rocksdb/db_stress_tool/db_stress_gflags.cc

812 lines
34 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"
static bool ValidateUint32Range(const char* flagname, uint64_t value) {
if (value > std::numeric_limits<uint32_t>::max()) {
fprintf(stderr, "Invalid value for --%s: %lu, overflow\n", flagname,
(unsigned long)value);
return false;
}
return true;
}
DEFINE_uint64(seed, 2341234, "Seed for PRNG");
static const bool FLAGS_seed_dummy __attribute__((__unused__)) =
RegisterFlagValidator(&FLAGS_seed, &ValidateUint32Range);
DEFINE_bool(read_only, false, "True if open DB in read-only mode during tests");
DEFINE_int64(max_key, 1 * KB * KB,
"Max number of key/values to place in database");
DEFINE_int32(max_key_len, 3, "Maximum length of a key in 8-byte units");
DEFINE_string(key_len_percent_dist, "",
"Percentages of keys of various lengths. For example, 1,30,69 "
"means 1% of keys are 8 bytes, 30% are 16 bytes, and 69% are "
"24 bytes. If not specified, it will be evenly distributed");
DEFINE_int32(key_window_scale_factor, 10,
"This value will be multiplied by 100 to come up with a window "
"size for varying the key length");
DEFINE_int32(column_families, 10, "Number of column families");
DEFINE_double(
hot_key_alpha, 0,
"Use Zipfian distribution to generate the key "
"distribution. If it is not specified, write path will use random "
"distribution to generate the keys. The parameter is [0, double_max]). "
"However, the larger alpha is, the more shewed will be. If alpha is "
"larger than 2, it is likely that only 1 key will be accessed. The "
"Recommended value is [0.8-1.5]. The distribution is also related to "
"max_key and total iterations of generating the hot key. ");
DEFINE_string(
options_file, "",
"The path to a RocksDB options file. If specified, then db_stress will "
"run with the RocksDB options in the default column family of the "
"specified options file. Note that, when an options file is provided, "
"db_stress will ignore the flag values for all options that may be passed "
"via options file.");
DEFINE_int64(
active_width, 0,
"Number of keys in active span of the key-range at any given time. The "
"span begins with its left endpoint at key 0, gradually moves rightwards, "
"and ends with its right endpoint at max_key. If set to 0, active_width "
"will be sanitized to be equal to max_key.");
// TODO(noetzli) Add support for single deletes
DEFINE_bool(test_batches_snapshots, false,
"If set, the test uses MultiGet(), MultiPut() and MultiDelete()"
" which read/write/delete multiple keys in a batch. In this mode,"
" we do not verify db content by comparing the content with the "
"pre-allocated array. Instead, we do partial verification inside"
" MultiGet() by checking various values in a batch. Benefit of"
" this mode:\n"
"\t(a) No need to acquire mutexes during writes (less cache "
"flushes in multi-core leading to speed up)\n"
"\t(b) No long validation at the end (more speed up)\n"
"\t(c) Test snapshot and atomicity of batch writes");
DEFINE_bool(atomic_flush, false,
"If set, enables atomic flush in the options.\n");
DEFINE_bool(test_cf_consistency, false,
"If set, runs the stress test dedicated to verifying writes to "
"multiple column families are consistent. Setting this implies "
"`atomic_flush=true` is set true if `disable_wal=false`.\n");
DEFINE_int32(threads, 32, "Number of concurrent threads to run.");
DEFINE_int32(ttl, -1,
"Opens the db with this ttl value if this is not -1. "
"Carefully specify a large value such that verifications on "
"deleted values don't fail");
DEFINE_int32(value_size_mult, 8,
"Size of value will be this number times rand_int(1,3) bytes");
DEFINE_int32(compaction_readahead_size, 0, "Compaction readahead size");
DEFINE_bool(enable_pipelined_write, false, "Pipeline WAL/memtable writes");
DEFINE_bool(verify_before_write, false, "Verify before write");
DEFINE_bool(histogram, false, "Print histogram of operation timings");
DEFINE_bool(destroy_db_initially, true,
"Destroys the database dir before start if this is true");
DEFINE_bool(verbose, false, "Verbose");
DEFINE_bool(progress_reports, true,
"If true, db_stress will report number of finished operations");
DEFINE_uint64(db_write_buffer_size,
ROCKSDB_NAMESPACE::Options().db_write_buffer_size,
"Number of bytes to buffer in all memtables before compacting");
DEFINE_int32(
write_buffer_size,
static_cast<int32_t>(ROCKSDB_NAMESPACE::Options().write_buffer_size),
"Number of bytes to buffer in memtable before compacting");
DEFINE_int32(max_write_buffer_number,
ROCKSDB_NAMESPACE::Options().max_write_buffer_number,
"The number of in-memory memtables. "
"Each memtable is of size FLAGS_write_buffer_size.");
DEFINE_int32(min_write_buffer_number_to_merge,
ROCKSDB_NAMESPACE::Options().min_write_buffer_number_to_merge,
"The minimum number of write buffers that will be merged together "
"before writing to storage. This is cheap because it is an "
"in-memory merge. If this feature is not enabled, then all these "
"write buffers are flushed to L0 as separate files and this "
"increases read amplification because a get request has to check "
"in all of these files. Also, an in-memory merge may result in "
"writing less data to storage if there are duplicate records in"
" each of these individual write buffers.");
DEFINE_int32(max_write_buffer_number_to_maintain,
ROCKSDB_NAMESPACE::Options().max_write_buffer_number_to_maintain,
"The total maximum number of write buffers to maintain in memory "
"including copies of buffers that have already been flushed. "
"Unlike max_write_buffer_number, this parameter does not affect "
"flushing. This controls the minimum amount of write history "
"that will be available in memory for conflict checking when "
"Transactions are used. If this value is too low, some "
"transactions may fail at commit time due to not being able to "
"determine whether there were any write conflicts. Setting this "
"value to 0 will cause write buffers to be freed immediately "
"after they are flushed. If this value is set to -1, "
"'max_write_buffer_number' will be used.");
DEFINE_int64(max_write_buffer_size_to_maintain,
ROCKSDB_NAMESPACE::Options().max_write_buffer_size_to_maintain,
"The total maximum size of write buffers to maintain in memory "
"including copies of buffers that have already been flushed. "
"Unlike max_write_buffer_number, this parameter does not affect "
"flushing. This controls the minimum amount of write history "
"that will be available in memory for conflict checking when "
"Transactions are used. If this value is too low, some "
"transactions may fail at commit time due to not being able to "
"determine whether there were any write conflicts. Setting this "
"value to 0 will cause write buffers to be freed immediately "
"after they are flushed. If this value is set to -1, "
"'max_write_buffer_number' will be used.");
DEFINE_double(memtable_prefix_bloom_size_ratio,
ROCKSDB_NAMESPACE::Options().memtable_prefix_bloom_size_ratio,
"creates prefix blooms for memtables, each with size "
"`write_buffer_size * memtable_prefix_bloom_size_ratio`.");
DEFINE_bool(memtable_whole_key_filtering,
ROCKSDB_NAMESPACE::Options().memtable_whole_key_filtering,
"Enable whole key filtering in memtables.");
DEFINE_int32(open_files, ROCKSDB_NAMESPACE::Options().max_open_files,
"Maximum number of files to keep open at the same time "
"(use default if == 0)");
DEFINE_int64(compressed_cache_size, -1,
"Number of bytes to use as a cache of compressed data."
" Negative means use default settings.");
DEFINE_int32(compaction_style, ROCKSDB_NAMESPACE::Options().compaction_style,
"");
DEFINE_int32(num_levels, ROCKSDB_NAMESPACE::Options().num_levels,
"Number of levels in the DB");
DEFINE_int32(level0_file_num_compaction_trigger,
ROCKSDB_NAMESPACE::Options().level0_file_num_compaction_trigger,
"Level0 compaction start trigger");
DEFINE_int32(level0_slowdown_writes_trigger,
ROCKSDB_NAMESPACE::Options().level0_slowdown_writes_trigger,
"Number of files in level-0 that will slow down writes");
DEFINE_int32(level0_stop_writes_trigger,
ROCKSDB_NAMESPACE::Options().level0_stop_writes_trigger,
"Number of files in level-0 that will trigger put stop.");
DEFINE_int32(block_size,
static_cast<int32_t>(
ROCKSDB_NAMESPACE::BlockBasedTableOptions().block_size),
"Number of bytes in a block.");
DEFINE_int32(format_version,
static_cast<int32_t>(
ROCKSDB_NAMESPACE::BlockBasedTableOptions().format_version),
"Format version of SST files.");
DEFINE_int32(
index_block_restart_interval,
ROCKSDB_NAMESPACE::BlockBasedTableOptions().index_block_restart_interval,
"Number of keys between restart points "
"for delta encoding of keys in index block.");
DEFINE_int32(max_background_compactions,
ROCKSDB_NAMESPACE::Options().max_background_compactions,
"The maximum number of concurrent background compactions "
"that can occur in parallel.");
DEFINE_int32(num_bottom_pri_threads, 0,
"The number of threads in the bottom-priority thread pool (used "
"by universal compaction only).");
DEFINE_int32(compaction_thread_pool_adjust_interval, 0,
"The interval (in milliseconds) to adjust compaction thread pool "
"size. Don't change it periodically if the value is 0.");
DEFINE_int32(compaction_thread_pool_variations, 2,
"Range of background thread pool size variations when adjusted "
"periodically.");
DEFINE_int32(max_background_flushes,
ROCKSDB_NAMESPACE::Options().max_background_flushes,
"The maximum number of concurrent background flushes "
"that can occur in parallel.");
DEFINE_int32(universal_size_ratio, 0,
"The ratio of file sizes that trigger"
" compaction in universal style");
DEFINE_int32(universal_min_merge_width, 0,
"The minimum number of files to "
"compact in universal style compaction");
DEFINE_int32(universal_max_merge_width, 0,
"The max number of files to compact"
" in universal style compaction");
DEFINE_int32(universal_max_size_amplification_percent, 0,
"The max size amplification for universal style compaction");
DEFINE_int32(clear_column_family_one_in, 1000000,
"With a chance of 1/N, delete a column family and then recreate "
"it again. If N == 0, never drop/create column families. "
"When test_batches_snapshots is true, this flag has no effect");
DEFINE_int32(get_live_files_one_in, 1000000,
"With a chance of 1/N, call GetLiveFiles to verify if it returns "
"correctly. If N == 0, do not call the interface.");
DEFINE_int32(
get_sorted_wal_files_one_in, 1000000,
"With a chance of 1/N, call GetSortedWalFiles to verify if it returns "
"correctly. (Note that this API may legitimately return an error.) If N == "
"0, do not call the interface.");
DEFINE_int32(
get_current_wal_file_one_in, 1000000,
"With a chance of 1/N, call GetCurrentWalFile to verify if it returns "
"correctly. (Note that this API may legitimately return an error.) If N == "
"0, do not call the interface.");
DEFINE_int32(set_options_one_in, 0,
"With a chance of 1/N, change some random options");
DEFINE_int32(set_in_place_one_in, 0,
"With a chance of 1/N, toggle in place support option");
DEFINE_int64(cache_size, 2LL * KB * KB * KB,
"Number of bytes to use as a cache of uncompressed data.");
DEFINE_bool(cache_index_and_filter_blocks, false,
"True if indexes/filters should be cached in block cache.");
DEFINE_int32(
top_level_index_pinning,
static_cast<int32_t>(ROCKSDB_NAMESPACE::PinningTier::kFallback),
"Type of pinning for top-level indexes into metadata partitions (see "
"`enum PinningTier` in table.h)");
DEFINE_int32(
partition_pinning,
static_cast<int32_t>(ROCKSDB_NAMESPACE::PinningTier::kFallback),
"Type of pinning for metadata partitions (see `enum PinningTier` in "
"table.h)");
DEFINE_int32(
unpartitioned_pinning,
static_cast<int32_t>(ROCKSDB_NAMESPACE::PinningTier::kFallback),
"Type of pinning for unpartitioned metadata blocks (see `enum PinningTier` "
"in table.h)");
DEFINE_bool(use_clock_cache, false,
"Replace default LRU block cache with clock cache.");
DEFINE_uint64(subcompactions, 1,
"Maximum number of subcompactions to divide L0-L1 compactions "
"into.");
DEFINE_uint64(periodic_compaction_seconds, 1000,
"Files older than this value will be picked up for compaction.");
DEFINE_uint64(compaction_ttl, 1000,
"Files older than TTL will be compacted to the next level.");
DEFINE_bool(allow_concurrent_memtable_write, false,
"Allow multi-writers to update mem tables in parallel.");
DEFINE_bool(enable_write_thread_adaptive_yield, true,
"Use a yielding spin loop for brief writer thread waits.");
#ifndef ROCKSDB_LITE
// Options for StackableDB-based BlobDB
DEFINE_bool(use_blob_db, false, "[Stacked BlobDB] Use BlobDB.");
DEFINE_uint64(
blob_db_min_blob_size,
ROCKSDB_NAMESPACE::blob_db::BlobDBOptions().min_blob_size,
"[Stacked BlobDB] Smallest blob to store in a file. Blobs "
"smaller than this will be inlined with the key in the LSM tree.");
DEFINE_uint64(
blob_db_bytes_per_sync,
ROCKSDB_NAMESPACE::blob_db::BlobDBOptions().bytes_per_sync,
"[Stacked BlobDB] Sync blob files once per every N bytes written.");
DEFINE_uint64(blob_db_file_size,
ROCKSDB_NAMESPACE::blob_db::BlobDBOptions().blob_file_size,
"[Stacked BlobDB] Target size of each blob file.");
DEFINE_bool(
blob_db_enable_gc,
ROCKSDB_NAMESPACE::blob_db::BlobDBOptions().enable_garbage_collection,
"[Stacked BlobDB] Enable BlobDB garbage collection.");
DEFINE_double(
blob_db_gc_cutoff,
ROCKSDB_NAMESPACE::blob_db::BlobDBOptions().garbage_collection_cutoff,
"[Stacked BlobDB] Cutoff ratio for BlobDB garbage collection.");
#endif // !ROCKSDB_LITE
// Options for integrated BlobDB
DEFINE_bool(allow_setting_blob_options_dynamically, false,
"[Integrated BlobDB] Allow setting blob options dynamically.");
DEFINE_bool(
enable_blob_files,
ROCKSDB_NAMESPACE::AdvancedColumnFamilyOptions().enable_blob_files,
"[Integrated BlobDB] Enable writing large values to separate blob files.");
DEFINE_uint64(min_blob_size,
ROCKSDB_NAMESPACE::AdvancedColumnFamilyOptions().min_blob_size,
"[Integrated BlobDB] The size of the smallest value to be stored "
"separately in a blob file.");
DEFINE_uint64(blob_file_size,
ROCKSDB_NAMESPACE::AdvancedColumnFamilyOptions().blob_file_size,
"[Integrated BlobDB] The size limit for blob files.");
DEFINE_string(blob_compression_type, "none",
"[Integrated BlobDB] The compression algorithm to use for large "
"values stored in blob files.");
DEFINE_bool(enable_blob_garbage_collection,
ROCKSDB_NAMESPACE::AdvancedColumnFamilyOptions()
.enable_blob_garbage_collection,
"[Integrated BlobDB] Enable blob garbage collection.");
DEFINE_double(blob_garbage_collection_age_cutoff,
ROCKSDB_NAMESPACE::AdvancedColumnFamilyOptions()
.blob_garbage_collection_age_cutoff,
"[Integrated BlobDB] The cutoff in terms of blob file age for "
"garbage collection.");
static const bool FLAGS_subcompactions_dummy __attribute__((__unused__)) =
RegisterFlagValidator(&FLAGS_subcompactions, &ValidateUint32Range);
static bool ValidateInt32Positive(const char* flagname, int32_t value) {
if (value < 0) {
fprintf(stderr, "Invalid value for --%s: %d, must be >=0\n", flagname,
value);
return false;
}
return true;
}
DEFINE_int32(reopen, 10, "Number of times database reopens");
static const bool FLAGS_reopen_dummy __attribute__((__unused__)) =
RegisterFlagValidator(&FLAGS_reopen, &ValidateInt32Positive);
DEFINE_double(bloom_bits, 10,
"Bloom filter bits per key. "
"Negative means use default settings.");
DEFINE_bool(use_block_based_filter, false,
"use block based filter"
"instead of full filter for block based table");
Experimental (production candidate) SST schema for Ribbon filter (#7658) Summary: Added experimental public API for Ribbon filter: NewExperimentalRibbonFilterPolicy(). This experimental API will take a "Bloom equivalent" bits per key, and configure the Ribbon filter for the same FP rate as Bloom would have but ~30% space savings. (Note: optimize_filters_for_memory is not yet implemented for Ribbon filter. That can be added with no effect on schema.) Internally, the Ribbon filter is configured using a "one_in_fp_rate" value, which is 1 over desired FP rate. For example, use 100 for 1% FP rate. I'm expecting this will be used in the future for configuring Bloom-like filters, as I expect people to more commonly hold constant the filter accuracy and change the space vs. time trade-off, rather than hold constant the space (per key) and change the accuracy vs. time trade-off, though we might make that available. ### Benchmarking ``` $ ./filter_bench -impl=2 -quick -m_keys_total_max=200 -average_keys_per_filter=100000 -net_includes_hashing Building... Build avg ns/key: 34.1341 Number of filters: 1993 Total size (MB): 238.488 Reported total allocated memory (MB): 262.875 Reported internal fragmentation: 10.2255% Bits/key stored: 10.0029 ---------------------------- Mixed inside/outside queries... Single filter net ns/op: 18.7508 Random filter net ns/op: 258.246 Average FP rate %: 0.968672 ---------------------------- Done. (For more info, run with -legend or -help.) $ ./filter_bench -impl=3 -quick -m_keys_total_max=200 -average_keys_per_filter=100000 -net_includes_hashing Building... Build avg ns/key: 130.851 Number of filters: 1993 Total size (MB): 168.166 Reported total allocated memory (MB): 183.211 Reported internal fragmentation: 8.94626% Bits/key stored: 7.05341 ---------------------------- Mixed inside/outside queries... Single filter net ns/op: 58.4523 Random filter net ns/op: 363.717 Average FP rate %: 0.952978 ---------------------------- Done. (For more info, run with -legend or -help.) ``` 168.166 / 238.488 = 0.705 -> 29.5% space reduction 130.851 / 34.1341 = 3.83x construction time for this Ribbon filter vs. lastest Bloom filter (could make that as little as about 2.5x for less space reduction) ### Working around a hashing "flaw" bloom_test discovered a flaw in the simple hashing applied in StandardHasher when num_starts == 1 (num_slots == 128), showing an excessively high FP rate. The problem is that when many entries, on the order of number of hash bits or kCoeffBits, are associated with the same start location, the correlation between the CoeffRow and ResultRow (for efficiency) can lead to a solution that is "universal," or nearly so, for entries mapping to that start location. (Normally, variance in start location breaks the effective association between CoeffRow and ResultRow; the same value for CoeffRow is effectively different if start locations are different.) Without kUseSmash and with num_starts > 1 (thus num_starts ~= num_slots), this flaw should be completely irrelevant. Even with 10M slots, the chances of a single slot having just 16 (or more) entries map to it--not enough to cause an FP problem, which would be local to that slot if it happened--is 1 in millions. This spreadsheet formula shows that: =1/(10000000*(1 - POISSON(15, 1, TRUE))) As kUseSmash==false (the setting for Standard128RibbonBitsBuilder) is intended for CPU efficiency of filters with many more entries/slots than kCoeffBits, a very reasonable work-around is to disallow num_starts==1 when !kUseSmash, by making the minimum non-zero number of slots 2*kCoeffBits. This is the work-around I've applied. This also means that the new Ribbon filter schema (Standard128RibbonBitsBuilder) is not space-efficient for less than a few hundred entries. Because of this, I have made it fall back on constructing a Bloom filter, under existing schema, when that is more space efficient for small filters. (We can change this in the future if we want.) TODO: better unit tests for this case in ribbon_test, and probably update StandardHasher for kUseSmash case so that it can scale nicely to small filters. ### Other related changes * Add Ribbon filter to stress/crash test * Add Ribbon filter to filter_bench as -impl=3 * Add option string support, as in "filter_policy=experimental_ribbon:5.678;" where 5.678 is the Bloom equivalent bits per key. * Rename internal mode BloomFilterPolicy::kAuto to kAutoBloom * Add a general BuiltinFilterBitsBuilder::CalculateNumEntry based on binary searching CalculateSpace (inefficient), so that subclasses (especially experimental ones) don't have to provide an efficient implementation inverting CalculateSpace. * Minor refactor FastLocalBloomBitsBuilder for new base class XXH3pFilterBitsBuilder shared with new Standard128RibbonBitsBuilder, which allows the latter to fall back on Bloom construction in some extreme cases. * Mostly updated bloom_test for Ribbon filter, though a test like FullBloomTest::Schema is a next TODO to ensure schema stability (in case this becomes production-ready schema as it is). * Add some APIs to ribbon_impl.h for configuring Ribbon filters. Although these are reasonably covered by bloom_test, TODO more unit tests in ribbon_test * Added a "tool" FindOccupancyForSuccessRate to ribbon_test to get data for constructing the linear approximations in GetNumSlotsFor95PctSuccess. Pull Request resolved: https://github.com/facebook/rocksdb/pull/7658 Test Plan: Some unit tests updated but other testing is left TODO. This is considered experimental but laying down schema compatibility as early as possible in case it proves production-quality. Also tested in stress/crash test. Reviewed By: jay-zhuang Differential Revision: D24899349 Pulled By: pdillinger fbshipit-source-id: 9715f3e6371c959d923aea8077c9423c7a9f82b8
2020-11-13 04:45:02 +00:00
DEFINE_bool(use_ribbon_filter, false,
"Use Ribbon filter instead of Bloom filter");
DEFINE_bool(partition_filters, false,
"use partitioned filters "
"for block-based table");
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 20:30:57 +00:00
DEFINE_bool(
optimize_filters_for_memory,
ROCKSDB_NAMESPACE::BlockBasedTableOptions().optimize_filters_for_memory,
"Minimize memory footprint of filters");
DEFINE_int32(
index_type,
static_cast<int32_t>(
ROCKSDB_NAMESPACE::BlockBasedTableOptions::kBinarySearch),
"Type of block-based table index (see `enum IndexType` in table.h)");
DEFINE_string(db, "", "Use the db with the following name.");
DEFINE_string(secondaries_base, "",
"Use this path as the base path for secondary instances.");
DEFINE_bool(test_secondary, false, "Test secondary instance.");
DEFINE_string(
expected_values_path, "",
"File where the array of expected uint32_t values will be stored. If "
"provided and non-empty, the DB state will be verified against these "
"values after recovery. --max_key and --column_family must be kept the "
"same across invocations of this program that use the same "
"--expected_values_path.");
DEFINE_bool(verify_checksum, false,
"Verify checksum for every block read from storage");
DEFINE_bool(mmap_read, ROCKSDB_NAMESPACE::Options().allow_mmap_reads,
"Allow reads to occur via mmap-ing files");
DEFINE_bool(mmap_write, ROCKSDB_NAMESPACE::Options().allow_mmap_writes,
"Allow writes to occur via mmap-ing files");
DEFINE_bool(use_direct_reads, ROCKSDB_NAMESPACE::Options().use_direct_reads,
"Use O_DIRECT for reading data");
DEFINE_bool(use_direct_io_for_flush_and_compaction,
ROCKSDB_NAMESPACE::Options().use_direct_io_for_flush_and_compaction,
"Use O_DIRECT for writing data");
DEFINE_bool(mock_direct_io, false,
"Mock direct IO by not using O_DIRECT for direct IO read");
DEFINE_bool(statistics, false, "Create database statistics");
DEFINE_bool(sync, false, "Sync all writes to disk");
DEFINE_bool(use_fsync, false, "If true, issue fsync instead of fdatasync");
DEFINE_int32(kill_random_test, 0,
"If non-zero, kill at various points in source code with "
"probability 1/this");
static const bool FLAGS_kill_random_test_dummy __attribute__((__unused__)) =
RegisterFlagValidator(&FLAGS_kill_random_test, &ValidateInt32Positive);
extern int rocksdb_kill_odds;
DEFINE_string(kill_exclude_prefixes, "",
"If non-empty, kill points with prefix in the list given will be"
" skipped. Items are comma-separated.");
extern std::vector<std::string> rocksdb_kill_exclude_prefixes;
DEFINE_bool(disable_wal, false, "If true, do not write WAL for write.");
DEFINE_uint64(recycle_log_file_num,
ROCKSDB_NAMESPACE::Options().recycle_log_file_num,
"Number of old WAL files to keep around for later recycling");
DEFINE_int64(target_file_size_base,
ROCKSDB_NAMESPACE::Options().target_file_size_base,
"Target level-1 file size for compaction");
DEFINE_int32(target_file_size_multiplier, 1,
"A multiplier to compute target level-N file size (N >= 2)");
DEFINE_uint64(max_bytes_for_level_base,
ROCKSDB_NAMESPACE::Options().max_bytes_for_level_base,
"Max bytes for level-1");
DEFINE_double(max_bytes_for_level_multiplier, 2,
"A multiplier to compute max bytes for level-N (N >= 2)");
DEFINE_int32(range_deletion_width, 10,
"The width of the range deletion intervals.");
DEFINE_uint64(rate_limiter_bytes_per_sec, 0, "Set options.rate_limiter value.");
DEFINE_bool(rate_limit_bg_reads, false,
"Use options.rate_limiter on compaction reads");
DEFINE_uint64(sst_file_manager_bytes_per_sec, 0,
"Set `Options::sst_file_manager` to delete at this rate. By "
"default the deletion rate is unbounded.");
DEFINE_uint64(sst_file_manager_bytes_per_truncate, 0,
"Set `Options::sst_file_manager` to delete in chunks of this "
"many bytes. By default whole files will be deleted.");
DEFINE_bool(use_txn, false,
"Use TransactionDB. Currently the default write policy is "
"TxnDBWritePolicy::WRITE_PREPARED");
DEFINE_uint64(txn_write_policy, 0,
"The transaction write policy. Default is "
"TxnDBWritePolicy::WRITE_COMMITTED. Note that this should not be "
"changed accross crashes.");
DEFINE_bool(unordered_write, false,
"Turn on the unordered_write feature. This options is currently "
"tested only in combination with use_txn=true and "
"txn_write_policy=TxnDBWritePolicy::WRITE_PREPARED.");
DEFINE_int32(backup_one_in, 0,
"If non-zero, then CreateNewBackup() will be called once for "
"every N operations on average. 0 indicates CreateNewBackup() "
"is disabled.");
DEFINE_uint64(backup_max_size, 100 * 1024 * 1024,
"If non-zero, skip checking backup/restore when DB size in "
"bytes exceeds this setting.");
DEFINE_int32(checkpoint_one_in, 0,
"If non-zero, then CreateCheckpoint() will be called once for "
"every N operations on average. 0 indicates CreateCheckpoint() "
"is disabled.");
DEFINE_int32(ingest_external_file_one_in, 0,
"If non-zero, then IngestExternalFile() will be called once for "
"every N operations on average. 0 indicates IngestExternalFile() "
"is disabled.");
DEFINE_int32(ingest_external_file_width, 1000,
"The width of the ingested external files.");
DEFINE_int32(compact_files_one_in, 0,
"If non-zero, then CompactFiles() will be called once for every N "
"operations on average. 0 indicates CompactFiles() is disabled.");
DEFINE_int32(compact_range_one_in, 0,
"If non-zero, then CompactRange() will be called once for every N "
"operations on average. 0 indicates CompactRange() is disabled.");
DEFINE_int32(mark_for_compaction_one_file_in, 0,
"A `TablePropertiesCollectorFactory` will be registered, which "
"creates a `TablePropertiesCollector` with `NeedCompact()` "
"returning true once for every N files on average. 0 or negative "
"mean `NeedCompact()` always returns false.");
DEFINE_int32(flush_one_in, 0,
"If non-zero, then Flush() will be called once for every N ops "
"on average. 0 indicates calls to Flush() are disabled.");
DEFINE_int32(pause_background_one_in, 0,
"If non-zero, then PauseBackgroundWork()+Continue will be called "
"once for every N ops on average. 0 disables.");
DEFINE_int32(compact_range_width, 10000,
"The width of the ranges passed to CompactRange().");
DEFINE_int32(acquire_snapshot_one_in, 0,
"If non-zero, then acquires a snapshot once every N operations on "
"average.");
DEFINE_bool(compare_full_db_state_snapshot, false,
"If set we compare state of entire db (in one of the threads) with"
"each snapshot.");
DEFINE_uint64(snapshot_hold_ops, 0,
"If non-zero, then releases snapshots N operations after they're "
"acquired.");
DEFINE_bool(long_running_snapshots, false,
"If set, hold on some some snapshots for much longer time.");
DEFINE_bool(use_multiget, false,
"If set, use the batched MultiGet API for reads");
static bool ValidateInt32Percent(const char* flagname, int32_t value) {
if (value < 0 || value > 100) {
fprintf(stderr, "Invalid value for --%s: %d, 0<= pct <=100 \n", flagname,
value);
return false;
}
return true;
}
DEFINE_int32(readpercent, 10,
"Ratio of reads to total workload (expressed as a percentage)");
static const bool FLAGS_readpercent_dummy __attribute__((__unused__)) =
RegisterFlagValidator(&FLAGS_readpercent, &ValidateInt32Percent);
DEFINE_int32(prefixpercent, 20,
"Ratio of prefix iterators to total workload (expressed as a"
" percentage)");
static const bool FLAGS_prefixpercent_dummy __attribute__((__unused__)) =
RegisterFlagValidator(&FLAGS_prefixpercent, &ValidateInt32Percent);
DEFINE_int32(writepercent, 45,
"Ratio of writes to total workload (expressed as a percentage)");
static const bool FLAGS_writepercent_dummy __attribute__((__unused__)) =
RegisterFlagValidator(&FLAGS_writepercent, &ValidateInt32Percent);
DEFINE_int32(delpercent, 15,
"Ratio of deletes to total workload (expressed as a percentage)");
static const bool FLAGS_delpercent_dummy __attribute__((__unused__)) =
RegisterFlagValidator(&FLAGS_delpercent, &ValidateInt32Percent);
DEFINE_int32(delrangepercent, 0,
"Ratio of range deletions to total workload (expressed as a "
"percentage). Cannot be used with test_batches_snapshots");
static const bool FLAGS_delrangepercent_dummy __attribute__((__unused__)) =
RegisterFlagValidator(&FLAGS_delrangepercent, &ValidateInt32Percent);
DEFINE_int32(nooverwritepercent, 60,
"Ratio of keys without overwrite to total workload (expressed as "
" a percentage)");
static const bool FLAGS_nooverwritepercent_dummy __attribute__((__unused__)) =
RegisterFlagValidator(&FLAGS_nooverwritepercent, &ValidateInt32Percent);
DEFINE_int32(iterpercent, 10,
"Ratio of iterations to total workload"
" (expressed as a percentage)");
static const bool FLAGS_iterpercent_dummy __attribute__((__unused__)) =
RegisterFlagValidator(&FLAGS_iterpercent, &ValidateInt32Percent);
DEFINE_uint64(num_iterations, 10, "Number of iterations per MultiIterate run");
static const bool FLAGS_num_iterations_dummy __attribute__((__unused__)) =
RegisterFlagValidator(&FLAGS_num_iterations, &ValidateUint32Range);
DEFINE_string(compression_type, "snappy",
"Algorithm to use to compress the database");
DEFINE_int32(compression_max_dict_bytes, 0,
"Maximum size of dictionary used to prime the compression "
"library.");
DEFINE_int32(compression_zstd_max_train_bytes, 0,
"Maximum size of training data passed to zstd's dictionary "
"trainer.");
DEFINE_int32(compression_parallel_threads, 1,
"Number of threads for parallel compression.");
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 22:06:59 +00:00
DEFINE_uint64(compression_max_dict_buffer_bytes, 0,
"Buffering limit for SST file data to sample for dictionary "
"compression.");
DEFINE_string(bottommost_compression_type, "disable",
"Algorithm to use to compress bottommost level of the database. "
"\"disable\" means disabling the feature");
DEFINE_string(checksum_type, "kCRC32c", "Algorithm to use to checksum blocks");
DEFINE_string(hdfs, "",
"Name of hdfs environment. Mutually exclusive with"
" --env_uri and --fs_uri.");
DEFINE_string(
env_uri, "",
"URI for env lookup. Mutually exclusive with --hdfs and --fs_uri");
DEFINE_string(fs_uri, "",
"URI for registry Filesystem lookup. Mutually exclusive"
" with --hdfs and --env_uri."
" Creates a default environment with the specified filesystem.");
DEFINE_uint64(ops_per_thread, 1200000, "Number of operations per thread.");
static const bool FLAGS_ops_per_thread_dummy __attribute__((__unused__)) =
RegisterFlagValidator(&FLAGS_ops_per_thread, &ValidateUint32Range);
DEFINE_uint64(log2_keys_per_lock, 2, "Log2 of number of keys per lock");
static const bool FLAGS_log2_keys_per_lock_dummy __attribute__((__unused__)) =
RegisterFlagValidator(&FLAGS_log2_keys_per_lock, &ValidateUint32Range);
DEFINE_uint64(max_manifest_file_size, 16384, "Maximum size of a MANIFEST file");
DEFINE_bool(in_place_update, false, "On true, does inplace update in memtable");
DEFINE_int32(secondary_catch_up_one_in, 0,
"If non-zero, the secondaries attemp to catch up with the primary "
"once for every N operations on average. 0 indicates the "
"secondaries do not try to catch up after open.");
DEFINE_string(memtablerep, "skip_list", "");
inline static bool ValidatePrefixSize(const char* flagname, int32_t value) {
if (value < -1 || value > 8) {
fprintf(stderr, "Invalid value for --%s: %d. -1 <= PrefixSize <= 8\n",
flagname, value);
return false;
}
return true;
}
DEFINE_int32(prefix_size, 7,
"Control the prefix size for HashSkipListRep. "
"-1 is disabled.");
static const bool FLAGS_prefix_size_dummy __attribute__((__unused__)) =
RegisterFlagValidator(&FLAGS_prefix_size, &ValidatePrefixSize);
DEFINE_bool(use_merge, false,
"On true, replaces all writes with a Merge "
"that behaves like a Put");
DEFINE_bool(use_full_merge_v1, false,
"On true, use a merge operator that implement the deprecated "
"version of FullMerge");
DEFINE_int32(sync_wal_one_in, 0,
"If non-zero, then SyncWAL() will be called once for every N ops "
"on average. 0 indicates that calls to SyncWAL() are disabled.");
DEFINE_bool(avoid_unnecessary_blocking_io,
ROCKSDB_NAMESPACE::Options().avoid_unnecessary_blocking_io,
"If true, some expensive cleaning up operations will be moved from "
"user reads to high-pri background threads.");
DEFINE_bool(write_dbid_to_manifest,
ROCKSDB_NAMESPACE::Options().write_dbid_to_manifest,
"Write DB_ID to manifest");
DEFINE_bool(avoid_flush_during_recovery,
ROCKSDB_NAMESPACE::Options().avoid_flush_during_recovery,
"Avoid flush during recovery");
DEFINE_uint64(max_write_batch_group_size_bytes,
ROCKSDB_NAMESPACE::Options().max_write_batch_group_size_bytes,
"Max write batch group size");
DEFINE_bool(level_compaction_dynamic_level_bytes,
ROCKSDB_NAMESPACE::Options().level_compaction_dynamic_level_bytes,
"Use dynamic level");
DEFINE_int32(verify_checksum_one_in, 0,
"If non-zero, then DB::VerifyChecksum() will be called to do"
" checksum verification of all the files in the database once for"
" every N ops on average. 0 indicates that calls to"
" VerifyChecksum() are disabled.");
DEFINE_int32(verify_db_one_in, 0,
"If non-zero, call VerifyDb() once for every N ops. 0 indicates "
"that VerifyDb() will not be called in OperateDb(). Note that "
"enabling this can slow down tests.");
DEFINE_int32(continuous_verification_interval, 1000,
"While test is running, verify db every N milliseconds. 0 "
"disables continuous verification.");
DEFINE_int32(approximate_size_one_in, 64,
"If non-zero, DB::GetApproximateSizes() will be called against"
" random key ranges.");
DEFINE_int32(read_fault_one_in, 1000,
"On non-zero, enables fault injection on read");
DEFINE_int32(get_property_one_in, 1000,
"If non-zero, then DB::GetProperty() will be called to get various"
" properties for every N ops on average. 0 indicates that"
" GetProperty() will be not be called.");
DEFINE_bool(sync_fault_injection, false,
"If true, FaultInjectionTestFS will be used for write operations, "
" and unsynced data in DB will lost after crash.");
DEFINE_bool(best_efforts_recovery, false,
"If true, use best efforts recovery.");
DEFINE_bool(skip_verifydb, false, "If true, skip VerifyDb() calls.");
DEFINE_bool(enable_compaction_filter, false,
"If true, configures a compaction filter that returns a kRemove "
"decision for deleted keys.");
DEFINE_bool(paranoid_file_checks, true,
"After writing every SST file, reopen it and read all the keys "
"and validate checksums");
Integrity protection for live updates to WriteBatch (#7748) Summary: This PR adds the foundation classes for key-value integrity protection and the first use case: protecting live updates from the source buffers added to `WriteBatch` through the destination buffer in `MemTable`. The width of the protection info is not yet configurable -- only eight bytes per key is supported. This PR allows users to enable protection by constructing `WriteBatch` with `protection_bytes_per_key == 8`. It does not yet expose a way for users to get integrity protection via other write APIs (e.g., `Put()`, `Merge()`, `Delete()`, etc.). The foundation classes (`ProtectionInfo.*`) embed the coverage info in their type, and provide `Protect.*()` and `Strip.*()` functions to navigate between types with different coverage. For making bytes per key configurable (for powers of two up to eight) in the future, these classes are templated on the unsigned integer type used to store the protection info. That integer contains the XOR'd result of hashes with independent seeds for all covered fields. For integer fields, the hash is computed on the raw unadjusted bytes, so the result is endian-dependent. The most significant bytes are truncated when the hash value (8 bytes) is wider than the protection integer. When `WriteBatch` is constructed with `protection_bytes_per_key == 8`, we hold a `ProtectionInfoKVOTC` (i.e., one that covers key, value, optype aka `ValueType`, timestamp, and CF ID) for each entry added to the batch. The protection info is generated from the original buffers passed by the user, as well as the original metadata generated internally. When writing to memtable, each entry is transformed to a `ProtectionInfoKVOTS` (i.e., dropping coverage of CF ID and adding coverage of sequence number), since at that point we know the sequence number, and have already selected a memtable corresponding to a particular CF. This protection info is verified once the entry is encoded in the `MemTable` buffer. Pull Request resolved: https://github.com/facebook/rocksdb/pull/7748 Test Plan: - an integration test to verify a wide variety of single-byte changes to the encoded `MemTable` buffer are caught - add to stress/crash test to verify it works in variety of configs/operations without intentional corruption - [deferred] unit tests for `ProtectionInfo.*` classes for edge cases like KV swap, `SliceParts` and `Slice` APIs are interchangeable, etc. Reviewed By: pdillinger Differential Revision: D25754492 Pulled By: ajkr fbshipit-source-id: e481bac6c03c2ab268be41359730f1ceb9964866
2021-01-29 20:17:17 +00:00
DEFINE_uint64(batch_protection_bytes_per_key, 0,
"If nonzero, enables integrity protection in `WriteBatch` at the "
"specified number of bytes per key. Currently the only supported "
"nonzero value is eight.");
DEFINE_string(file_checksum_impl, "none",
"Name of an implementation for file_checksum_gen_factory, or "
"\"none\" for null.");
DEFINE_int32(write_fault_one_in, 0,
"On non-zero, enables fault injection on write");
DEFINE_uint64(user_timestamp_size, 0,
"Number of bytes for a user-defined timestamp. Currently, only "
"8-byte is supported");
#endif // GFLAGS