2016-02-09 23:12:00 +00:00
|
|
|
// Copyright (c) 2011-present, Facebook, Inc. All rights reserved.
|
2013-10-29 00:54:09 +00:00
|
|
|
// This source code is licensed under the BSD-style license found in the
|
|
|
|
// LICENSE file in the root directory of this source tree. An additional grant
|
|
|
|
// of patent rights can be found in the PATENTS file in the same directory.
|
|
|
|
//
|
|
|
|
// Copyright (c) 2011 The LevelDB Authors. All rights reserved.
|
|
|
|
// Use of this source code is governed by a BSD-style license that can be
|
|
|
|
// found in the LICENSE file. See the AUTHORS file for names of contributors.
|
|
|
|
|
|
|
|
|
|
|
|
#include "table/block_based_table_factory.h"
|
|
|
|
|
|
|
|
#include <memory>
|
2014-03-01 02:19:07 +00:00
|
|
|
#include <string>
|
2013-10-29 00:54:09 +00:00
|
|
|
#include <stdint.h>
|
2014-03-01 02:19:07 +00:00
|
|
|
|
2015-01-15 00:24:24 +00:00
|
|
|
#include "port/port.h"
|
2014-03-01 02:19:07 +00:00
|
|
|
#include "rocksdb/flush_block_policy.h"
|
2014-08-25 21:22:05 +00:00
|
|
|
#include "rocksdb/cache.h"
|
2013-10-29 00:54:09 +00:00
|
|
|
#include "table/block_based_table_builder.h"
|
2013-10-30 17:52:33 +00:00
|
|
|
#include "table/block_based_table_reader.h"
|
2015-01-15 00:24:24 +00:00
|
|
|
#include "table/format.h"
|
2013-10-29 00:54:09 +00:00
|
|
|
|
|
|
|
namespace rocksdb {
|
|
|
|
|
2014-03-01 00:39:27 +00:00
|
|
|
BlockBasedTableFactory::BlockBasedTableFactory(
|
2015-11-18 00:41:54 +00:00
|
|
|
const BlockBasedTableOptions& _table_options)
|
|
|
|
: table_options_(_table_options) {
|
2014-03-01 00:39:27 +00:00
|
|
|
if (table_options_.flush_block_policy_factory == nullptr) {
|
|
|
|
table_options_.flush_block_policy_factory.reset(
|
|
|
|
new FlushBlockBySizePolicyFactory());
|
|
|
|
}
|
2014-08-25 21:22:05 +00:00
|
|
|
if (table_options_.no_block_cache) {
|
|
|
|
table_options_.block_cache.reset();
|
|
|
|
} else if (table_options_.block_cache == nullptr) {
|
|
|
|
table_options_.block_cache = NewLRUCache(8 << 20);
|
|
|
|
}
|
|
|
|
if (table_options_.block_size_deviation < 0 ||
|
|
|
|
table_options_.block_size_deviation > 100) {
|
|
|
|
table_options_.block_size_deviation = 0;
|
|
|
|
}
|
2016-01-04 18:51:00 +00:00
|
|
|
if (table_options_.block_restart_interval < 1) {
|
|
|
|
table_options_.block_restart_interval = 1;
|
|
|
|
}
|
2016-02-05 18:22:37 +00:00
|
|
|
if (table_options_.index_block_restart_interval < 1) {
|
|
|
|
table_options_.index_block_restart_interval = 1;
|
|
|
|
}
|
2014-03-01 00:39:27 +00:00
|
|
|
}
|
|
|
|
|
2014-01-28 05:58:46 +00:00
|
|
|
Status BlockBasedTableFactory::NewTableReader(
|
2015-09-11 18:36:33 +00:00
|
|
|
const TableReaderOptions& table_reader_options,
|
|
|
|
unique_ptr<RandomAccessFileReader>&& file, uint64_t file_size,
|
|
|
|
unique_ptr<TableReader>* table_reader) const {
|
|
|
|
return NewTableReader(table_reader_options, std::move(file), file_size,
|
|
|
|
table_reader,
|
|
|
|
/*prefetch_index_and_filter=*/true);
|
|
|
|
}
|
|
|
|
|
|
|
|
Status BlockBasedTableFactory::NewTableReader(
|
|
|
|
const TableReaderOptions& table_reader_options,
|
Move rate_limiter, write buffering, most perf context instrumentation and most random kill out of Env
Summary: We want to keep Env a think layer for better portability. Less platform dependent codes should be moved out of Env. In this patch, I create a wrapper of file readers and writers, and put rate limiting, write buffering, as well as most perf context instrumentation and random kill out of Env. It will make it easier to maintain multiple Env in the future.
Test Plan: Run all existing unit tests.
Reviewers: anthony, kradhakrishnan, IslamAbdelRahman, yhchiang, igor
Reviewed By: igor
Subscribers: leveldb, dhruba
Differential Revision: https://reviews.facebook.net/D42321
2015-07-17 23:16:11 +00:00
|
|
|
unique_ptr<RandomAccessFileReader>&& file, uint64_t file_size,
|
2015-02-26 00:34:26 +00:00
|
|
|
unique_ptr<TableReader>* table_reader, const bool prefetch_enabled) const {
|
2015-09-11 18:36:33 +00:00
|
|
|
return BlockBasedTable::Open(
|
|
|
|
table_reader_options.ioptions, table_reader_options.env_options,
|
|
|
|
table_options_, table_reader_options.internal_comparator, std::move(file),
|
Skip bottom-level filter block caching when hit-optimized
Summary:
When Get() or NewIterator() trigger file loads, skip caching the filter block if
(1) optimize_filters_for_hits is set and (2) the file is on the bottommost
level. Also skip checking filters under the same conditions, which means that
for a preloaded file or a file that was trivially-moved to the bottom level, its
filter block will eventually expire from the cache.
- added parameters/instance variables in various places in order to propagate the config ("skip_filters") from version_set to block_based_table_reader
- in BlockBasedTable::Rep, this optimization prevents filter from being loaded when the file is opened simply by setting filter_policy = nullptr
- in BlockBasedTable::Get/BlockBasedTable::NewIterator, this optimization prevents filter from being used (even if it was loaded already) by setting filter = nullptr
Test Plan:
updated unit test:
$ ./db_test --gtest_filter=DBTest.OptimizeFiltersForHits
will also run 'make check'
Reviewers: sdong, igor, paultuckfield, anthony, rven, kradhakrishnan, IslamAbdelRahman, yhchiang
Reviewed By: yhchiang
Subscribers: leveldb
Differential Revision: https://reviews.facebook.net/D51633
2015-12-23 18:15:07 +00:00
|
|
|
file_size, table_reader, prefetch_enabled,
|
Adding pin_l0_filter_and_index_blocks_in_cache feature and related fixes.
Summary:
When a block based table file is opened, if prefetch_index_and_filter is true, it will prefetch the index and filter blocks, putting them into the block cache.
What this feature adds: when a L0 block based table file is opened, if pin_l0_filter_and_index_blocks_in_cache is true in the options (and prefetch_index_and_filter is true), then the filter and index blocks aren't released back to the block cache at the end of BlockBasedTableReader::Open(). Instead the table reader takes ownership of them, hence pinning them, ie. the LRU cache will never push them out. Meanwhile in the table reader, further accesses will not hit the block cache, thus avoiding lock contention.
Test Plan:
'export TEST_TMPDIR=/dev/shm/ && DISABLE_JEMALLOC=1 OPT=-g make all valgrind_check -j32' is OK.
I didn't run the Java tests, I don't have Java set up on my devserver.
Reviewers: sdong
Reviewed By: sdong
Subscribers: andrewkr, dhruba
Differential Revision: https://reviews.facebook.net/D56133
2016-04-01 17:42:39 +00:00
|
|
|
table_reader_options.skip_filters, table_reader_options.level);
|
2013-10-29 00:54:09 +00:00
|
|
|
}
|
|
|
|
|
2014-01-28 05:58:46 +00:00
|
|
|
TableBuilder* BlockBasedTableFactory::NewTableBuilder(
|
2015-10-08 23:57:35 +00:00
|
|
|
const TableBuilderOptions& table_builder_options, uint32_t column_family_id,
|
Move rate_limiter, write buffering, most perf context instrumentation and most random kill out of Env
Summary: We want to keep Env a think layer for better portability. Less platform dependent codes should be moved out of Env. In this patch, I create a wrapper of file readers and writers, and put rate limiting, write buffering, as well as most perf context instrumentation and random kill out of Env. It will make it easier to maintain multiple Env in the future.
Test Plan: Run all existing unit tests.
Reviewers: anthony, kradhakrishnan, IslamAbdelRahman, yhchiang, igor
Reviewed By: igor
Subscribers: leveldb, dhruba
Differential Revision: https://reviews.facebook.net/D42321
2015-07-17 23:16:11 +00:00
|
|
|
WritableFileWriter* file) const {
|
2014-03-01 02:19:07 +00:00
|
|
|
auto table_builder = new BlockBasedTableBuilder(
|
A new call back to TablePropertiesCollector to allow users know the entry is add, delete or merge
Summary:
Currently users have no idea a key is add, delete or merge from TablePropertiesCollector call back. Add a new function to add it.
Also refactor the codes so that
(1) make table property collector and internal table property collector two separate data structures with the later one now exposed
(2) table builders only receive internal table properties
Test Plan: Add cases in table_properties_collector_test to cover both of old and new ways of using TablePropertiesCollector.
Reviewers: yhchiang, igor.sugak, rven, igor
Reviewed By: rven, igor
Subscribers: meyering, yoshinorim, maykov, leveldb, dhruba
Differential Revision: https://reviews.facebook.net/D35373
2015-04-06 17:04:30 +00:00
|
|
|
table_builder_options.ioptions, table_options_,
|
|
|
|
table_builder_options.internal_comparator,
|
2015-10-08 23:57:35 +00:00
|
|
|
table_builder_options.int_tbl_prop_collector_factories, column_family_id,
|
|
|
|
file, table_builder_options.compression_type,
|
A new call back to TablePropertiesCollector to allow users know the entry is add, delete or merge
Summary:
Currently users have no idea a key is add, delete or merge from TablePropertiesCollector call back. Add a new function to add it.
Also refactor the codes so that
(1) make table property collector and internal table property collector two separate data structures with the later one now exposed
(2) table builders only receive internal table properties
Test Plan: Add cases in table_properties_collector_test to cover both of old and new ways of using TablePropertiesCollector.
Reviewers: yhchiang, igor.sugak, rven, igor
Reviewed By: rven, igor
Subscribers: meyering, yoshinorim, maykov, leveldb, dhruba
Differential Revision: https://reviews.facebook.net/D35373
2015-04-06 17:04:30 +00:00
|
|
|
table_builder_options.compression_opts,
|
Shared dictionary compression using reference block
Summary:
This adds a new metablock containing a shared dictionary that is used
to compress all data blocks in the SST file. The size of the shared dictionary
is configurable in CompressionOptions and defaults to 0. It's currently only
used for zlib/lz4/lz4hc, but the block will be stored in the SST regardless of
the compression type if the user chooses a nonzero dictionary size.
During compaction, computes the dictionary by randomly sampling the first
output file in each subcompaction. It pre-computes the intervals to sample
by assuming the output file will have the maximum allowable length. In case
the file is smaller, some of the pre-computed sampling intervals can be beyond
end-of-file, in which case we skip over those samples and the dictionary will
be a bit smaller. After the dictionary is generated using the first file in a
subcompaction, it is loaded into the compression library before writing each
block in each subsequent file of that subcompaction.
On the read path, gets the dictionary from the metablock, if it exists. Then,
loads that dictionary into the compression library before reading each block.
Test Plan: new unit test
Reviewers: yhchiang, IslamAbdelRahman, cyan, sdong
Reviewed By: sdong
Subscribers: andrewkr, yoshinorim, kradhakrishnan, dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D52287
2016-04-28 00:36:03 +00:00
|
|
|
table_builder_options.compression_dict,
|
2016-04-07 06:10:32 +00:00
|
|
|
table_builder_options.skip_filters,
|
|
|
|
table_builder_options.column_family_name);
|
2013-11-20 06:00:48 +00:00
|
|
|
|
|
|
|
return table_builder;
|
2013-10-29 00:54:09 +00:00
|
|
|
}
|
2013-11-20 06:00:48 +00:00
|
|
|
|
2014-10-18 04:18:36 +00:00
|
|
|
Status BlockBasedTableFactory::SanitizeOptions(
|
|
|
|
const DBOptions& db_opts,
|
|
|
|
const ColumnFamilyOptions& cf_opts) const {
|
|
|
|
if (table_options_.index_type == BlockBasedTableOptions::kHashSearch &&
|
|
|
|
cf_opts.prefix_extractor == nullptr) {
|
|
|
|
return Status::InvalidArgument("Hash index is specified for block-based "
|
|
|
|
"table, but prefix_extractor is not given");
|
|
|
|
}
|
2014-10-22 18:52:35 +00:00
|
|
|
if (table_options_.cache_index_and_filter_blocks &&
|
|
|
|
table_options_.no_block_cache) {
|
|
|
|
return Status::InvalidArgument("Enable cache_index_and_filter_blocks, "
|
|
|
|
", but block cache is disabled");
|
|
|
|
}
|
Adding pin_l0_filter_and_index_blocks_in_cache feature and related fixes.
Summary:
When a block based table file is opened, if prefetch_index_and_filter is true, it will prefetch the index and filter blocks, putting them into the block cache.
What this feature adds: when a L0 block based table file is opened, if pin_l0_filter_and_index_blocks_in_cache is true in the options (and prefetch_index_and_filter is true), then the filter and index blocks aren't released back to the block cache at the end of BlockBasedTableReader::Open(). Instead the table reader takes ownership of them, hence pinning them, ie. the LRU cache will never push them out. Meanwhile in the table reader, further accesses will not hit the block cache, thus avoiding lock contention.
Test Plan:
'export TEST_TMPDIR=/dev/shm/ && DISABLE_JEMALLOC=1 OPT=-g make all valgrind_check -j32' is OK.
I didn't run the Java tests, I don't have Java set up on my devserver.
Reviewers: sdong
Reviewed By: sdong
Subscribers: andrewkr, dhruba
Differential Revision: https://reviews.facebook.net/D56133
2016-04-01 17:42:39 +00:00
|
|
|
if (table_options_.pin_l0_filter_and_index_blocks_in_cache &&
|
|
|
|
table_options_.no_block_cache) {
|
|
|
|
return Status::InvalidArgument(
|
|
|
|
"Enable pin_l0_filter_and_index_blocks_in_cache, "
|
|
|
|
", but block cache is disabled");
|
|
|
|
}
|
2015-01-15 00:24:24 +00:00
|
|
|
if (!BlockBasedTableSupportedVersion(table_options_.format_version)) {
|
2015-01-13 22:33:04 +00:00
|
|
|
return Status::InvalidArgument(
|
2015-01-15 00:24:24 +00:00
|
|
|
"Unsupported BlockBasedTable format_version. Please check "
|
|
|
|
"include/rocksdb/table.h for more info");
|
2015-01-13 22:33:04 +00:00
|
|
|
}
|
2014-10-18 04:18:36 +00:00
|
|
|
return Status::OK();
|
|
|
|
}
|
|
|
|
|
2014-08-25 21:24:09 +00:00
|
|
|
std::string BlockBasedTableFactory::GetPrintableTableOptions() const {
|
|
|
|
std::string ret;
|
|
|
|
ret.reserve(20000);
|
|
|
|
const int kBufferSize = 200;
|
|
|
|
char buffer[kBufferSize];
|
|
|
|
|
|
|
|
snprintf(buffer, kBufferSize, " flush_block_policy_factory: %s (%p)\n",
|
|
|
|
table_options_.flush_block_policy_factory->Name(),
|
2016-03-16 21:57:57 +00:00
|
|
|
static_cast<void*>(table_options_.flush_block_policy_factory.get()));
|
2014-08-25 21:24:09 +00:00
|
|
|
ret.append(buffer);
|
|
|
|
snprintf(buffer, kBufferSize, " cache_index_and_filter_blocks: %d\n",
|
|
|
|
table_options_.cache_index_and_filter_blocks);
|
|
|
|
ret.append(buffer);
|
Adding pin_l0_filter_and_index_blocks_in_cache feature and related fixes.
Summary:
When a block based table file is opened, if prefetch_index_and_filter is true, it will prefetch the index and filter blocks, putting them into the block cache.
What this feature adds: when a L0 block based table file is opened, if pin_l0_filter_and_index_blocks_in_cache is true in the options (and prefetch_index_and_filter is true), then the filter and index blocks aren't released back to the block cache at the end of BlockBasedTableReader::Open(). Instead the table reader takes ownership of them, hence pinning them, ie. the LRU cache will never push them out. Meanwhile in the table reader, further accesses will not hit the block cache, thus avoiding lock contention.
Test Plan:
'export TEST_TMPDIR=/dev/shm/ && DISABLE_JEMALLOC=1 OPT=-g make all valgrind_check -j32' is OK.
I didn't run the Java tests, I don't have Java set up on my devserver.
Reviewers: sdong
Reviewed By: sdong
Subscribers: andrewkr, dhruba
Differential Revision: https://reviews.facebook.net/D56133
2016-04-01 17:42:39 +00:00
|
|
|
snprintf(buffer, kBufferSize,
|
|
|
|
" pin_l0_filter_and_index_blocks_in_cache: %d\n",
|
|
|
|
table_options_.pin_l0_filter_and_index_blocks_in_cache);
|
|
|
|
ret.append(buffer);
|
2014-08-25 21:24:09 +00:00
|
|
|
snprintf(buffer, kBufferSize, " index_type: %d\n",
|
|
|
|
table_options_.index_type);
|
|
|
|
ret.append(buffer);
|
|
|
|
snprintf(buffer, kBufferSize, " hash_index_allow_collision: %d\n",
|
|
|
|
table_options_.hash_index_allow_collision);
|
|
|
|
ret.append(buffer);
|
|
|
|
snprintf(buffer, kBufferSize, " checksum: %d\n",
|
|
|
|
table_options_.checksum);
|
|
|
|
ret.append(buffer);
|
|
|
|
snprintf(buffer, kBufferSize, " no_block_cache: %d\n",
|
|
|
|
table_options_.no_block_cache);
|
|
|
|
ret.append(buffer);
|
|
|
|
snprintf(buffer, kBufferSize, " block_cache: %p\n",
|
2016-03-16 21:57:57 +00:00
|
|
|
static_cast<void*>(table_options_.block_cache.get()));
|
2014-08-25 21:24:09 +00:00
|
|
|
ret.append(buffer);
|
|
|
|
if (table_options_.block_cache) {
|
2015-07-01 23:13:49 +00:00
|
|
|
snprintf(buffer, kBufferSize, " block_cache_size: %" ROCKSDB_PRIszt "\n",
|
2014-08-25 21:24:09 +00:00
|
|
|
table_options_.block_cache->GetCapacity());
|
|
|
|
ret.append(buffer);
|
|
|
|
}
|
|
|
|
snprintf(buffer, kBufferSize, " block_cache_compressed: %p\n",
|
2016-03-16 21:57:57 +00:00
|
|
|
static_cast<void*>(table_options_.block_cache_compressed.get()));
|
2014-08-25 21:24:09 +00:00
|
|
|
ret.append(buffer);
|
|
|
|
if (table_options_.block_cache_compressed) {
|
2015-07-13 19:11:05 +00:00
|
|
|
snprintf(buffer, kBufferSize,
|
|
|
|
" block_cache_compressed_size: %" ROCKSDB_PRIszt "\n",
|
2014-08-25 21:24:09 +00:00
|
|
|
table_options_.block_cache_compressed->GetCapacity());
|
|
|
|
ret.append(buffer);
|
|
|
|
}
|
2015-07-01 23:13:49 +00:00
|
|
|
snprintf(buffer, kBufferSize, " block_size: %" ROCKSDB_PRIszt "\n",
|
2014-08-25 21:24:09 +00:00
|
|
|
table_options_.block_size);
|
|
|
|
ret.append(buffer);
|
|
|
|
snprintf(buffer, kBufferSize, " block_size_deviation: %d\n",
|
|
|
|
table_options_.block_size_deviation);
|
|
|
|
ret.append(buffer);
|
|
|
|
snprintf(buffer, kBufferSize, " block_restart_interval: %d\n",
|
|
|
|
table_options_.block_restart_interval);
|
|
|
|
ret.append(buffer);
|
2016-02-05 18:22:37 +00:00
|
|
|
snprintf(buffer, kBufferSize, " index_block_restart_interval: %d\n",
|
|
|
|
table_options_.index_block_restart_interval);
|
|
|
|
ret.append(buffer);
|
2014-08-25 21:24:09 +00:00
|
|
|
snprintf(buffer, kBufferSize, " filter_policy: %s\n",
|
|
|
|
table_options_.filter_policy == nullptr ?
|
|
|
|
"nullptr" : table_options_.filter_policy->Name());
|
|
|
|
ret.append(buffer);
|
|
|
|
snprintf(buffer, kBufferSize, " whole_key_filtering: %d\n",
|
|
|
|
table_options_.whole_key_filtering);
|
2015-10-31 01:33:01 +00:00
|
|
|
ret.append(buffer);
|
|
|
|
snprintf(buffer, kBufferSize, " skip_table_builder_flush: %d\n",
|
|
|
|
table_options_.skip_table_builder_flush);
|
|
|
|
ret.append(buffer);
|
2015-01-13 22:33:04 +00:00
|
|
|
snprintf(buffer, kBufferSize, " format_version: %d\n",
|
|
|
|
table_options_.format_version);
|
2014-08-25 21:24:09 +00:00
|
|
|
ret.append(buffer);
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
2015-10-30 22:58:46 +00:00
|
|
|
const BlockBasedTableOptions& BlockBasedTableFactory::table_options() const {
|
2014-11-21 03:24:39 +00:00
|
|
|
return table_options_;
|
|
|
|
}
|
|
|
|
|
2014-01-28 05:58:46 +00:00
|
|
|
TableFactory* NewBlockBasedTableFactory(
|
2015-11-18 00:41:54 +00:00
|
|
|
const BlockBasedTableOptions& _table_options) {
|
|
|
|
return new BlockBasedTableFactory(_table_options);
|
2014-01-28 05:58:46 +00:00
|
|
|
}
|
|
|
|
|
2014-03-01 02:19:07 +00:00
|
|
|
const std::string BlockBasedTablePropertyNames::kIndexType =
|
|
|
|
"rocksdb.block.based.table.index.type";
|
2015-02-05 01:03:57 +00:00
|
|
|
const std::string BlockBasedTablePropertyNames::kWholeKeyFiltering =
|
|
|
|
"rocksdb.block.based.table.whole.key.filtering";
|
|
|
|
const std::string BlockBasedTablePropertyNames::kPrefixFiltering =
|
|
|
|
"rocksdb.block.based.table.prefix.filtering";
|
2014-05-15 21:09:03 +00:00
|
|
|
const std::string kHashIndexPrefixesBlock = "rocksdb.hashindex.prefixes";
|
|
|
|
const std::string kHashIndexPrefixesMetadataBlock =
|
|
|
|
"rocksdb.hashindex.metadata";
|
2015-02-05 01:03:57 +00:00
|
|
|
const std::string kPropTrue = "1";
|
|
|
|
const std::string kPropFalse = "0";
|
2014-03-01 02:19:07 +00:00
|
|
|
|
2013-10-29 00:54:09 +00:00
|
|
|
} // namespace rocksdb
|