger = (1<<30);
level0_slowdown_writes_trigger = (1<<30);
level0_stop_writes_trigger = (1<<30);
soft_pending_compaction_bytes_limit = 0;
hard_pending_compaction_bytes_limit = 0;
// no auto compactions please. The application should issue a
// manual compaction after all data is loaded into L0.
disable_auto_compactions = true;
// A manual compaction run should pick all files in L0 in
// a single compaction run.
max_compaction_bytes = (static_cast<uint64_t>(1) << 60);
// It is better to have only 2 levels, otherwise a manual
// compaction would compact at every possible level, thereby
// increasing the total time needed for compactions.
num_levels = 2;
// Need to allow more write buffers to allow more parallism
// of flushes.
max_write_buffer_number = 6;
min_write_buffer_number_to_merge = 1;
// When compaction is disabled, more parallel flush threads can
// help with write throughput.
max_background_flushes = 4;
// Prevent a memtable flush to automatically promote files
// to L1. This is helpful so that all files that are
// input to the manual compaction are all at L0.
max_background_compactions = 2;
// The compaction would create large files in L1.
target_file_size_base = 256 * 1024 * 1024;
return this;
}
Options* Options::OptimizeForSmallDb() {
ColumnFamilyOptions::OptimizeForSmallDb();
DBOptions::OptimizeForSmallDb();
return this;
}
Options* Options::OldDefaults(int rocksdb_major_version,
int rocksdb_minor_version) {
ColumnFamilyOptions::OldDefaults(rocksdb_major_version,
rocksdb_minor_version);
DBOptions::OldDefaults(rocksdb_major_version, rocksdb_minor_version);
return this;
}
DBOptions* DBOptions::OldDefaults(int rocksdb_major_version,
int rocksdb_minor_version) {
if (rocksdb_major_version < 4 ||
(rocksdb_major_version == 4 && rocksdb_minor_version < 7)) {
max_file_opening_threads = 1;
table_cache_numshardbits = 4;
}
if (rocksdb_major_version < 5 ||
(rocksdb_major_version == 5 && rocksdb_minor_version < 2)) {
delayed_write_rate = 2 * 1024U * 1024U;
} else if (rocksdb_major_version < 5 ||
(rocksdb_major_version == 5 && rocksdb_minor_version < 6)) {
delayed_write_rate = 16 * 1024U * 1024U;
}
max_open_files = 5000;
wal_recovery_mode = WALRecoveryMode::kTolerateCorruptedTailRecords;
return this;
}
ColumnFamilyOptions* ColumnFamilyOptions::OldDefaults(
int rocksdb_major_version, int rocksdb_minor_version) {
if (rocksdb_major_version < 4 ||
(rocksdb_major_version == 4 && rocksdb_minor_version < 7)) {
write_buffer_size = 4 << 20;
target_file_size_base = 2 * 1048576;
max_bytes_for_level_base = 10 * 1048576;
soft_pending_compaction_bytes_limit = 0;
hard_pending_compaction_bytes_limit = 0;
}
if (rocksdb_major_version < 5) {
level0_stop_writes_trigger = 24;
} else if (rocksdb_major_version == 5 && rocksdb_minor_version < 2) {
level0_stop_writes_trigger = 30;
}
compaction_pri = CompactionPri::kByCompensatedSize;
return this;
}
// Optimization functions
DBOptions* DBOptions::OptimizeForSmallDb() {
max_file_opening_threads = 1;
max_open_files = 5000;
return this;
}
ColumnFamilyOptions* ColumnFamilyOptions::OptimizeForSmallDb() {
write_buffer_size = 2 << 20;
target_file_size_base = 2 * 1048576;
max_bytes_for_level_base = 10 * 1048576;
soft_pending_compaction_bytes_limit = 256 * 1048576;
hard_pending_compaction_bytes_limit = 1073741824ul;
return this;
}
#ifndef ROCKSDB_LITE
ColumnFamilyOptions* ColumnFamilyOptions::OptimizeForPointLookup(
uint64_t block_cache_size_mb) {
prefix_extractor.reset(NewNoopTransform());
BlockBasedTableOptions block_based_options;
block_based_options.index_type = BlockBasedTableOptions::kHashSearch;
block_based_options.filter_policy.reset(NewBloomFilterPolicy(10));
block_based_options.block_cache =
NewLRUCache(static_cast<size_t>(block_cache_size_mb * 1024 * 1024));
table_factory.reset(new BlockBasedTableFactory(block_based_options));
memtable_prefix_bloom_size_ratio = 0.02;
return this;
}
ColumnFamilyOptions* ColumnFamilyOptions::OptimizeLevelStyleCompaction(
uint64_t memtable_memory_budget) {
write_buffer_size = static_cast<size_t>(memtable_memory_budget / 4);
// merge two memtables when flushing to L0
min_write_buffer_number_to_merge = 2;
// this means we'll use 50% extra memory in the worst case, but will reduce
// write stalls.
max_write_buffer_number = 6;
// start flushing L0->L1 as soon as possible. each file on level0 is
// (memtable_memory_budget / 2). This will flush level 0 when it's bigger than
// memtable_memory_budget.
level0_file_num_compaction_trigger = 2;
// doesn't really matter much, but we don't want to create too many files
target_file_size_base = memtable_memory_budget / 8;
// make Level1 size equal to Level0 size, so that L0->L1 compactions are fast
max_bytes_for_level_base = memtable_memory_budget;
// level style compaction
compaction_style = kCompactionStyleLevel;
// only compress levels >= 2
compression_per_level.resize(num_levels);
for (int i = 0; i < num_levels; ++i) {
if (i < 2) {
compression_per_level[i] = kNoCompression;
} else {
compression_per_level[i] = kSnappyCompression;
}
}
return this;
}
ColumnFamilyOptions* ColumnFamilyOptions::OptimizeUniversalStyleCompaction(
uint64_t memtable_memory_budget) {
write_buffer_size = static_cast<size_t>(memtable_memory_budget / 4);
// merge two memtables when flushing to L0
min_write_buffer_number_to_merge = 2;
// this means we'll use 50% extra memory in the worst case, but will reduce
// write stalls.
max_write_buffer_number = 6;
// universal style compaction
compaction_style = kCompactionStyleUniversal;
compaction_options_universal.compression_size_percent = 80;
return this;
}
DBOptions* DBOptions::IncreaseParallelism(int total_threads) {
max_background_jobs = total_threads;
env->SetBackgroundThreads(total_threads, Env::LOW);
env->SetBackgroundThreads(1, Env::HIGH);
return this;
}
#endif // !ROCKSDB_LITE
ReadOptions::ReadOptions()
: snapshot(nullptr),
iterate_lower_bound(nullptr),
iterate_upper_bound(nullptr),
readahead_size(0),
max_skippable_internal_keys(0),
read_tier(kReadAllTier),
verify_checksums(true),
fill_cache(true),
tailing(false),
managed(false),
total_order_seek(false),
prefix_same_as_start(false),
pin_data(false),
background_purge_on_iterator_cleanup(false),
ignore_range_deletions(false),
iter_start_seqnum(0) {}
ReadOptions::ReadOptions(bool cksum, bool cache)
: snapshot(nullptr),
iterate_lower_bound(nullptr),
iterate_upper_bound(nullptr),
readahead_size(0),
max_skippable_internal_keys(0),
read_tier(kReadAllTier),
verify_checksums(cksum),
fill_cache(cache),
tailing(false),
managed(false),
total_order_seek(false),
prefix_same_as_start(false),
pin_data(false),
background_purge_on_iterator_cleanup(false),
ignore_range_deletions(false),
iter_start_seqnum(0) {}
} // namespace rocksdb
Reference in New Issue View Git Blame Copy Permalink