2013-10-16 23:59:46 +02:00
|
|
|
// Copyright (c) 2013, Facebook, Inc. All rights reserved.
|
|
|
|
// 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.
|
|
|
|
//
|
2011-03-18 23:37:00 +01:00
|
|
|
// 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 "db/table_cache.h"
|
|
|
|
|
|
|
|
#include "db/filename.h"
|
2014-01-07 05:29:17 +01:00
|
|
|
#include "db/version_edit.h"
|
2013-02-25 22:58:34 +01:00
|
|
|
|
2013-08-23 17:38:13 +02:00
|
|
|
#include "rocksdb/statistics.h"
|
2014-01-28 06:58:46 +01:00
|
|
|
#include "table/table_reader.h"
|
2011-03-18 23:37:00 +01:00
|
|
|
#include "util/coding.h"
|
2013-06-07 19:02:28 +02:00
|
|
|
#include "util/stop_watch.h"
|
2011-03-18 23:37:00 +01:00
|
|
|
|
2013-10-04 06:49:15 +02:00
|
|
|
namespace rocksdb {
|
2011-03-18 23:37:00 +01:00
|
|
|
|
|
|
|
static void DeleteEntry(const Slice& key, void* value) {
|
2013-10-30 18:52:33 +01:00
|
|
|
TableReader* table_reader = reinterpret_cast<TableReader*>(value);
|
|
|
|
delete table_reader;
|
2011-03-18 23:37:00 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
static void UnrefEntry(void* arg1, void* arg2) {
|
|
|
|
Cache* cache = reinterpret_cast<Cache*>(arg1);
|
|
|
|
Cache::Handle* h = reinterpret_cast<Cache::Handle*>(arg2);
|
|
|
|
cache->Release(h);
|
|
|
|
}
|
|
|
|
|
2014-01-02 19:29:48 +01:00
|
|
|
static Slice GetSliceForFileNumber(uint64_t* file_number) {
|
|
|
|
return Slice(reinterpret_cast<const char*>(file_number),
|
|
|
|
sizeof(*file_number));
|
2013-12-27 01:25:45 +01:00
|
|
|
}
|
|
|
|
|
2014-02-05 21:20:40 +01:00
|
|
|
TableCache::TableCache(const std::string& dbname, const Options* options,
|
[CF] Rethink table cache
Summary:
Adapting table cache to column families is interesting. We want table cache to be global LRU, so if some column families are use not as often as others, we want them to be evicted from cache. However, current TableCache object also constructs tables on its own. If table is not found in the cache, TableCache automatically creates new table. We want each column family to be able to specify different table factory.
To solve the problem, we still have a single LRU, but we provide the LRUCache object to TableCache on construction. We have one TableCache per column family, but the underyling cache is shared by all TableCache objects.
This allows us to have a global LRU, but still be able to support different table factories for different column families. Also, in the future it will also be able to support different directories for different column families.
Test Plan: make check
Reviewers: dhruba, haobo, kailiu, sdong
CC: leveldb
Differential Revision: https://reviews.facebook.net/D15915
2014-02-05 18:07:55 +01:00
|
|
|
const EnvOptions& storage_options, Cache* const cache)
|
2014-02-05 21:20:40 +01:00
|
|
|
: env_(options->env),
|
2011-03-18 23:37:00 +01:00
|
|
|
dbname_(dbname),
|
2014-02-05 21:20:40 +01:00
|
|
|
options_(options),
|
2013-03-15 01:00:04 +01:00
|
|
|
storage_options_(storage_options),
|
[CF] Rethink table cache
Summary:
Adapting table cache to column families is interesting. We want table cache to be global LRU, so if some column families are use not as often as others, we want them to be evicted from cache. However, current TableCache object also constructs tables on its own. If table is not found in the cache, TableCache automatically creates new table. We want each column family to be able to specify different table factory.
To solve the problem, we still have a single LRU, but we provide the LRUCache object to TableCache on construction. We have one TableCache per column family, but the underyling cache is shared by all TableCache objects.
This allows us to have a global LRU, but still be able to support different table factories for different column families. Also, in the future it will also be able to support different directories for different column families.
Test Plan: make check
Reviewers: dhruba, haobo, kailiu, sdong
CC: leveldb
Differential Revision: https://reviews.facebook.net/D15915
2014-02-05 18:07:55 +01:00
|
|
|
cache_(cache) {}
|
2011-03-18 23:37:00 +01:00
|
|
|
|
|
|
|
TableCache::~TableCache() {
|
|
|
|
}
|
|
|
|
|
2014-01-07 05:29:17 +01:00
|
|
|
TableReader* TableCache::GetTableReaderFromHandle(Cache::Handle* handle) {
|
|
|
|
return reinterpret_cast<TableReader*>(cache_->Value(handle));
|
|
|
|
}
|
|
|
|
|
|
|
|
void TableCache::ReleaseHandle(Cache::Handle* handle) {
|
|
|
|
cache_->Release(handle);
|
|
|
|
}
|
|
|
|
|
2013-03-15 01:00:04 +01:00
|
|
|
Status TableCache::FindTable(const EnvOptions& toptions,
|
2014-01-27 22:53:22 +01:00
|
|
|
const InternalKeyComparator& internal_comparator,
|
2013-03-15 01:00:04 +01:00
|
|
|
uint64_t file_number, uint64_t file_size,
|
2013-07-13 01:56:52 +02:00
|
|
|
Cache::Handle** handle, bool* table_io,
|
|
|
|
const bool no_io) {
|
2012-04-17 17:36:46 +02:00
|
|
|
Status s;
|
2014-01-02 19:29:48 +01:00
|
|
|
Slice key = GetSliceForFileNumber(&file_number);
|
2012-04-17 17:36:46 +02:00
|
|
|
*handle = cache_->Lookup(key);
|
2013-02-25 22:58:34 +01:00
|
|
|
if (*handle == nullptr) {
|
2013-07-13 01:56:52 +02:00
|
|
|
if (no_io) { // Dont do IO and return a not-found status
|
2013-08-25 07:48:51 +02:00
|
|
|
return Status::Incomplete("Table not found in table_cache, no_io is set");
|
2013-07-13 01:56:52 +02:00
|
|
|
}
|
|
|
|
if (table_io != nullptr) {
|
|
|
|
*table_io = true; // we had to do IO from storage
|
2012-09-27 10:05:38 +02:00
|
|
|
}
|
2011-03-18 23:37:00 +01:00
|
|
|
std::string fname = TableFileName(dbname_, file_number);
|
2013-01-20 11:07:13 +01:00
|
|
|
unique_ptr<RandomAccessFile> file;
|
2013-10-30 18:52:33 +01:00
|
|
|
unique_ptr<TableReader> table_reader;
|
2013-03-15 01:00:04 +01:00
|
|
|
s = env_->NewRandomAccessFile(fname, &file, toptions);
|
2014-02-05 21:20:40 +01:00
|
|
|
RecordTick(options_->statistics.get(), NO_FILE_OPENS);
|
2011-03-18 23:37:00 +01:00
|
|
|
if (s.ok()) {
|
2014-02-05 21:20:40 +01:00
|
|
|
if (options_->advise_random_on_open) {
|
2013-05-18 00:53:01 +02:00
|
|
|
file->Hint(RandomAccessFile::RANDOM);
|
|
|
|
}
|
2014-02-05 21:20:40 +01:00
|
|
|
StopWatch sw(env_, options_->statistics.get(), TABLE_OPEN_IO_MICROS);
|
2014-01-28 06:58:46 +01:00
|
|
|
s = options_->table_factory->NewTableReader(
|
2014-01-27 22:53:22 +01:00
|
|
|
*options_, toptions, internal_comparator, std::move(file), file_size,
|
|
|
|
&table_reader);
|
2011-03-18 23:37:00 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
if (!s.ok()) {
|
2013-10-30 18:52:33 +01:00
|
|
|
assert(table_reader == nullptr);
|
2014-02-05 21:20:40 +01:00
|
|
|
RecordTick(options_->statistics.get(), NO_FILE_ERRORS);
|
2011-03-18 23:37:00 +01:00
|
|
|
// We do not cache error results so that if the error is transient,
|
|
|
|
// or somebody repairs the file, we recover automatically.
|
2012-04-17 17:36:46 +02:00
|
|
|
} else {
|
2013-03-04 22:33:16 +01:00
|
|
|
assert(file.get() == nullptr);
|
2013-10-30 18:52:33 +01:00
|
|
|
*handle = cache_->Insert(key, table_reader.release(), 1, &DeleteEntry);
|
2011-03-18 23:37:00 +01:00
|
|
|
}
|
2012-04-17 17:36:46 +02:00
|
|
|
}
|
|
|
|
return s;
|
|
|
|
}
|
2011-03-18 23:37:00 +01:00
|
|
|
|
2012-04-17 17:36:46 +02:00
|
|
|
Iterator* TableCache::NewIterator(const ReadOptions& options,
|
2013-03-15 01:00:04 +01:00
|
|
|
const EnvOptions& toptions,
|
2014-01-27 22:53:22 +01:00
|
|
|
const InternalKeyComparator& icomparator,
|
2014-01-07 05:29:17 +01:00
|
|
|
const FileMetaData& file_meta,
|
2013-10-30 18:52:33 +01:00
|
|
|
TableReader** table_reader_ptr,
|
2013-05-18 00:53:01 +02:00
|
|
|
bool for_compaction) {
|
2013-10-30 18:52:33 +01:00
|
|
|
if (table_reader_ptr != nullptr) {
|
|
|
|
*table_reader_ptr = nullptr;
|
2012-04-17 17:36:46 +02:00
|
|
|
}
|
2014-01-07 05:29:17 +01:00
|
|
|
Cache::Handle* handle = file_meta.table_reader_handle;
|
|
|
|
Status s;
|
|
|
|
if (!handle) {
|
2014-01-27 22:53:22 +01:00
|
|
|
s = FindTable(toptions, icomparator, file_meta.number, file_meta.file_size,
|
|
|
|
&handle, nullptr, options.read_tier == kBlockCacheTier);
|
2014-01-07 05:29:17 +01:00
|
|
|
}
|
2012-04-17 17:36:46 +02:00
|
|
|
if (!s.ok()) {
|
|
|
|
return NewErrorIterator(s);
|
2011-03-18 23:37:00 +01:00
|
|
|
}
|
|
|
|
|
2014-01-07 05:29:17 +01:00
|
|
|
TableReader* table_reader = GetTableReaderFromHandle(handle);
|
2013-10-30 18:52:33 +01:00
|
|
|
Iterator* result = table_reader->NewIterator(options);
|
2014-01-07 05:29:17 +01:00
|
|
|
if (!file_meta.table_reader_handle) {
|
2014-02-07 00:42:16 +01:00
|
|
|
result->RegisterCleanup(&UnrefEntry, cache_, handle);
|
2014-01-07 05:29:17 +01:00
|
|
|
}
|
2013-10-30 18:52:33 +01:00
|
|
|
if (table_reader_ptr != nullptr) {
|
|
|
|
*table_reader_ptr = table_reader;
|
2011-03-18 23:37:00 +01:00
|
|
|
}
|
2013-05-18 00:53:01 +02:00
|
|
|
|
|
|
|
if (for_compaction) {
|
2013-10-30 18:52:33 +01:00
|
|
|
table_reader->SetupForCompaction();
|
2013-05-18 00:53:01 +02:00
|
|
|
}
|
|
|
|
|
2011-03-18 23:37:00 +01:00
|
|
|
return result;
|
|
|
|
}
|
|
|
|
|
2012-04-17 17:36:46 +02:00
|
|
|
Status TableCache::Get(const ReadOptions& options,
|
2014-01-27 22:53:22 +01:00
|
|
|
const InternalKeyComparator& internal_comparator,
|
|
|
|
const FileMetaData& file_meta, const Slice& k, void* arg,
|
|
|
|
bool (*saver)(void*, const ParsedInternalKey&,
|
|
|
|
const Slice&, bool),
|
|
|
|
bool* table_io, void (*mark_key_may_exist)(void*)) {
|
2014-01-07 05:29:17 +01:00
|
|
|
Cache::Handle* handle = file_meta.table_reader_handle;
|
|
|
|
Status s;
|
|
|
|
if (!handle) {
|
2014-01-27 22:53:22 +01:00
|
|
|
s = FindTable(storage_options_, internal_comparator, file_meta.number,
|
|
|
|
file_meta.file_size, &handle, table_io,
|
|
|
|
options.read_tier == kBlockCacheTier);
|
2014-01-07 05:29:17 +01:00
|
|
|
}
|
2012-04-17 17:36:46 +02:00
|
|
|
if (s.ok()) {
|
2014-01-07 05:29:17 +01:00
|
|
|
TableReader* t = GetTableReaderFromHandle(handle);
|
2013-10-29 01:54:09 +01:00
|
|
|
s = t->Get(options, k, arg, saver, mark_key_may_exist);
|
2014-01-07 05:29:17 +01:00
|
|
|
if (!file_meta.table_reader_handle) {
|
|
|
|
ReleaseHandle(handle);
|
|
|
|
}
|
2013-08-25 07:48:51 +02:00
|
|
|
} else if (options.read_tier && s.IsIncomplete()) {
|
2013-07-13 01:56:52 +02:00
|
|
|
// Couldnt find Table in cache but treat as kFound if no_io set
|
|
|
|
(*mark_key_may_exist)(arg);
|
|
|
|
return Status::OK();
|
2012-04-17 17:36:46 +02:00
|
|
|
}
|
|
|
|
return s;
|
|
|
|
}
|
2014-02-14 01:28:21 +01:00
|
|
|
Status TableCache::GetTableProperties(
|
|
|
|
const EnvOptions& toptions,
|
|
|
|
const InternalKeyComparator& internal_comparator,
|
|
|
|
const FileMetaData& file_meta,
|
|
|
|
std::shared_ptr<const TableProperties>* properties, bool no_io) {
|
|
|
|
Status s;
|
|
|
|
auto table_handle = file_meta.table_reader_handle;
|
|
|
|
// table already been pre-loaded?
|
|
|
|
if (table_handle) {
|
|
|
|
auto table = GetTableReaderFromHandle(table_handle);
|
|
|
|
*properties = table->GetTableProperties();
|
|
|
|
return s;
|
|
|
|
}
|
|
|
|
|
|
|
|
bool table_io;
|
|
|
|
s = FindTable(toptions, internal_comparator, file_meta.number,
|
|
|
|
file_meta.file_size, &table_handle, &table_io, no_io);
|
|
|
|
if (!s.ok()) {
|
|
|
|
return s;
|
|
|
|
}
|
|
|
|
assert(table_handle);
|
|
|
|
auto table = GetTableReaderFromHandle(table_handle);
|
|
|
|
*properties = table->GetTableProperties();
|
|
|
|
ReleaseHandle(table_handle);
|
|
|
|
return s;
|
|
|
|
}
|
2012-04-17 17:36:46 +02:00
|
|
|
|
2013-08-23 23:49:57 +02:00
|
|
|
bool TableCache::PrefixMayMatch(const ReadOptions& options,
|
2014-01-27 22:53:22 +01:00
|
|
|
const InternalKeyComparator& icomparator,
|
2014-03-21 01:32:55 +01:00
|
|
|
const FileMetaData& file_meta,
|
2014-01-27 22:53:22 +01:00
|
|
|
const Slice& internal_prefix, bool* table_io) {
|
2013-08-23 23:49:57 +02:00
|
|
|
bool may_match = true;
|
2014-03-21 01:32:55 +01:00
|
|
|
auto table_handle = file_meta.table_reader_handle;
|
|
|
|
if (table_handle == nullptr) {
|
|
|
|
// Need to get table handle from file number
|
|
|
|
Status s = FindTable(storage_options_, icomparator, file_meta.number,
|
|
|
|
file_meta.file_size, &table_handle, table_io);
|
|
|
|
if (!s.ok()) {
|
|
|
|
return may_match;
|
|
|
|
}
|
2013-08-23 23:49:57 +02:00
|
|
|
}
|
2014-03-21 01:32:55 +01:00
|
|
|
|
|
|
|
auto table = GetTableReaderFromHandle(table_handle);
|
|
|
|
may_match = table->PrefixMayMatch(internal_prefix);
|
|
|
|
|
|
|
|
if (file_meta.table_reader_handle == nullptr) {
|
|
|
|
// Need to release handle if it is generated from here.
|
|
|
|
ReleaseHandle(table_handle);
|
|
|
|
}
|
|
|
|
|
2013-08-23 23:49:57 +02:00
|
|
|
return may_match;
|
|
|
|
}
|
|
|
|
|
[CF] Rethink table cache
Summary:
Adapting table cache to column families is interesting. We want table cache to be global LRU, so if some column families are use not as often as others, we want them to be evicted from cache. However, current TableCache object also constructs tables on its own. If table is not found in the cache, TableCache automatically creates new table. We want each column family to be able to specify different table factory.
To solve the problem, we still have a single LRU, but we provide the LRUCache object to TableCache on construction. We have one TableCache per column family, but the underyling cache is shared by all TableCache objects.
This allows us to have a global LRU, but still be able to support different table factories for different column families. Also, in the future it will also be able to support different directories for different column families.
Test Plan: make check
Reviewers: dhruba, haobo, kailiu, sdong
CC: leveldb
Differential Revision: https://reviews.facebook.net/D15915
2014-02-05 18:07:55 +01:00
|
|
|
void TableCache::Evict(Cache* cache, uint64_t file_number) {
|
|
|
|
cache->Erase(GetSliceForFileNumber(&file_number));
|
2011-03-18 23:37:00 +01:00
|
|
|
}
|
|
|
|
|
2013-10-04 06:49:15 +02:00
|
|
|
} // namespace rocksdb
|