2016-02-10 00:12:00 +01:00
|
|
|
// Copyright (c) 2011-present, Facebook, Inc. All rights reserved.
|
2017-07-16 01:03:42 +02:00
|
|
|
// 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).
|
2013-10-16 23:59:46 +02:00
|
|
|
//
|
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.
|
2014-02-05 01:21:47 +01:00
|
|
|
|
|
|
|
#include <stdio.h>
|
2014-01-24 20:09:04 +01:00
|
|
|
#include <algorithm>
|
2014-06-13 04:03:22 +02:00
|
|
|
#include <iostream>
|
2011-03-18 23:37:00 +01:00
|
|
|
#include <map>
|
2013-07-23 23:42:27 +02:00
|
|
|
#include <memory>
|
2015-09-02 22:58:22 +02:00
|
|
|
#include <string>
|
2013-10-10 20:43:24 +02:00
|
|
|
#include <vector>
|
|
|
|
|
2019-05-30 23:47:29 +02:00
|
|
|
#include "block_fetcher.h"
|
2017-05-06 05:10:56 +02:00
|
|
|
#include "cache/lru_cache.h"
|
2011-03-18 23:37:00 +01:00
|
|
|
#include "db/dbformat.h"
|
|
|
|
#include "db/memtable.h"
|
|
|
|
#include "db/write_batch_internal.h"
|
2015-10-16 23:10:33 +02:00
|
|
|
#include "memtable/stl_wrappers.h"
|
2019-05-30 23:47:29 +02:00
|
|
|
#include "meta_blocks.h"
|
2017-04-06 04:02:00 +02:00
|
|
|
#include "monitoring/statistics.h"
|
2017-02-06 23:43:55 +01:00
|
|
|
#include "port/port.h"
|
2013-11-13 07:46:51 +01:00
|
|
|
#include "rocksdb/cache.h"
|
2013-08-23 17:38:13 +02:00
|
|
|
#include "rocksdb/db.h"
|
|
|
|
#include "rocksdb/env.h"
|
2020-02-11 00:42:46 +01:00
|
|
|
#include "rocksdb/file_checksum.h"
|
Introduce a new storage specific Env API (#5761)
Summary:
The current Env API encompasses both storage/file operations, as well as OS related operations. Most of the APIs return a Status, which does not have enough metadata about an error, such as whether its retry-able or not, scope (i.e fault domain) of the error etc., that may be required in order to properly handle a storage error. The file APIs also do not provide enough control over the IO SLA, such as timeout, prioritization, hinting about placement and redundancy etc.
This PR separates out the file/storage APIs from Env into a new FileSystem class. The APIs are updated to return an IOStatus with metadata about the error, as well as to take an IOOptions structure as input in order to allow more control over the IO.
The user can set both ```options.env``` and ```options.file_system``` to specify that RocksDB should use the former for OS related operations and the latter for storage operations. Internally, a ```CompositeEnvWrapper``` has been introduced that inherits from ```Env``` and redirects individual methods to either an ```Env``` implementation or the ```FileSystem``` as appropriate. When options are sanitized during ```DB::Open```, ```options.env``` is replaced with a newly allocated ```CompositeEnvWrapper``` instance if both env and file_system have been specified. This way, the rest of the RocksDB code can continue to function as before.
This PR also ports PosixEnv to the new API by splitting it into two - PosixEnv and PosixFileSystem. PosixEnv is defined as a sub-class of CompositeEnvWrapper, and threading/time functions are overridden with Posix specific implementations in order to avoid an extra level of indirection.
The ```CompositeEnvWrapper``` translates ```IOStatus``` return code to ```Status```, and sets the severity to ```kSoftError``` if the io_status is retryable. The error handling code in RocksDB can then recover the DB automatically.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/5761
Differential Revision: D18868376
Pulled By: anand1976
fbshipit-source-id: 39efe18a162ea746fabac6360ff529baba48486f
2019-12-13 23:47:08 +01:00
|
|
|
#include "rocksdb/file_system.h"
|
2013-08-23 17:38:13 +02:00
|
|
|
#include "rocksdb/iterator.h"
|
|
|
|
#include "rocksdb/memtablerep.h"
|
2015-09-03 00:36:47 +02:00
|
|
|
#include "rocksdb/perf_context.h"
|
2014-03-01 03:19:07 +01:00
|
|
|
#include "rocksdb/slice_transform.h"
|
|
|
|
#include "rocksdb/statistics.h"
|
2016-06-21 03:01:03 +02:00
|
|
|
#include "rocksdb/write_buffer_manager.h"
|
2019-05-30 23:47:29 +02:00
|
|
|
#include "table/block_based/block.h"
|
|
|
|
#include "table/block_based/block_based_table_builder.h"
|
|
|
|
#include "table/block_based/block_based_table_factory.h"
|
|
|
|
#include "table/block_based/block_based_table_reader.h"
|
|
|
|
#include "table/block_based/block_builder.h"
|
|
|
|
#include "table/block_based/flush_block_policy.h"
|
2011-03-18 23:37:00 +01:00
|
|
|
#include "table/format.h"
|
2015-09-02 22:58:22 +02:00
|
|
|
#include "table/get_context.h"
|
2015-10-13 00:06:38 +02:00
|
|
|
#include "table/internal_iterator.h"
|
2019-05-30 23:47:29 +02:00
|
|
|
#include "table/plain/plain_table_factory.h"
|
2015-10-13 00:06:38 +02:00
|
|
|
#include "table/scoped_arena_iterator.h"
|
2016-10-19 01:59:37 +02:00
|
|
|
#include "table/sst_file_writer_collectors.h"
|
2019-05-30 20:21:38 +02:00
|
|
|
#include "test_util/sync_point.h"
|
|
|
|
#include "test_util/testharness.h"
|
|
|
|
#include "test_util/testutil.h"
|
2019-05-31 02:39:43 +02:00
|
|
|
#include "util/compression.h"
|
2020-02-11 00:42:46 +01:00
|
|
|
#include "util/file_checksum_helper.h"
|
2019-05-31 02:39:43 +02:00
|
|
|
#include "util/random.h"
|
|
|
|
#include "util/string_util.h"
|
2016-04-21 19:16:28 +02:00
|
|
|
#include "utilities/merge_operators.h"
|
2015-03-03 02:07:03 +01:00
|
|
|
|
2020-02-20 21:07:53 +01:00
|
|
|
namespace ROCKSDB_NAMESPACE {
|
2011-03-18 23:37:00 +01:00
|
|
|
|
2014-05-01 20:09:32 +02:00
|
|
|
extern const uint64_t kLegacyBlockBasedTableMagicNumber;
|
|
|
|
extern const uint64_t kLegacyPlainTableMagicNumber;
|
|
|
|
extern const uint64_t kBlockBasedTableMagicNumber;
|
|
|
|
extern const uint64_t kPlainTableMagicNumber;
|
|
|
|
|
2013-11-08 06:27:21 +01:00
|
|
|
namespace {
|
2014-01-24 20:09:04 +01:00
|
|
|
|
2019-07-17 22:02:00 +02:00
|
|
|
const std::string kDummyValue(10000, 'o');
|
|
|
|
|
2016-04-21 19:16:28 +02:00
|
|
|
// DummyPropertiesCollector used to test BlockBasedTableProperties
|
|
|
|
class DummyPropertiesCollector : public TablePropertiesCollector {
|
|
|
|
public:
|
2020-10-02 00:21:17 +02:00
|
|
|
const char* Name() const override { return "DummyPropertiesCollector"; }
|
2016-04-21 19:16:28 +02:00
|
|
|
|
2019-02-14 22:52:47 +01:00
|
|
|
Status Finish(UserCollectedProperties* /*properties*/) override {
|
2018-03-05 22:08:17 +01:00
|
|
|
return Status::OK();
|
|
|
|
}
|
2016-04-21 19:16:28 +02:00
|
|
|
|
2019-02-14 22:52:47 +01:00
|
|
|
Status Add(const Slice& /*user_key*/, const Slice& /*value*/) override {
|
2018-03-05 22:08:17 +01:00
|
|
|
return Status::OK();
|
|
|
|
}
|
2016-04-21 19:16:28 +02:00
|
|
|
|
2019-02-14 22:52:47 +01:00
|
|
|
UserCollectedProperties GetReadableProperties() const override {
|
2016-04-21 19:16:28 +02:00
|
|
|
return UserCollectedProperties{};
|
|
|
|
}
|
|
|
|
};
|
|
|
|
|
|
|
|
class DummyPropertiesCollectorFactory1
|
|
|
|
: public TablePropertiesCollectorFactory {
|
|
|
|
public:
|
2019-02-14 22:52:47 +01:00
|
|
|
TablePropertiesCollector* CreateTablePropertiesCollector(
|
|
|
|
TablePropertiesCollectorFactory::Context /*context*/) override {
|
2016-04-21 19:16:28 +02:00
|
|
|
return new DummyPropertiesCollector();
|
|
|
|
}
|
2020-10-02 00:21:17 +02:00
|
|
|
const char* Name() const override {
|
|
|
|
return "DummyPropertiesCollectorFactory1";
|
|
|
|
}
|
2016-04-21 19:16:28 +02:00
|
|
|
};
|
|
|
|
|
|
|
|
class DummyPropertiesCollectorFactory2
|
|
|
|
: public TablePropertiesCollectorFactory {
|
|
|
|
public:
|
2019-02-14 22:52:47 +01:00
|
|
|
TablePropertiesCollector* CreateTablePropertiesCollector(
|
|
|
|
TablePropertiesCollectorFactory::Context /*context*/) override {
|
2016-04-21 19:16:28 +02:00
|
|
|
return new DummyPropertiesCollector();
|
|
|
|
}
|
2020-10-02 00:21:17 +02:00
|
|
|
const char* Name() const override {
|
|
|
|
return "DummyPropertiesCollectorFactory2";
|
|
|
|
}
|
2016-04-21 19:16:28 +02:00
|
|
|
};
|
|
|
|
|
2011-03-18 23:37:00 +01:00
|
|
|
// Return reverse of "key".
|
|
|
|
// Used to test non-lexicographic comparators.
|
2014-01-24 20:09:04 +01:00
|
|
|
std::string Reverse(const Slice& key) {
|
|
|
|
auto rev = key.ToString();
|
|
|
|
std::reverse(rev.begin(), rev.end());
|
2011-03-18 23:37:00 +01:00
|
|
|
return rev;
|
|
|
|
}
|
|
|
|
|
|
|
|
class ReverseKeyComparator : public Comparator {
|
|
|
|
public:
|
2019-02-14 22:52:47 +01:00
|
|
|
const char* Name() const override {
|
2013-10-05 07:32:05 +02:00
|
|
|
return "rocksdb.ReverseBytewiseComparator";
|
2011-03-18 23:37:00 +01:00
|
|
|
}
|
|
|
|
|
2019-02-14 22:52:47 +01:00
|
|
|
int Compare(const Slice& a, const Slice& b) const override {
|
2011-03-18 23:37:00 +01:00
|
|
|
return BytewiseComparator()->Compare(Reverse(a), Reverse(b));
|
|
|
|
}
|
|
|
|
|
2019-02-14 22:52:47 +01:00
|
|
|
void FindShortestSeparator(std::string* start,
|
|
|
|
const Slice& limit) const override {
|
2011-03-18 23:37:00 +01:00
|
|
|
std::string s = Reverse(*start);
|
|
|
|
std::string l = Reverse(limit);
|
|
|
|
BytewiseComparator()->FindShortestSeparator(&s, l);
|
|
|
|
*start = Reverse(s);
|
|
|
|
}
|
|
|
|
|
2019-02-14 22:52:47 +01:00
|
|
|
void FindShortSuccessor(std::string* key) const override {
|
2011-03-18 23:37:00 +01:00
|
|
|
std::string s = Reverse(*key);
|
|
|
|
BytewiseComparator()->FindShortSuccessor(&s);
|
|
|
|
*key = Reverse(s);
|
|
|
|
}
|
|
|
|
};
|
|
|
|
|
2014-01-24 20:09:04 +01:00
|
|
|
ReverseKeyComparator reverse_key_comparator;
|
|
|
|
|
|
|
|
void Increment(const Comparator* cmp, std::string* key) {
|
2011-03-18 23:37:00 +01:00
|
|
|
if (cmp == BytewiseComparator()) {
|
|
|
|
key->push_back('\0');
|
|
|
|
} else {
|
|
|
|
assert(cmp == &reverse_key_comparator);
|
|
|
|
std::string rev = Reverse(*key);
|
|
|
|
rev.push_back('\0');
|
|
|
|
*key = Reverse(rev);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2011-10-31 18:22:06 +01:00
|
|
|
} // namespace
|
2011-03-18 23:37:00 +01:00
|
|
|
|
|
|
|
// Helper class for tests to unify the interface between
|
|
|
|
// BlockBuilder/TableBuilder and Block/Table.
|
|
|
|
class Constructor {
|
|
|
|
public:
|
2015-09-02 22:58:22 +02:00
|
|
|
explicit Constructor(const Comparator* cmp)
|
|
|
|
: data_(stl_wrappers::LessOfComparator(cmp)) {}
|
2011-03-18 23:37:00 +01:00
|
|
|
virtual ~Constructor() { }
|
|
|
|
|
|
|
|
void Add(const std::string& key, const Slice& value) {
|
|
|
|
data_[key] = value.ToString();
|
|
|
|
}
|
|
|
|
|
|
|
|
// Finish constructing the data structure with all the keys that have
|
|
|
|
// been added so far. Returns the keys in sorted order in "*keys"
|
|
|
|
// and stores the key/value pairs in "*kvmap"
|
2015-09-02 22:58:22 +02:00
|
|
|
void Finish(const Options& options, const ImmutableCFOptions& ioptions,
|
2018-05-21 23:33:55 +02:00
|
|
|
const MutableCFOptions& moptions,
|
2014-08-25 23:22:05 +02:00
|
|
|
const BlockBasedTableOptions& table_options,
|
2014-01-27 22:53:22 +01:00
|
|
|
const InternalKeyComparator& internal_comparator,
|
2015-09-02 22:58:22 +02:00
|
|
|
std::vector<std::string>* keys, stl_wrappers::KVMap* kvmap) {
|
2014-01-27 22:53:22 +01:00
|
|
|
last_internal_key_ = &internal_comparator;
|
2011-03-18 23:37:00 +01:00
|
|
|
*kvmap = data_;
|
|
|
|
keys->clear();
|
2015-09-02 22:58:22 +02:00
|
|
|
for (const auto& kv : data_) {
|
|
|
|
keys->push_back(kv.first);
|
2011-03-18 23:37:00 +01:00
|
|
|
}
|
|
|
|
data_.clear();
|
2018-05-21 23:33:55 +02:00
|
|
|
Status s = FinishImpl(options, ioptions, moptions, table_options,
|
2014-09-05 01:18:36 +02:00
|
|
|
internal_comparator, *kvmap);
|
2011-03-18 23:37:00 +01:00
|
|
|
ASSERT_TRUE(s.ok()) << s.ToString();
|
|
|
|
}
|
|
|
|
|
|
|
|
// Construct the data structure from the data in "data"
|
2014-01-27 22:53:22 +01:00
|
|
|
virtual Status FinishImpl(const Options& options,
|
2014-09-05 01:18:36 +02:00
|
|
|
const ImmutableCFOptions& ioptions,
|
2018-05-21 23:33:55 +02:00
|
|
|
const MutableCFOptions& moptions,
|
2014-08-25 23:22:05 +02:00
|
|
|
const BlockBasedTableOptions& table_options,
|
2014-01-27 22:53:22 +01:00
|
|
|
const InternalKeyComparator& internal_comparator,
|
2015-09-02 22:58:22 +02:00
|
|
|
const stl_wrappers::KVMap& data) = 0;
|
2011-03-18 23:37:00 +01:00
|
|
|
|
2018-05-21 23:33:55 +02:00
|
|
|
virtual InternalIterator* NewIterator(
|
|
|
|
const SliceTransform* prefix_extractor = nullptr) const = 0;
|
2011-03-18 23:37:00 +01:00
|
|
|
|
2015-09-02 22:58:22 +02:00
|
|
|
virtual const stl_wrappers::KVMap& data() { return data_; }
|
2011-03-18 23:37:00 +01:00
|
|
|
|
2014-09-05 02:40:41 +02:00
|
|
|
virtual bool IsArenaMode() const { return false; }
|
|
|
|
|
2013-03-01 03:04:58 +01:00
|
|
|
virtual DB* db() const { return nullptr; } // Overridden in DBConstructor
|
2011-03-21 20:40:57 +01:00
|
|
|
|
2014-09-05 02:40:41 +02:00
|
|
|
virtual bool AnywayDeleteIterator() const { return false; }
|
|
|
|
|
2014-01-27 22:53:22 +01:00
|
|
|
protected:
|
|
|
|
const InternalKeyComparator* last_internal_key_;
|
|
|
|
|
2011-03-18 23:37:00 +01:00
|
|
|
private:
|
2015-09-02 22:58:22 +02:00
|
|
|
stl_wrappers::KVMap data_;
|
2011-03-18 23:37:00 +01:00
|
|
|
};
|
|
|
|
|
2013-12-20 18:35:24 +01:00
|
|
|
// A helper class that converts internal format keys into user keys
|
2015-10-13 00:06:38 +02:00
|
|
|
class KeyConvertingIterator : public InternalIterator {
|
2011-03-18 23:37:00 +01:00
|
|
|
public:
|
2015-10-13 00:06:38 +02:00
|
|
|
explicit KeyConvertingIterator(InternalIterator* iter,
|
|
|
|
bool arena_mode = false)
|
2014-09-05 02:40:41 +02:00
|
|
|
: iter_(iter), arena_mode_(arena_mode) {}
|
2019-02-14 22:52:47 +01:00
|
|
|
~KeyConvertingIterator() override {
|
2014-09-05 02:40:41 +02:00
|
|
|
if (arena_mode_) {
|
2015-10-13 00:06:38 +02:00
|
|
|
iter_->~InternalIterator();
|
2014-09-05 02:40:41 +02:00
|
|
|
} else {
|
|
|
|
delete iter_;
|
|
|
|
}
|
|
|
|
}
|
2019-02-14 22:52:47 +01:00
|
|
|
bool Valid() const override { return iter_->Valid() && status_.ok(); }
|
|
|
|
void Seek(const Slice& target) override {
|
2013-12-20 18:35:24 +01:00
|
|
|
ParsedInternalKey ikey(target, kMaxSequenceNumber, kTypeValue);
|
|
|
|
std::string encoded;
|
|
|
|
AppendInternalKey(&encoded, ikey);
|
|
|
|
iter_->Seek(encoded);
|
|
|
|
}
|
2019-02-14 22:52:47 +01:00
|
|
|
void SeekForPrev(const Slice& target) override {
|
2016-09-28 03:20:57 +02:00
|
|
|
ParsedInternalKey ikey(target, kMaxSequenceNumber, kTypeValue);
|
|
|
|
std::string encoded;
|
|
|
|
AppendInternalKey(&encoded, ikey);
|
|
|
|
iter_->SeekForPrev(encoded);
|
|
|
|
}
|
2019-02-14 22:52:47 +01:00
|
|
|
void SeekToFirst() override { iter_->SeekToFirst(); }
|
|
|
|
void SeekToLast() override { iter_->SeekToLast(); }
|
|
|
|
void Next() override { iter_->Next(); }
|
|
|
|
void Prev() override { iter_->Prev(); }
|
2020-08-05 19:42:56 +02:00
|
|
|
IterBoundCheck UpperBoundCheckResult() override {
|
|
|
|
return iter_->UpperBoundCheckResult();
|
|
|
|
}
|
2013-12-20 18:35:24 +01:00
|
|
|
|
2019-02-14 22:52:47 +01:00
|
|
|
Slice key() const override {
|
2013-12-20 18:35:24 +01:00
|
|
|
assert(Valid());
|
2014-11-06 20:14:28 +01:00
|
|
|
ParsedInternalKey parsed_key;
|
2020-10-28 18:11:13 +01:00
|
|
|
Status pik_status =
|
|
|
|
ParseInternalKey(iter_->key(), &parsed_key, true /* log_err_key */);
|
|
|
|
if (!pik_status.ok()) {
|
|
|
|
status_ = pik_status;
|
|
|
|
return Slice(status_.getState());
|
2013-12-20 18:35:24 +01:00
|
|
|
}
|
2014-11-06 20:14:28 +01:00
|
|
|
return parsed_key.user_key;
|
2011-03-18 23:37:00 +01:00
|
|
|
}
|
2014-01-24 19:57:15 +01:00
|
|
|
|
2019-02-14 22:52:47 +01:00
|
|
|
Slice value() const override { return iter_->value(); }
|
|
|
|
Status status() const override {
|
2013-12-20 18:35:24 +01:00
|
|
|
return status_.ok() ? iter_->status() : status_;
|
|
|
|
}
|
|
|
|
|
|
|
|
private:
|
|
|
|
mutable Status status_;
|
2015-10-13 00:06:38 +02:00
|
|
|
InternalIterator* iter_;
|
2014-09-05 02:40:41 +02:00
|
|
|
bool arena_mode_;
|
2013-12-20 18:35:24 +01:00
|
|
|
|
|
|
|
// No copying allowed
|
|
|
|
KeyConvertingIterator(const KeyConvertingIterator&);
|
|
|
|
void operator=(const KeyConvertingIterator&);
|
|
|
|
};
|
|
|
|
|
2020-07-08 02:25:08 +02:00
|
|
|
// `BlockConstructor` APIs always accept/return user keys.
|
|
|
|
class BlockConstructor : public Constructor {
|
|
|
|
public:
|
|
|
|
explicit BlockConstructor(const Comparator* cmp)
|
|
|
|
: Constructor(cmp), comparator_(cmp), block_(nullptr) {}
|
|
|
|
~BlockConstructor() override { delete block_; }
|
|
|
|
Status FinishImpl(const Options& /*options*/,
|
|
|
|
const ImmutableCFOptions& /*ioptions*/,
|
|
|
|
const MutableCFOptions& /*moptions*/,
|
|
|
|
const BlockBasedTableOptions& table_options,
|
|
|
|
const InternalKeyComparator& /*internal_comparator*/,
|
|
|
|
const stl_wrappers::KVMap& kv_map) override {
|
|
|
|
delete block_;
|
|
|
|
block_ = nullptr;
|
|
|
|
BlockBuilder builder(table_options.block_restart_interval);
|
|
|
|
|
|
|
|
for (const auto& kv : kv_map) {
|
|
|
|
// `DataBlockIter` assumes it reads only internal keys. `BlockConstructor`
|
|
|
|
// clients provide user keys, so we need to convert to internal key format
|
|
|
|
// before writing the data block.
|
|
|
|
ParsedInternalKey ikey(kv.first, kMaxSequenceNumber, kTypeValue);
|
|
|
|
std::string encoded;
|
|
|
|
AppendInternalKey(&encoded, ikey);
|
|
|
|
builder.Add(encoded, kv.second);
|
|
|
|
}
|
|
|
|
// Open the block
|
|
|
|
data_ = builder.Finish().ToString();
|
|
|
|
BlockContents contents;
|
|
|
|
contents.data = data_;
|
|
|
|
block_ = new Block(std::move(contents));
|
|
|
|
return Status::OK();
|
|
|
|
}
|
|
|
|
InternalIterator* NewIterator(
|
|
|
|
const SliceTransform* /*prefix_extractor*/) const override {
|
|
|
|
// `DataBlockIter` returns the internal keys it reads.
|
|
|
|
// `KeyConvertingIterator` converts them to user keys before they are
|
|
|
|
// exposed to the `BlockConstructor` clients.
|
|
|
|
return new KeyConvertingIterator(
|
|
|
|
block_->NewDataIterator(comparator_, kDisableGlobalSequenceNumber));
|
|
|
|
}
|
|
|
|
|
|
|
|
private:
|
|
|
|
const Comparator* comparator_;
|
|
|
|
std::string data_;
|
|
|
|
Block* block_;
|
|
|
|
|
|
|
|
BlockConstructor();
|
|
|
|
};
|
|
|
|
|
|
|
|
class TableConstructor : public Constructor {
|
2013-12-20 18:35:24 +01:00
|
|
|
public:
|
2014-01-28 19:35:48 +01:00
|
|
|
explicit TableConstructor(const Comparator* cmp,
|
2018-04-10 01:17:15 +02:00
|
|
|
bool convert_to_internal_key = false,
|
Add an option to put first key of each sst block in the index (#5289)
Summary:
The first key is used to defer reading the data block until this file gets to the top of merging iterator's heap. For short range scans, most files never make it to the top of the heap, so this change can reduce read amplification by a lot sometimes.
Consider the following workload. There are a few data streams (we'll be calling them "logs"), each stream consisting of a sequence of blobs (we'll be calling them "records"). Each record is identified by log ID and a sequence number within the log. RocksDB key is concatenation of log ID and sequence number (big endian). Reads are mostly relatively short range scans, each within a single log. Writes are mostly sequential for each log, but writes to different logs are randomly interleaved. Compactions are disabled; instead, when we accumulate a few tens of sst files, we create a new column family and start writing to it.
So, a typical sst file consists of a few ranges of blocks, each range corresponding to one log ID (we use FlushBlockPolicy to cut blocks at log boundaries). A typical read would go like this. First, iterator Seek() reads one block from each sst file. Then a series of Next()s move through one sst file (since writes to each log are mostly sequential) until the subiterator reaches the end of this log in this sst file; then Next() switches to the next sst file and reads sequentially from that, and so on. Often a range scan will only return records from a small number of blocks in small number of sst files; in this case, the cost of initial Seek() reading one block from each file may be bigger than the cost of reading the actually useful blocks.
Neither iterate_upper_bound nor bloom filters can prevent reading one block from each file in Seek(). But this PR can: if the index contains first key from each block, we don't have to read the block until this block actually makes it to the top of merging iterator's heap, so for short range scans we won't read any blocks from most of the sst files.
This PR does the deferred block loading inside value() call. This is not ideal: there's no good way to report an IO error from inside value(). As discussed with siying offline, it would probably be better to change InternalIterator's interface to explicitly fetch deferred value and get status. I'll do it in a separate PR.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/5289
Differential Revision: D15256423
Pulled By: al13n321
fbshipit-source-id: 750e4c39ce88e8d41662f701cf6275d9388ba46a
2019-06-25 05:50:35 +02:00
|
|
|
int level = -1, SequenceNumber largest_seqno = 0)
|
2014-02-08 01:25:38 +01:00
|
|
|
: Constructor(cmp),
|
Add an option to put first key of each sst block in the index (#5289)
Summary:
The first key is used to defer reading the data block until this file gets to the top of merging iterator's heap. For short range scans, most files never make it to the top of the heap, so this change can reduce read amplification by a lot sometimes.
Consider the following workload. There are a few data streams (we'll be calling them "logs"), each stream consisting of a sequence of blobs (we'll be calling them "records"). Each record is identified by log ID and a sequence number within the log. RocksDB key is concatenation of log ID and sequence number (big endian). Reads are mostly relatively short range scans, each within a single log. Writes are mostly sequential for each log, but writes to different logs are randomly interleaved. Compactions are disabled; instead, when we accumulate a few tens of sst files, we create a new column family and start writing to it.
So, a typical sst file consists of a few ranges of blocks, each range corresponding to one log ID (we use FlushBlockPolicy to cut blocks at log boundaries). A typical read would go like this. First, iterator Seek() reads one block from each sst file. Then a series of Next()s move through one sst file (since writes to each log are mostly sequential) until the subiterator reaches the end of this log in this sst file; then Next() switches to the next sst file and reads sequentially from that, and so on. Often a range scan will only return records from a small number of blocks in small number of sst files; in this case, the cost of initial Seek() reading one block from each file may be bigger than the cost of reading the actually useful blocks.
Neither iterate_upper_bound nor bloom filters can prevent reading one block from each file in Seek(). But this PR can: if the index contains first key from each block, we don't have to read the block until this block actually makes it to the top of merging iterator's heap, so for short range scans we won't read any blocks from most of the sst files.
This PR does the deferred block loading inside value() call. This is not ideal: there's no good way to report an IO error from inside value(). As discussed with siying offline, it would probably be better to change InternalIterator's interface to explicitly fetch deferred value and get status. I'll do it in a separate PR.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/5289
Differential Revision: D15256423
Pulled By: al13n321
fbshipit-source-id: 750e4c39ce88e8d41662f701cf6275d9388ba46a
2019-06-25 05:50:35 +02:00
|
|
|
largest_seqno_(largest_seqno),
|
2018-04-10 01:17:15 +02:00
|
|
|
convert_to_internal_key_(convert_to_internal_key),
|
2019-07-17 22:02:00 +02:00
|
|
|
level_(level) {
|
2020-02-20 21:07:53 +01:00
|
|
|
env_ = ROCKSDB_NAMESPACE::Env::Default();
|
2019-07-17 22:02:00 +02:00
|
|
|
}
|
2019-02-14 22:52:47 +01:00
|
|
|
~TableConstructor() override { Reset(); }
|
2014-01-24 19:57:15 +01:00
|
|
|
|
2019-02-14 22:52:47 +01:00
|
|
|
Status FinishImpl(const Options& options, const ImmutableCFOptions& ioptions,
|
|
|
|
const MutableCFOptions& moptions,
|
|
|
|
const BlockBasedTableOptions& /*table_options*/,
|
|
|
|
const InternalKeyComparator& internal_comparator,
|
|
|
|
const stl_wrappers::KVMap& kv_map) override {
|
2011-03-18 23:37:00 +01:00
|
|
|
Reset();
|
2015-09-17 01:57:43 +02:00
|
|
|
soptions.use_mmap_reads = ioptions.allow_mmap_reads;
|
2021-01-05 00:59:52 +01:00
|
|
|
std::unique_ptr<FSWritableFile> sink(new test::StringSink());
|
|
|
|
file_writer_.reset(new WritableFileWriter(
|
|
|
|
std::move(sink), "" /* don't care */, FileOptions()));
|
2018-11-09 20:17:34 +01:00
|
|
|
std::unique_ptr<TableBuilder> builder;
|
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 19:04:30 +02:00
|
|
|
std::vector<std::unique_ptr<IntTblPropCollectorFactory>>
|
|
|
|
int_tbl_prop_collector_factories;
|
Add an option to put first key of each sst block in the index (#5289)
Summary:
The first key is used to defer reading the data block until this file gets to the top of merging iterator's heap. For short range scans, most files never make it to the top of the heap, so this change can reduce read amplification by a lot sometimes.
Consider the following workload. There are a few data streams (we'll be calling them "logs"), each stream consisting of a sequence of blobs (we'll be calling them "records"). Each record is identified by log ID and a sequence number within the log. RocksDB key is concatenation of log ID and sequence number (big endian). Reads are mostly relatively short range scans, each within a single log. Writes are mostly sequential for each log, but writes to different logs are randomly interleaved. Compactions are disabled; instead, when we accumulate a few tens of sst files, we create a new column family and start writing to it.
So, a typical sst file consists of a few ranges of blocks, each range corresponding to one log ID (we use FlushBlockPolicy to cut blocks at log boundaries). A typical read would go like this. First, iterator Seek() reads one block from each sst file. Then a series of Next()s move through one sst file (since writes to each log are mostly sequential) until the subiterator reaches the end of this log in this sst file; then Next() switches to the next sst file and reads sequentially from that, and so on. Often a range scan will only return records from a small number of blocks in small number of sst files; in this case, the cost of initial Seek() reading one block from each file may be bigger than the cost of reading the actually useful blocks.
Neither iterate_upper_bound nor bloom filters can prevent reading one block from each file in Seek(). But this PR can: if the index contains first key from each block, we don't have to read the block until this block actually makes it to the top of merging iterator's heap, so for short range scans we won't read any blocks from most of the sst files.
This PR does the deferred block loading inside value() call. This is not ideal: there's no good way to report an IO error from inside value(). As discussed with siying offline, it would probably be better to change InternalIterator's interface to explicitly fetch deferred value and get status. I'll do it in a separate PR.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/5289
Differential Revision: D15256423
Pulled By: al13n321
fbshipit-source-id: 750e4c39ce88e8d41662f701cf6275d9388ba46a
2019-06-25 05:50:35 +02:00
|
|
|
|
|
|
|
if (largest_seqno_ != 0) {
|
|
|
|
// Pretend that it's an external file written by SstFileWriter.
|
|
|
|
int_tbl_prop_collector_factories.emplace_back(
|
|
|
|
new SstFileWriterPropertiesCollectorFactory(2 /* version */,
|
|
|
|
0 /* global_seqno*/));
|
|
|
|
}
|
|
|
|
|
2016-04-07 08:10:32 +02:00
|
|
|
std::string column_family_name;
|
2014-09-05 01:18:36 +02:00
|
|
|
builder.reset(ioptions.table_factory->NewTableBuilder(
|
Reduce scope of compression dictionary to single SST (#4952)
Summary:
Our previous approach was to train one compression dictionary per compaction, using the first output SST to train a dictionary, and then applying it on subsequent SSTs in the same compaction. While this was great for minimizing CPU/memory/I/O overhead, it did not achieve good compression ratios in practice. In our most promising potential use case, moderate reductions in a dictionary's scope make a major difference on compression ratio.
So, this PR changes compression dictionary to be scoped per-SST. It accepts the tradeoff during table building to use more memory and CPU. Important changes include:
- The `BlockBasedTableBuilder` has a new state when dictionary compression is in-use: `kBuffered`. In that state it accumulates uncompressed data in-memory whenever `Add` is called.
- After accumulating target file size bytes or calling `BlockBasedTableBuilder::Finish`, a `BlockBasedTableBuilder` moves to the `kUnbuffered` state. The transition (`EnterUnbuffered()`) involves sampling the buffered data, training a dictionary, and compressing/writing out all buffered data. In the `kUnbuffered` state, a `BlockBasedTableBuilder` behaves the same as before -- blocks are compressed/written out as soon as they fill up.
- Samples are now whole uncompressed data blocks, except the final sample may be a partial data block so we don't breach the user's configured `max_dict_bytes` or `zstd_max_train_bytes`. The dictionary trainer is supposed to work better when we pass it real units of compression. Previously we were passing 64-byte KV samples which was not realistic.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4952
Differential Revision: D13967980
Pulled By: ajkr
fbshipit-source-id: 82bea6f7537e1529c7a1a4cdee84585f5949300f
2019-02-12 04:42:25 +01:00
|
|
|
TableBuilderOptions(ioptions, moptions, internal_comparator,
|
|
|
|
&int_tbl_prop_collector_factories,
|
2019-03-18 20:07:35 +01:00
|
|
|
options.compression, options.sample_for_compression,
|
2019-04-05 01:18:07 +02:00
|
|
|
options.compression_opts, false /* skip_filters */,
|
2019-03-18 20:07:35 +01:00
|
|
|
column_family_name, level_),
|
2015-10-09 01:57:35 +02:00
|
|
|
TablePropertiesCollectorFactory::Context::kUnknownColumnFamily,
|
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
2015-07-18 01:16:11 +02:00
|
|
|
file_writer_.get()));
|
2011-03-18 23:37:00 +01:00
|
|
|
|
2020-06-25 01:20:55 +02:00
|
|
|
for (const auto& kv : kv_map) {
|
2013-12-20 18:35:24 +01:00
|
|
|
if (convert_to_internal_key_) {
|
2014-11-06 20:14:28 +01:00
|
|
|
ParsedInternalKey ikey(kv.first, kMaxSequenceNumber, kTypeValue);
|
2013-12-20 18:35:24 +01:00
|
|
|
std::string encoded;
|
|
|
|
AppendInternalKey(&encoded, ikey);
|
2014-11-06 20:14:28 +01:00
|
|
|
builder->Add(encoded, kv.second);
|
2013-12-20 18:35:24 +01:00
|
|
|
} else {
|
2014-11-06 20:14:28 +01:00
|
|
|
builder->Add(kv.first, kv.second);
|
2013-12-20 18:35:24 +01:00
|
|
|
}
|
2020-12-24 01:54:05 +01:00
|
|
|
EXPECT_OK(builder->status());
|
2011-03-18 23:37:00 +01:00
|
|
|
}
|
2013-12-20 18:35:24 +01:00
|
|
|
Status s = builder->Finish();
|
2020-12-24 01:54:05 +01:00
|
|
|
EXPECT_OK(file_writer_->Flush());
|
rocksdb: Replace ASSERT* with EXPECT* in functions that does not return void value
Summary:
gtest does not use exceptions to fail a unit test by design, and `ASSERT*`s are implemented using `return`. As a consequence we cannot use `ASSERT*` in a function that does not return `void` value ([[ https://code.google.com/p/googletest/wiki/AdvancedGuide#Assertion_Placement | 1]]), and have to fix our existing code. This diff does this in a generic way, with no manual changes.
In order to detect all existing `ASSERT*` that are used in functions that doesn't return void value, I change the code to generate compile errors for such cases.
In `util/testharness.h` I defined `EXPECT*` assertions, the same way as `ASSERT*`, and redefined `ASSERT*` to return `void`. Then executed:
```lang=bash
% USE_CLANG=1 make all -j55 -k 2> build.log
% perl -naF: -e 'print "-- -number=".$F[1]." ".$F[0]."\n" if /: error:/' \
build.log | xargs -L 1 perl -spi -e 's/ASSERT/EXPECT/g if $. == $number'
% make format
```
After that I reverted back change to `ASSERT*` in `util/testharness.h`. But preserved introduced `EXPECT*`, which is the same as `ASSERT*`. This will be deleted once switched to gtest.
This diff is independent and contains manual changes only in `util/testharness.h`.
Test Plan:
Make sure all tests are passing.
```lang=bash
% USE_CLANG=1 make check
```
Reviewers: igor, lgalanis, sdong, yufei.zhu, rven, meyering
Reviewed By: meyering
Subscribers: dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D33333
2015-03-17 04:52:32 +01:00
|
|
|
EXPECT_TRUE(s.ok()) << s.ToString();
|
2011-03-18 23:37:00 +01:00
|
|
|
|
2018-07-20 18:00:33 +02:00
|
|
|
EXPECT_EQ(TEST_GetSink()->contents().size(), builder->FileSize());
|
2011-03-18 23:37:00 +01:00
|
|
|
|
|
|
|
// Open the table
|
2013-02-01 00:20:24 +01:00
|
|
|
uniq_id_ = cur_uniq_id_++;
|
2021-01-05 00:59:52 +01:00
|
|
|
std::unique_ptr<FSRandomAccessFile> source(new test::StringSource(
|
|
|
|
TEST_GetSink()->contents(), uniq_id_, ioptions.allow_mmap_reads));
|
|
|
|
|
|
|
|
file_reader_.reset(new RandomAccessFileReader(std::move(source), "test"));
|
2018-06-28 02:09:29 +02:00
|
|
|
const bool kSkipFilters = true;
|
|
|
|
const bool kImmortal = true;
|
2014-09-05 01:18:36 +02:00
|
|
|
return ioptions.table_factory->NewTableReader(
|
2018-05-21 23:33:55 +02:00
|
|
|
TableReaderOptions(ioptions, moptions.prefix_extractor.get(), soptions,
|
2018-06-28 02:09:29 +02:00
|
|
|
internal_comparator, !kSkipFilters, !kImmortal,
|
2020-06-10 01:49:07 +02:00
|
|
|
false, level_, largest_seqno_, &block_cache_tracer_,
|
|
|
|
moptions.write_buffer_size),
|
2018-07-20 23:34:07 +02:00
|
|
|
std::move(file_reader_), TEST_GetSink()->contents().size(),
|
|
|
|
&table_reader_);
|
2011-03-18 23:37:00 +01:00
|
|
|
}
|
|
|
|
|
2019-02-14 22:52:47 +01:00
|
|
|
InternalIterator* NewIterator(
|
2018-05-21 23:33:55 +02:00
|
|
|
const SliceTransform* prefix_extractor) const override {
|
2019-06-20 23:28:22 +02:00
|
|
|
InternalIterator* iter = table_reader_->NewIterator(
|
2020-08-04 00:21:56 +02:00
|
|
|
read_options_, prefix_extractor, /*arena=*/nullptr,
|
|
|
|
/*skip_filters=*/false, TableReaderCaller::kUncategorized);
|
2013-12-20 18:35:24 +01:00
|
|
|
if (convert_to_internal_key_) {
|
|
|
|
return new KeyConvertingIterator(iter);
|
|
|
|
} else {
|
|
|
|
return iter;
|
|
|
|
}
|
2011-03-18 23:37:00 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
uint64_t ApproximateOffsetOf(const Slice& key) const {
|
2016-08-20 00:10:31 +02:00
|
|
|
if (convert_to_internal_key_) {
|
|
|
|
InternalKey ikey(key, kMaxSequenceNumber, kTypeValue);
|
|
|
|
const Slice skey = ikey.Encode();
|
2019-06-20 23:28:22 +02:00
|
|
|
return table_reader_->ApproximateOffsetOf(
|
|
|
|
skey, TableReaderCaller::kUncategorized);
|
2016-08-20 00:10:31 +02:00
|
|
|
}
|
2019-06-20 23:28:22 +02:00
|
|
|
return table_reader_->ApproximateOffsetOf(
|
|
|
|
key, TableReaderCaller::kUncategorized);
|
2011-03-18 23:37:00 +01:00
|
|
|
}
|
|
|
|
|
2018-05-21 23:33:55 +02:00
|
|
|
virtual Status Reopen(const ImmutableCFOptions& ioptions,
|
|
|
|
const MutableCFOptions& moptions) {
|
2021-01-05 00:59:52 +01:00
|
|
|
std::unique_ptr<FSRandomAccessFile> source(new test::StringSource(
|
|
|
|
TEST_GetSink()->contents(), uniq_id_, ioptions.allow_mmap_reads));
|
|
|
|
|
|
|
|
file_reader_.reset(new RandomAccessFileReader(std::move(source), "test"));
|
2014-09-05 01:18:36 +02:00
|
|
|
return ioptions.table_factory->NewTableReader(
|
2018-05-21 23:33:55 +02:00
|
|
|
TableReaderOptions(ioptions, moptions.prefix_extractor.get(), soptions,
|
|
|
|
*last_internal_key_),
|
2018-07-20 23:34:07 +02:00
|
|
|
std::move(file_reader_), TEST_GetSink()->contents().size(),
|
|
|
|
&table_reader_);
|
2013-02-01 00:20:24 +01:00
|
|
|
}
|
|
|
|
|
2018-07-20 23:34:07 +02:00
|
|
|
virtual TableReader* GetTableReader() { return table_reader_.get(); }
|
2013-02-01 00:20:24 +01:00
|
|
|
|
2019-02-14 22:52:47 +01:00
|
|
|
bool AnywayDeleteIterator() const override {
|
2014-09-05 02:40:41 +02:00
|
|
|
return convert_to_internal_key_;
|
|
|
|
}
|
|
|
|
|
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 19:42:39 +02:00
|
|
|
void ResetTableReader() { table_reader_.reset(); }
|
|
|
|
|
2016-08-20 00:10:31 +02:00
|
|
|
bool ConvertToInternalKey() { return convert_to_internal_key_; }
|
|
|
|
|
2018-07-20 18:00:33 +02:00
|
|
|
test::StringSink* TEST_GetSink() {
|
2021-01-05 00:59:52 +01:00
|
|
|
return static_cast<test::StringSink*>(file_writer_->writable_file());
|
2018-07-20 18:00:33 +02:00
|
|
|
}
|
|
|
|
|
2019-07-17 22:02:00 +02:00
|
|
|
BlockCacheTracer block_cache_tracer_;
|
|
|
|
|
2011-03-18 23:37:00 +01:00
|
|
|
private:
|
|
|
|
void Reset() {
|
2013-02-01 00:20:24 +01:00
|
|
|
uniq_id_ = 0;
|
2013-10-30 18:52:33 +01:00
|
|
|
table_reader_.reset();
|
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
2015-07-18 01:16:11 +02:00
|
|
|
file_writer_.reset();
|
|
|
|
file_reader_.reset();
|
|
|
|
}
|
|
|
|
|
2020-08-04 00:21:56 +02:00
|
|
|
const ReadOptions read_options_;
|
2013-02-01 00:20:24 +01:00
|
|
|
uint64_t uniq_id_;
|
2018-11-09 20:17:34 +01:00
|
|
|
std::unique_ptr<WritableFileWriter> file_writer_;
|
|
|
|
std::unique_ptr<RandomAccessFileReader> file_reader_;
|
|
|
|
std::unique_ptr<TableReader> table_reader_;
|
Add an option to put first key of each sst block in the index (#5289)
Summary:
The first key is used to defer reading the data block until this file gets to the top of merging iterator's heap. For short range scans, most files never make it to the top of the heap, so this change can reduce read amplification by a lot sometimes.
Consider the following workload. There are a few data streams (we'll be calling them "logs"), each stream consisting of a sequence of blobs (we'll be calling them "records"). Each record is identified by log ID and a sequence number within the log. RocksDB key is concatenation of log ID and sequence number (big endian). Reads are mostly relatively short range scans, each within a single log. Writes are mostly sequential for each log, but writes to different logs are randomly interleaved. Compactions are disabled; instead, when we accumulate a few tens of sst files, we create a new column family and start writing to it.
So, a typical sst file consists of a few ranges of blocks, each range corresponding to one log ID (we use FlushBlockPolicy to cut blocks at log boundaries). A typical read would go like this. First, iterator Seek() reads one block from each sst file. Then a series of Next()s move through one sst file (since writes to each log are mostly sequential) until the subiterator reaches the end of this log in this sst file; then Next() switches to the next sst file and reads sequentially from that, and so on. Often a range scan will only return records from a small number of blocks in small number of sst files; in this case, the cost of initial Seek() reading one block from each file may be bigger than the cost of reading the actually useful blocks.
Neither iterate_upper_bound nor bloom filters can prevent reading one block from each file in Seek(). But this PR can: if the index contains first key from each block, we don't have to read the block until this block actually makes it to the top of merging iterator's heap, so for short range scans we won't read any blocks from most of the sst files.
This PR does the deferred block loading inside value() call. This is not ideal: there's no good way to report an IO error from inside value(). As discussed with siying offline, it would probably be better to change InternalIterator's interface to explicitly fetch deferred value and get status. I'll do it in a separate PR.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/5289
Differential Revision: D15256423
Pulled By: al13n321
fbshipit-source-id: 750e4c39ce88e8d41662f701cf6275d9388ba46a
2019-06-25 05:50:35 +02:00
|
|
|
SequenceNumber largest_seqno_;
|
2014-09-05 02:40:41 +02:00
|
|
|
bool convert_to_internal_key_;
|
2018-04-10 01:17:15 +02:00
|
|
|
int level_;
|
2011-03-18 23:37:00 +01:00
|
|
|
|
2013-12-20 18:35:24 +01:00
|
|
|
TableConstructor();
|
2013-02-01 00:20:24 +01:00
|
|
|
|
|
|
|
static uint64_t cur_uniq_id_;
|
2015-09-17 01:57:43 +02:00
|
|
|
EnvOptions soptions;
|
2019-07-17 22:02:00 +02:00
|
|
|
Env* env_;
|
2011-03-18 23:37:00 +01:00
|
|
|
};
|
2013-12-20 18:35:24 +01:00
|
|
|
uint64_t TableConstructor::cur_uniq_id_ = 1;
|
2011-03-18 23:37:00 +01:00
|
|
|
|
|
|
|
class MemTableConstructor: public Constructor {
|
|
|
|
public:
|
2016-06-21 03:01:03 +02:00
|
|
|
explicit MemTableConstructor(const Comparator* cmp, WriteBufferManager* wb)
|
2011-03-18 23:37:00 +01:00
|
|
|
: Constructor(cmp),
|
2013-07-23 23:42:27 +02:00
|
|
|
internal_comparator_(cmp),
|
2016-06-21 03:01:03 +02:00
|
|
|
write_buffer_manager_(wb),
|
2013-07-23 23:42:27 +02:00
|
|
|
table_factory_(new SkipListFactory) {
|
2014-12-02 21:09:20 +01:00
|
|
|
options_.memtable_factory = table_factory_;
|
|
|
|
ImmutableCFOptions ioptions(options_);
|
2016-09-14 06:11:59 +02:00
|
|
|
memtable_ =
|
|
|
|
new MemTable(internal_comparator_, ioptions, MutableCFOptions(options_),
|
2017-06-02 21:08:01 +02:00
|
|
|
wb, kMaxSequenceNumber, 0 /* column_family_id */);
|
2011-05-21 04:17:43 +02:00
|
|
|
memtable_->Ref();
|
2011-03-18 23:37:00 +01:00
|
|
|
}
|
2019-02-14 22:52:47 +01:00
|
|
|
~MemTableConstructor() override { delete memtable_->Unref(); }
|
|
|
|
Status FinishImpl(const Options&, const ImmutableCFOptions& ioptions,
|
|
|
|
const MutableCFOptions& /*moptions*/,
|
|
|
|
const BlockBasedTableOptions& /*table_options*/,
|
|
|
|
const InternalKeyComparator& /*internal_comparator*/,
|
|
|
|
const stl_wrappers::KVMap& kv_map) override {
|
2013-12-02 06:23:44 +01:00
|
|
|
delete memtable_->Unref();
|
2014-12-02 21:09:20 +01:00
|
|
|
ImmutableCFOptions mem_ioptions(ioptions);
|
2014-10-02 01:19:16 +02:00
|
|
|
memtable_ = new MemTable(internal_comparator_, mem_ioptions,
|
2016-09-14 06:11:59 +02:00
|
|
|
MutableCFOptions(options_), write_buffer_manager_,
|
2017-06-02 21:08:01 +02:00
|
|
|
kMaxSequenceNumber, 0 /* column_family_id */);
|
2011-05-21 04:17:43 +02:00
|
|
|
memtable_->Ref();
|
2011-03-18 23:37:00 +01:00
|
|
|
int seq = 1;
|
2020-06-25 01:20:55 +02:00
|
|
|
for (const auto& kv : kv_map) {
|
2020-11-24 01:27:46 +01:00
|
|
|
Status s = memtable_->Add(seq, kTypeValue, kv.first, kv.second);
|
|
|
|
if (!s.ok()) {
|
|
|
|
return s;
|
|
|
|
}
|
2011-03-18 23:37:00 +01:00
|
|
|
seq++;
|
|
|
|
}
|
|
|
|
return Status::OK();
|
|
|
|
}
|
2019-02-14 22:52:47 +01:00
|
|
|
InternalIterator* NewIterator(
|
2018-05-21 23:33:55 +02:00
|
|
|
const SliceTransform* /*prefix_extractor*/) const override {
|
2014-09-05 02:40:41 +02:00
|
|
|
return new KeyConvertingIterator(
|
|
|
|
memtable_->NewIterator(ReadOptions(), &arena_), true);
|
2011-03-18 23:37:00 +01:00
|
|
|
}
|
|
|
|
|
2019-02-14 22:52:47 +01:00
|
|
|
bool AnywayDeleteIterator() const override { return true; }
|
2014-09-05 02:40:41 +02:00
|
|
|
|
2019-02-14 22:52:47 +01:00
|
|
|
bool IsArenaMode() const override { return true; }
|
2014-09-05 02:40:41 +02:00
|
|
|
|
2011-03-18 23:37:00 +01:00
|
|
|
private:
|
2014-09-05 02:40:41 +02:00
|
|
|
mutable Arena arena_;
|
2011-03-18 23:37:00 +01:00
|
|
|
InternalKeyComparator internal_comparator_;
|
2014-12-02 21:09:20 +01:00
|
|
|
Options options_;
|
2016-06-21 03:01:03 +02:00
|
|
|
WriteBufferManager* write_buffer_manager_;
|
2011-03-18 23:37:00 +01:00
|
|
|
MemTable* memtable_;
|
2013-07-23 23:42:27 +02:00
|
|
|
std::shared_ptr<SkipListFactory> table_factory_;
|
2011-03-18 23:37:00 +01:00
|
|
|
};
|
|
|
|
|
2015-10-13 00:06:38 +02:00
|
|
|
class InternalIteratorFromIterator : public InternalIterator {
|
|
|
|
public:
|
|
|
|
explicit InternalIteratorFromIterator(Iterator* it) : it_(it) {}
|
2019-02-14 22:52:47 +01:00
|
|
|
bool Valid() const override { return it_->Valid(); }
|
|
|
|
void Seek(const Slice& target) override { it_->Seek(target); }
|
|
|
|
void SeekForPrev(const Slice& target) override { it_->SeekForPrev(target); }
|
|
|
|
void SeekToFirst() override { it_->SeekToFirst(); }
|
|
|
|
void SeekToLast() override { it_->SeekToLast(); }
|
|
|
|
void Next() override { it_->Next(); }
|
|
|
|
void Prev() override { it_->Prev(); }
|
2015-10-13 00:06:38 +02:00
|
|
|
Slice key() const override { return it_->key(); }
|
|
|
|
Slice value() const override { return it_->value(); }
|
2019-02-14 22:52:47 +01:00
|
|
|
Status status() const override { return it_->status(); }
|
2015-10-13 00:06:38 +02:00
|
|
|
|
|
|
|
private:
|
2018-11-09 20:17:34 +01:00
|
|
|
std::unique_ptr<Iterator> it_;
|
2015-10-13 00:06:38 +02:00
|
|
|
};
|
|
|
|
|
2011-03-18 23:37:00 +01:00
|
|
|
class DBConstructor: public Constructor {
|
|
|
|
public:
|
|
|
|
explicit DBConstructor(const Comparator* cmp)
|
|
|
|
: Constructor(cmp),
|
|
|
|
comparator_(cmp) {
|
2013-03-01 03:04:58 +01:00
|
|
|
db_ = nullptr;
|
2011-03-18 23:37:00 +01:00
|
|
|
NewDB();
|
|
|
|
}
|
2019-02-14 22:52:47 +01:00
|
|
|
~DBConstructor() override { delete db_; }
|
|
|
|
Status FinishImpl(const Options& /*options*/,
|
|
|
|
const ImmutableCFOptions& /*ioptions*/,
|
|
|
|
const MutableCFOptions& /*moptions*/,
|
|
|
|
const BlockBasedTableOptions& /*table_options*/,
|
|
|
|
const InternalKeyComparator& /*internal_comparator*/,
|
|
|
|
const stl_wrappers::KVMap& kv_map) override {
|
2011-03-18 23:37:00 +01:00
|
|
|
delete db_;
|
2013-03-01 03:04:58 +01:00
|
|
|
db_ = nullptr;
|
2011-03-18 23:37:00 +01:00
|
|
|
NewDB();
|
2020-06-25 01:20:55 +02:00
|
|
|
for (const auto& kv : kv_map) {
|
2011-03-18 23:37:00 +01:00
|
|
|
WriteBatch batch;
|
2020-10-08 20:11:52 +02:00
|
|
|
EXPECT_OK(batch.Put(kv.first, kv.second));
|
rocksdb: Replace ASSERT* with EXPECT* in functions that does not return void value
Summary:
gtest does not use exceptions to fail a unit test by design, and `ASSERT*`s are implemented using `return`. As a consequence we cannot use `ASSERT*` in a function that does not return `void` value ([[ https://code.google.com/p/googletest/wiki/AdvancedGuide#Assertion_Placement | 1]]), and have to fix our existing code. This diff does this in a generic way, with no manual changes.
In order to detect all existing `ASSERT*` that are used in functions that doesn't return void value, I change the code to generate compile errors for such cases.
In `util/testharness.h` I defined `EXPECT*` assertions, the same way as `ASSERT*`, and redefined `ASSERT*` to return `void`. Then executed:
```lang=bash
% USE_CLANG=1 make all -j55 -k 2> build.log
% perl -naF: -e 'print "-- -number=".$F[1]." ".$F[0]."\n" if /: error:/' \
build.log | xargs -L 1 perl -spi -e 's/ASSERT/EXPECT/g if $. == $number'
% make format
```
After that I reverted back change to `ASSERT*` in `util/testharness.h`. But preserved introduced `EXPECT*`, which is the same as `ASSERT*`. This will be deleted once switched to gtest.
This diff is independent and contains manual changes only in `util/testharness.h`.
Test Plan:
Make sure all tests are passing.
```lang=bash
% USE_CLANG=1 make check
```
Reviewers: igor, lgalanis, sdong, yufei.zhu, rven, meyering
Reviewed By: meyering
Subscribers: dhruba, leveldb
Differential Revision: https://reviews.facebook.net/D33333
2015-03-17 04:52:32 +01:00
|
|
|
EXPECT_TRUE(db_->Write(WriteOptions(), &batch).ok());
|
2011-03-18 23:37:00 +01:00
|
|
|
}
|
|
|
|
return Status::OK();
|
|
|
|
}
|
2015-10-13 00:06:38 +02:00
|
|
|
|
2019-02-14 22:52:47 +01:00
|
|
|
InternalIterator* NewIterator(
|
2018-05-21 23:33:55 +02:00
|
|
|
const SliceTransform* /*prefix_extractor*/) const override {
|
2015-10-13 00:06:38 +02:00
|
|
|
return new InternalIteratorFromIterator(db_->NewIterator(ReadOptions()));
|
2011-03-18 23:37:00 +01:00
|
|
|
}
|
|
|
|
|
2019-02-14 22:52:47 +01:00
|
|
|
DB* db() const override { return db_; }
|
2011-03-21 20:40:57 +01:00
|
|
|
|
2011-03-18 23:37:00 +01:00
|
|
|
private:
|
|
|
|
void NewDB() {
|
2018-07-14 02:18:39 +02:00
|
|
|
std::string name = test::PerThreadDBPath("table_testdb");
|
2011-03-18 23:37:00 +01:00
|
|
|
|
2013-11-20 07:00:48 +01:00
|
|
|
Options options;
|
2011-03-18 23:37:00 +01:00
|
|
|
options.comparator = comparator_;
|
|
|
|
Status status = DestroyDB(name, options);
|
|
|
|
ASSERT_TRUE(status.ok()) << status.ToString();
|
|
|
|
|
|
|
|
options.create_if_missing = true;
|
|
|
|
options.error_if_exists = true;
|
2011-03-21 20:40:57 +01:00
|
|
|
options.write_buffer_size = 10000; // Something small to force merging
|
2011-03-18 23:37:00 +01:00
|
|
|
status = DB::Open(options, name, &db_);
|
|
|
|
ASSERT_TRUE(status.ok()) << status.ToString();
|
|
|
|
}
|
|
|
|
|
|
|
|
const Comparator* comparator_;
|
|
|
|
DB* db_;
|
|
|
|
};
|
|
|
|
|
|
|
|
enum TestType {
|
2013-12-20 18:35:24 +01:00
|
|
|
BLOCK_BASED_TABLE_TEST,
|
2015-07-20 20:09:14 +02:00
|
|
|
#ifndef ROCKSDB_LITE
|
2013-12-20 18:35:24 +01:00
|
|
|
PLAIN_TABLE_SEMI_FIXED_PREFIX,
|
|
|
|
PLAIN_TABLE_FULL_STR_PREFIX,
|
2014-02-08 01:25:38 +01:00
|
|
|
PLAIN_TABLE_TOTAL_ORDER,
|
2015-07-20 20:09:14 +02:00
|
|
|
#endif // !ROCKSDB_LITE
|
2011-03-18 23:37:00 +01:00
|
|
|
BLOCK_TEST,
|
|
|
|
MEMTABLE_TEST,
|
2011-07-20 01:36:47 +02:00
|
|
|
DB_TEST
|
2011-03-18 23:37:00 +01:00
|
|
|
};
|
|
|
|
|
|
|
|
struct TestArgs {
|
|
|
|
TestType type;
|
|
|
|
bool reverse_compare;
|
|
|
|
int restart_interval;
|
2012-06-28 08:41:33 +02:00
|
|
|
CompressionType compression;
|
2020-04-02 01:37:54 +02:00
|
|
|
uint32_t compression_parallel_threads;
|
2015-01-15 01:24:24 +01:00
|
|
|
uint32_t format_version;
|
2015-09-17 01:57:43 +02:00
|
|
|
bool use_mmap;
|
2011-03-18 23:37:00 +01:00
|
|
|
};
|
|
|
|
|
2020-06-15 23:06:47 +02:00
|
|
|
std::ostream& operator<<(std::ostream& os, const TestArgs& args) {
|
|
|
|
os << "type: " << args.type << " reverse_compare: " << args.reverse_compare
|
|
|
|
<< " restart_interval: " << args.restart_interval
|
|
|
|
<< " compression: " << args.compression
|
|
|
|
<< " compression_parallel_threads: " << args.compression_parallel_threads
|
|
|
|
<< " format_version: " << args.format_version
|
|
|
|
<< " use_mmap: " << args.use_mmap;
|
|
|
|
|
|
|
|
return os;
|
|
|
|
}
|
|
|
|
|
2013-11-20 07:00:48 +01:00
|
|
|
static std::vector<TestArgs> GenerateArgList() {
|
2014-01-24 20:09:04 +01:00
|
|
|
std::vector<TestArgs> test_args;
|
|
|
|
std::vector<TestType> test_types = {
|
2015-07-20 20:09:14 +02:00
|
|
|
BLOCK_BASED_TABLE_TEST,
|
|
|
|
#ifndef ROCKSDB_LITE
|
|
|
|
PLAIN_TABLE_SEMI_FIXED_PREFIX,
|
|
|
|
PLAIN_TABLE_FULL_STR_PREFIX,
|
|
|
|
PLAIN_TABLE_TOTAL_ORDER,
|
|
|
|
#endif // !ROCKSDB_LITE
|
|
|
|
BLOCK_TEST,
|
|
|
|
MEMTABLE_TEST, DB_TEST};
|
2014-01-24 20:09:04 +01:00
|
|
|
std::vector<bool> reverse_compare_types = {false, true};
|
|
|
|
std::vector<int> restart_intervals = {16, 1, 1024};
|
2020-04-02 01:37:54 +02:00
|
|
|
std::vector<uint32_t> compression_parallel_threads = {1, 4};
|
2012-06-28 08:41:33 +02:00
|
|
|
|
|
|
|
// Only add compression if it is supported
|
2015-01-15 01:24:24 +01:00
|
|
|
std::vector<std::pair<CompressionType, bool>> compression_types;
|
|
|
|
compression_types.emplace_back(kNoCompression, false);
|
2015-04-06 21:50:44 +02:00
|
|
|
if (Snappy_Supported()) {
|
2015-01-15 01:24:24 +01:00
|
|
|
compression_types.emplace_back(kSnappyCompression, false);
|
2014-01-24 20:09:04 +01:00
|
|
|
}
|
2015-04-06 21:50:44 +02:00
|
|
|
if (Zlib_Supported()) {
|
2015-01-15 01:24:24 +01:00
|
|
|
compression_types.emplace_back(kZlibCompression, false);
|
|
|
|
compression_types.emplace_back(kZlibCompression, true);
|
2014-01-24 20:09:04 +01:00
|
|
|
}
|
2015-04-06 21:50:44 +02:00
|
|
|
if (BZip2_Supported()) {
|
2015-01-15 01:24:24 +01:00
|
|
|
compression_types.emplace_back(kBZip2Compression, false);
|
|
|
|
compression_types.emplace_back(kBZip2Compression, true);
|
2014-01-24 20:09:04 +01:00
|
|
|
}
|
2015-04-06 21:50:44 +02:00
|
|
|
if (LZ4_Supported()) {
|
2015-01-15 01:24:24 +01:00
|
|
|
compression_types.emplace_back(kLZ4Compression, false);
|
|
|
|
compression_types.emplace_back(kLZ4Compression, true);
|
|
|
|
compression_types.emplace_back(kLZ4HCCompression, false);
|
|
|
|
compression_types.emplace_back(kLZ4HCCompression, true);
|
2014-02-08 03:12:30 +01:00
|
|
|
}
|
2016-04-20 07:54:24 +02:00
|
|
|
if (XPRESS_Supported()) {
|
|
|
|
compression_types.emplace_back(kXpressCompression, false);
|
|
|
|
compression_types.emplace_back(kXpressCompression, true);
|
|
|
|
}
|
2015-08-28 00:40:42 +02:00
|
|
|
if (ZSTD_Supported()) {
|
2016-09-02 00:28:40 +02:00
|
|
|
compression_types.emplace_back(kZSTD, false);
|
|
|
|
compression_types.emplace_back(kZSTD, true);
|
2015-08-28 00:40:42 +02:00
|
|
|
}
|
2012-06-29 04:26:43 +02:00
|
|
|
|
2014-01-24 20:09:04 +01:00
|
|
|
for (auto test_type : test_types) {
|
|
|
|
for (auto reverse_compare : reverse_compare_types) {
|
2015-07-20 20:09:14 +02:00
|
|
|
#ifndef ROCKSDB_LITE
|
2014-01-24 20:09:04 +01:00
|
|
|
if (test_type == PLAIN_TABLE_SEMI_FIXED_PREFIX ||
|
2015-09-17 01:57:43 +02:00
|
|
|
test_type == PLAIN_TABLE_FULL_STR_PREFIX ||
|
|
|
|
test_type == PLAIN_TABLE_TOTAL_ORDER) {
|
2013-12-20 18:35:24 +01:00
|
|
|
// Plain table doesn't use restart index or compression.
|
|
|
|
TestArgs one_arg;
|
2014-01-24 20:09:04 +01:00
|
|
|
one_arg.type = test_type;
|
|
|
|
one_arg.reverse_compare = reverse_compare;
|
|
|
|
one_arg.restart_interval = restart_intervals[0];
|
2015-01-15 01:24:24 +01:00
|
|
|
one_arg.compression = compression_types[0].first;
|
2020-04-02 01:37:54 +02:00
|
|
|
one_arg.compression_parallel_threads = 1;
|
2020-06-15 23:06:47 +02:00
|
|
|
one_arg.format_version = 0;
|
2015-09-17 01:57:43 +02:00
|
|
|
one_arg.use_mmap = true;
|
|
|
|
test_args.push_back(one_arg);
|
|
|
|
one_arg.use_mmap = false;
|
2014-01-24 20:09:04 +01:00
|
|
|
test_args.push_back(one_arg);
|
2013-12-20 18:35:24 +01:00
|
|
|
continue;
|
|
|
|
}
|
2015-07-20 20:09:14 +02:00
|
|
|
#endif // !ROCKSDB_LITE
|
2012-06-28 08:41:33 +02:00
|
|
|
|
2014-01-24 20:09:04 +01:00
|
|
|
for (auto restart_interval : restart_intervals) {
|
|
|
|
for (auto compression_type : compression_types) {
|
2020-04-02 01:37:54 +02:00
|
|
|
for (auto num_threads : compression_parallel_threads) {
|
|
|
|
TestArgs one_arg;
|
|
|
|
one_arg.type = test_type;
|
|
|
|
one_arg.reverse_compare = reverse_compare;
|
|
|
|
one_arg.restart_interval = restart_interval;
|
|
|
|
one_arg.compression = compression_type.first;
|
|
|
|
one_arg.compression_parallel_threads = num_threads;
|
2020-06-15 23:06:47 +02:00
|
|
|
one_arg.format_version = compression_type.second ? 2 : 1;
|
2020-04-02 01:37:54 +02:00
|
|
|
one_arg.use_mmap = false;
|
|
|
|
test_args.push_back(one_arg);
|
|
|
|
}
|
2013-12-20 18:35:24 +01:00
|
|
|
}
|
2014-01-24 20:09:04 +01:00
|
|
|
}
|
2013-12-20 18:35:24 +01:00
|
|
|
}
|
2014-01-24 20:09:04 +01:00
|
|
|
}
|
|
|
|
return test_args;
|
2012-06-28 08:41:33 +02:00
|
|
|
}
|
2011-03-18 23:37:00 +01:00
|
|
|
|
2013-12-20 18:35:24 +01:00
|
|
|
// In order to make all tests run for plain table format, including
|
|
|
|
// those operating on empty keys, create a new prefix transformer which
|
|
|
|
// return fixed prefix if the slice is not shorter than the prefix length,
|
|
|
|
// and the full slice if it is shorter.
|
|
|
|
class FixedOrLessPrefixTransform : public SliceTransform {
|
|
|
|
private:
|
|
|
|
const size_t prefix_len_;
|
|
|
|
|
|
|
|
public:
|
|
|
|
explicit FixedOrLessPrefixTransform(size_t prefix_len) :
|
|
|
|
prefix_len_(prefix_len) {
|
|
|
|
}
|
|
|
|
|
2019-02-14 22:52:47 +01:00
|
|
|
const char* Name() const override { return "rocksdb.FixedPrefix"; }
|
2013-12-20 18:35:24 +01:00
|
|
|
|
2019-02-14 22:52:47 +01:00
|
|
|
Slice Transform(const Slice& src) const override {
|
2013-12-20 18:35:24 +01:00
|
|
|
assert(InDomain(src));
|
|
|
|
if (src.size() < prefix_len_) {
|
|
|
|
return src;
|
|
|
|
}
|
|
|
|
return Slice(src.data(), prefix_len_);
|
|
|
|
}
|
|
|
|
|
2019-02-14 22:52:47 +01:00
|
|
|
bool InDomain(const Slice& /*src*/) const override { return true; }
|
2013-12-20 18:35:24 +01:00
|
|
|
|
2019-02-14 22:52:47 +01:00
|
|
|
bool InRange(const Slice& dst) const override {
|
2013-12-20 18:35:24 +01:00
|
|
|
return (dst.size() <= prefix_len_);
|
|
|
|
}
|
2019-02-14 22:52:47 +01:00
|
|
|
bool FullLengthEnabled(size_t* /*len*/) const override { return false; }
|
2013-12-20 18:35:24 +01:00
|
|
|
};
|
|
|
|
|
2015-03-17 22:08:00 +01:00
|
|
|
class HarnessTest : public testing::Test {
|
2011-03-18 23:37:00 +01:00
|
|
|
public:
|
2020-06-13 02:29:22 +02:00
|
|
|
explicit HarnessTest(const TestArgs& args)
|
|
|
|
: args_(args),
|
|
|
|
ioptions_(options_),
|
2018-05-21 23:33:55 +02:00
|
|
|
moptions_(options_),
|
2020-06-13 02:29:22 +02:00
|
|
|
write_buffer_(options_.db_write_buffer_size),
|
|
|
|
support_prev_(true),
|
|
|
|
only_support_prefix_seek_(false) {
|
|
|
|
options_.compression = args_.compression;
|
2020-04-02 01:37:54 +02:00
|
|
|
options_.compression_opts.parallel_threads =
|
2020-06-13 02:29:22 +02:00
|
|
|
args_.compression_parallel_threads;
|
2011-03-18 23:37:00 +01:00
|
|
|
// Use shorter block size for tests to exercise block boundary
|
|
|
|
// conditions more.
|
2020-06-13 02:29:22 +02:00
|
|
|
if (args_.reverse_compare) {
|
2011-03-18 23:37:00 +01:00
|
|
|
options_.comparator = &reverse_key_comparator;
|
|
|
|
}
|
2014-01-27 22:53:22 +01:00
|
|
|
|
|
|
|
internal_comparator_.reset(
|
|
|
|
new test::PlainInternalKeyComparator(options_.comparator));
|
|
|
|
|
2020-06-13 02:29:22 +02:00
|
|
|
options_.allow_mmap_reads = args_.use_mmap;
|
|
|
|
switch (args_.type) {
|
2013-12-20 18:35:24 +01:00
|
|
|
case BLOCK_BASED_TABLE_TEST:
|
2014-08-25 23:22:05 +02:00
|
|
|
table_options_.flush_block_policy_factory.reset(
|
2014-03-01 01:39:27 +01:00
|
|
|
new FlushBlockBySizePolicyFactory());
|
2014-08-25 23:22:05 +02:00
|
|
|
table_options_.block_size = 256;
|
2020-06-13 02:29:22 +02:00
|
|
|
table_options_.block_restart_interval = args_.restart_interval;
|
|
|
|
table_options_.index_block_restart_interval = args_.restart_interval;
|
|
|
|
table_options_.format_version = args_.format_version;
|
2014-08-25 23:22:05 +02:00
|
|
|
options_.table_factory.reset(
|
|
|
|
new BlockBasedTableFactory(table_options_));
|
2020-06-13 02:29:22 +02:00
|
|
|
constructor_.reset(new TableConstructor(
|
|
|
|
options_.comparator, true /* convert_to_internal_key_ */));
|
2016-08-20 00:10:31 +02:00
|
|
|
internal_comparator_.reset(
|
|
|
|
new InternalKeyComparator(options_.comparator));
|
2013-12-20 18:35:24 +01:00
|
|
|
break;
|
2015-07-20 20:09:14 +02:00
|
|
|
// Plain table is not supported in ROCKSDB_LITE
|
|
|
|
#ifndef ROCKSDB_LITE
|
2013-12-20 18:35:24 +01:00
|
|
|
case PLAIN_TABLE_SEMI_FIXED_PREFIX:
|
|
|
|
support_prev_ = false;
|
|
|
|
only_support_prefix_seek_ = true;
|
2014-03-10 20:56:46 +01:00
|
|
|
options_.prefix_extractor.reset(new FixedOrLessPrefixTransform(2));
|
2014-02-08 01:25:38 +01:00
|
|
|
options_.table_factory.reset(NewPlainTableFactory());
|
2020-06-13 02:29:22 +02:00
|
|
|
constructor_.reset(new TableConstructor(
|
|
|
|
options_.comparator, true /* convert_to_internal_key_ */));
|
2014-01-27 22:53:22 +01:00
|
|
|
internal_comparator_.reset(
|
|
|
|
new InternalKeyComparator(options_.comparator));
|
2013-12-20 18:35:24 +01:00
|
|
|
break;
|
|
|
|
case PLAIN_TABLE_FULL_STR_PREFIX:
|
|
|
|
support_prev_ = false;
|
|
|
|
only_support_prefix_seek_ = true;
|
2014-03-10 20:56:46 +01:00
|
|
|
options_.prefix_extractor.reset(NewNoopTransform());
|
2014-02-08 01:25:38 +01:00
|
|
|
options_.table_factory.reset(NewPlainTableFactory());
|
2020-06-13 02:29:22 +02:00
|
|
|
constructor_.reset(new TableConstructor(
|
|
|
|
options_.comparator, true /* convert_to_internal_key_ */));
|
2014-02-08 01:25:38 +01:00
|
|
|
internal_comparator_.reset(
|
|
|
|
new InternalKeyComparator(options_.comparator));
|
|
|
|
break;
|
|
|
|
case PLAIN_TABLE_TOTAL_ORDER:
|
|
|
|
support_prev_ = false;
|
|
|
|
only_support_prefix_seek_ = false;
|
|
|
|
options_.prefix_extractor = nullptr;
|
2014-07-18 09:08:38 +02:00
|
|
|
|
|
|
|
{
|
|
|
|
PlainTableOptions plain_table_options;
|
|
|
|
plain_table_options.user_key_len = kPlainTableVariableLength;
|
|
|
|
plain_table_options.bloom_bits_per_key = 0;
|
|
|
|
plain_table_options.hash_table_ratio = 0;
|
|
|
|
|
|
|
|
options_.table_factory.reset(
|
|
|
|
NewPlainTableFactory(plain_table_options));
|
|
|
|
}
|
2020-06-13 02:29:22 +02:00
|
|
|
constructor_.reset(new TableConstructor(
|
|
|
|
options_.comparator, true /* convert_to_internal_key_ */));
|
2014-01-27 22:53:22 +01:00
|
|
|
internal_comparator_.reset(
|
|
|
|
new InternalKeyComparator(options_.comparator));
|
2011-03-18 23:37:00 +01:00
|
|
|
break;
|
2015-07-20 20:09:14 +02:00
|
|
|
#endif // !ROCKSDB_LITE
|
2011-03-18 23:37:00 +01:00
|
|
|
case BLOCK_TEST:
|
2014-08-25 23:22:05 +02:00
|
|
|
table_options_.block_size = 256;
|
|
|
|
options_.table_factory.reset(
|
|
|
|
new BlockBasedTableFactory(table_options_));
|
2020-06-13 02:29:22 +02:00
|
|
|
constructor_.reset(new BlockConstructor(options_.comparator));
|
2011-03-18 23:37:00 +01:00
|
|
|
break;
|
|
|
|
case MEMTABLE_TEST:
|
2014-08-25 23:22:05 +02:00
|
|
|
table_options_.block_size = 256;
|
|
|
|
options_.table_factory.reset(
|
|
|
|
new BlockBasedTableFactory(table_options_));
|
2020-06-13 02:29:22 +02:00
|
|
|
constructor_.reset(
|
|
|
|
new MemTableConstructor(options_.comparator, &write_buffer_));
|
2011-03-18 23:37:00 +01:00
|
|
|
break;
|
|
|
|
case DB_TEST:
|
2014-08-25 23:22:05 +02:00
|
|
|
table_options_.block_size = 256;
|
|
|
|
options_.table_factory.reset(
|
|
|
|
new BlockBasedTableFactory(table_options_));
|
2020-06-13 02:29:22 +02:00
|
|
|
constructor_.reset(new DBConstructor(options_.comparator));
|
2011-03-18 23:37:00 +01:00
|
|
|
break;
|
|
|
|
}
|
2014-09-05 01:18:36 +02:00
|
|
|
ioptions_ = ImmutableCFOptions(options_);
|
2018-05-21 23:33:55 +02:00
|
|
|
moptions_ = MutableCFOptions(options_);
|
2011-03-18 23:37:00 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
void Add(const std::string& key, const std::string& value) {
|
|
|
|
constructor_->Add(key, value);
|
|
|
|
}
|
|
|
|
|
|
|
|
void Test(Random* rnd) {
|
|
|
|
std::vector<std::string> keys;
|
2015-09-02 22:58:22 +02:00
|
|
|
stl_wrappers::KVMap data;
|
2018-05-21 23:33:55 +02:00
|
|
|
constructor_->Finish(options_, ioptions_, moptions_, table_options_,
|
2014-09-05 01:18:36 +02:00
|
|
|
*internal_comparator_, &keys, &data);
|
2011-03-18 23:37:00 +01:00
|
|
|
|
|
|
|
TestForwardScan(keys, data);
|
2013-12-20 18:35:24 +01:00
|
|
|
if (support_prev_) {
|
|
|
|
TestBackwardScan(keys, data);
|
|
|
|
}
|
2011-03-18 23:37:00 +01:00
|
|
|
TestRandomAccess(rnd, keys, data);
|
|
|
|
}
|
|
|
|
|
2018-03-05 22:08:17 +01:00
|
|
|
void TestForwardScan(const std::vector<std::string>& /*keys*/,
|
2015-09-02 22:58:22 +02:00
|
|
|
const stl_wrappers::KVMap& data) {
|
2015-10-13 00:06:38 +02:00
|
|
|
InternalIterator* iter = constructor_->NewIterator();
|
2011-03-18 23:37:00 +01:00
|
|
|
ASSERT_TRUE(!iter->Valid());
|
|
|
|
iter->SeekToFirst();
|
2020-10-08 20:11:52 +02:00
|
|
|
ASSERT_OK(iter->status());
|
2015-09-02 22:58:22 +02:00
|
|
|
for (stl_wrappers::KVMap::const_iterator model_iter = data.begin();
|
|
|
|
model_iter != data.end(); ++model_iter) {
|
2011-03-18 23:37:00 +01:00
|
|
|
ASSERT_EQ(ToString(data, model_iter), ToString(iter));
|
|
|
|
iter->Next();
|
2020-10-08 20:11:52 +02:00
|
|
|
ASSERT_OK(iter->status());
|
2011-03-18 23:37:00 +01:00
|
|
|
}
|
|
|
|
ASSERT_TRUE(!iter->Valid());
|
2020-10-08 20:11:52 +02:00
|
|
|
ASSERT_OK(iter->status());
|
2014-09-05 02:40:41 +02:00
|
|
|
if (constructor_->IsArenaMode() && !constructor_->AnywayDeleteIterator()) {
|
2015-10-13 00:06:38 +02:00
|
|
|
iter->~InternalIterator();
|
2014-09-05 02:40:41 +02:00
|
|
|
} else {
|
|
|
|
delete iter;
|
|
|
|
}
|
2011-03-18 23:37:00 +01:00
|
|
|
}
|
|
|
|
|
2018-03-05 22:08:17 +01:00
|
|
|
void TestBackwardScan(const std::vector<std::string>& /*keys*/,
|
2015-09-02 22:58:22 +02:00
|
|
|
const stl_wrappers::KVMap& data) {
|
2015-10-13 00:06:38 +02:00
|
|
|
InternalIterator* iter = constructor_->NewIterator();
|
2011-03-18 23:37:00 +01:00
|
|
|
ASSERT_TRUE(!iter->Valid());
|
|
|
|
iter->SeekToLast();
|
2020-10-08 20:11:52 +02:00
|
|
|
ASSERT_OK(iter->status());
|
2015-09-02 22:58:22 +02:00
|
|
|
for (stl_wrappers::KVMap::const_reverse_iterator model_iter = data.rbegin();
|
|
|
|
model_iter != data.rend(); ++model_iter) {
|
2011-03-18 23:37:00 +01:00
|
|
|
ASSERT_EQ(ToString(data, model_iter), ToString(iter));
|
|
|
|
iter->Prev();
|
2020-10-08 20:11:52 +02:00
|
|
|
ASSERT_OK(iter->status());
|
2011-03-18 23:37:00 +01:00
|
|
|
}
|
|
|
|
ASSERT_TRUE(!iter->Valid());
|
2020-10-08 20:11:52 +02:00
|
|
|
ASSERT_OK(iter->status());
|
2014-09-05 02:40:41 +02:00
|
|
|
if (constructor_->IsArenaMode() && !constructor_->AnywayDeleteIterator()) {
|
2015-10-13 00:06:38 +02:00
|
|
|
iter->~InternalIterator();
|
2014-09-05 02:40:41 +02:00
|
|
|
} else {
|
|
|
|
delete iter;
|
|
|
|
}
|
2011-03-18 23:37:00 +01:00
|
|
|
}
|
|
|
|
|
2015-09-02 22:58:22 +02:00
|
|
|
void TestRandomAccess(Random* rnd, const std::vector<std::string>& keys,
|
|
|
|
const stl_wrappers::KVMap& data) {
|
2011-03-18 23:37:00 +01:00
|
|
|
static const bool kVerbose = false;
|
2015-10-13 00:06:38 +02:00
|
|
|
InternalIterator* iter = constructor_->NewIterator();
|
2011-03-18 23:37:00 +01:00
|
|
|
ASSERT_TRUE(!iter->Valid());
|
2015-09-02 22:58:22 +02:00
|
|
|
stl_wrappers::KVMap::const_iterator model_iter = data.begin();
|
2011-03-18 23:37:00 +01:00
|
|
|
if (kVerbose) fprintf(stderr, "---\n");
|
|
|
|
for (int i = 0; i < 200; i++) {
|
2013-12-20 18:35:24 +01:00
|
|
|
const int toss = rnd->Uniform(support_prev_ ? 5 : 3);
|
2011-03-18 23:37:00 +01:00
|
|
|
switch (toss) {
|
|
|
|
case 0: {
|
|
|
|
if (iter->Valid()) {
|
|
|
|
if (kVerbose) fprintf(stderr, "Next\n");
|
|
|
|
iter->Next();
|
2020-10-08 20:11:52 +02:00
|
|
|
ASSERT_OK(iter->status());
|
2011-03-18 23:37:00 +01:00
|
|
|
++model_iter;
|
|
|
|
ASSERT_EQ(ToString(data, model_iter), ToString(iter));
|
|
|
|
}
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
case 1: {
|
|
|
|
if (kVerbose) fprintf(stderr, "SeekToFirst\n");
|
|
|
|
iter->SeekToFirst();
|
2020-10-08 20:11:52 +02:00
|
|
|
ASSERT_OK(iter->status());
|
2011-03-18 23:37:00 +01:00
|
|
|
model_iter = data.begin();
|
|
|
|
ASSERT_EQ(ToString(data, model_iter), ToString(iter));
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
case 2: {
|
|
|
|
std::string key = PickRandomKey(rnd, keys);
|
|
|
|
model_iter = data.lower_bound(key);
|
|
|
|
if (kVerbose) fprintf(stderr, "Seek '%s'\n",
|
|
|
|
EscapeString(key).c_str());
|
|
|
|
iter->Seek(Slice(key));
|
2020-10-08 20:11:52 +02:00
|
|
|
ASSERT_OK(iter->status());
|
2011-03-18 23:37:00 +01:00
|
|
|
ASSERT_EQ(ToString(data, model_iter), ToString(iter));
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
case 3: {
|
|
|
|
if (iter->Valid()) {
|
|
|
|
if (kVerbose) fprintf(stderr, "Prev\n");
|
|
|
|
iter->Prev();
|
2020-10-08 20:11:52 +02:00
|
|
|
ASSERT_OK(iter->status());
|
2011-03-18 23:37:00 +01:00
|
|
|
if (model_iter == data.begin()) {
|
|
|
|
model_iter = data.end(); // Wrap around to invalid value
|
|
|
|
} else {
|
|
|
|
--model_iter;
|
|
|
|
}
|
|
|
|
ASSERT_EQ(ToString(data, model_iter), ToString(iter));
|
|
|
|
}
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
case 4: {
|
|
|
|
if (kVerbose) fprintf(stderr, "SeekToLast\n");
|
|
|
|
iter->SeekToLast();
|
2020-10-08 20:11:52 +02:00
|
|
|
ASSERT_OK(iter->status());
|
2011-03-18 23:37:00 +01:00
|
|
|
if (keys.empty()) {
|
|
|
|
model_iter = data.end();
|
|
|
|
} else {
|
|
|
|
std::string last = data.rbegin()->first;
|
|
|
|
model_iter = data.lower_bound(last);
|
|
|
|
}
|
|
|
|
ASSERT_EQ(ToString(data, model_iter), ToString(iter));
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
2014-09-05 02:40:41 +02:00
|
|
|
if (constructor_->IsArenaMode() && !constructor_->AnywayDeleteIterator()) {
|
2015-10-13 00:06:38 +02:00
|
|
|
iter->~InternalIterator();
|
2014-09-05 02:40:41 +02:00
|
|
|
} else {
|
|
|
|
delete iter;
|
|
|
|
}
|
2011-03-18 23:37:00 +01:00
|
|
|
}
|
|
|
|
|
2015-09-02 22:58:22 +02:00
|
|
|
std::string ToString(const stl_wrappers::KVMap& data,
|
|
|
|
const stl_wrappers::KVMap::const_iterator& it) {
|
2011-03-18 23:37:00 +01:00
|
|
|
if (it == data.end()) {
|
|
|
|
return "END";
|
|
|
|
} else {
|
|
|
|
return "'" + it->first + "->" + it->second + "'";
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2015-09-02 22:58:22 +02:00
|
|
|
std::string ToString(const stl_wrappers::KVMap& data,
|
|
|
|
const stl_wrappers::KVMap::const_reverse_iterator& it) {
|
2011-03-18 23:37:00 +01:00
|
|
|
if (it == data.rend()) {
|
|
|
|
return "END";
|
|
|
|
} else {
|
|
|
|
return "'" + it->first + "->" + it->second + "'";
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2015-10-13 00:06:38 +02:00
|
|
|
std::string ToString(const InternalIterator* it) {
|
2011-03-18 23:37:00 +01:00
|
|
|
if (!it->Valid()) {
|
|
|
|
return "END";
|
|
|
|
} else {
|
|
|
|
return "'" + it->key().ToString() + "->" + it->value().ToString() + "'";
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
std::string PickRandomKey(Random* rnd, const std::vector<std::string>& keys) {
|
|
|
|
if (keys.empty()) {
|
|
|
|
return "foo";
|
|
|
|
} else {
|
2014-11-11 22:47:22 +01:00
|
|
|
const int index = rnd->Uniform(static_cast<int>(keys.size()));
|
2011-03-18 23:37:00 +01:00
|
|
|
std::string result = keys[index];
|
2013-12-20 18:35:24 +01:00
|
|
|
switch (rnd->Uniform(support_prev_ ? 3 : 1)) {
|
2011-03-18 23:37:00 +01:00
|
|
|
case 0:
|
|
|
|
// Return an existing key
|
|
|
|
break;
|
|
|
|
case 1: {
|
|
|
|
// Attempt to return something smaller than an existing key
|
2013-12-20 18:35:24 +01:00
|
|
|
if (result.size() > 0 && result[result.size() - 1] > '\0'
|
|
|
|
&& (!only_support_prefix_seek_
|
|
|
|
|| options_.prefix_extractor->Transform(result).size()
|
|
|
|
< result.size())) {
|
|
|
|
result[result.size() - 1]--;
|
2011-03-18 23:37:00 +01:00
|
|
|
}
|
|
|
|
break;
|
2013-12-20 18:35:24 +01:00
|
|
|
}
|
2011-03-18 23:37:00 +01:00
|
|
|
case 2: {
|
|
|
|
// Return something larger than an existing key
|
|
|
|
Increment(options_.comparator, &result);
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
return result;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2013-03-01 03:04:58 +01:00
|
|
|
// Returns nullptr if not running against a DB
|
2011-03-21 20:40:57 +01:00
|
|
|
DB* db() const { return constructor_->db(); }
|
|
|
|
|
2011-03-18 23:37:00 +01:00
|
|
|
private:
|
2020-06-13 02:29:22 +02:00
|
|
|
TestArgs args_;
|
|
|
|
Options options_;
|
2014-09-05 01:18:36 +02:00
|
|
|
ImmutableCFOptions ioptions_;
|
2018-05-21 23:33:55 +02:00
|
|
|
MutableCFOptions moptions_;
|
2020-06-13 02:29:22 +02:00
|
|
|
BlockBasedTableOptions table_options_;
|
|
|
|
std::unique_ptr<Constructor> constructor_;
|
2016-06-21 03:01:03 +02:00
|
|
|
WriteBufferManager write_buffer_;
|
2013-12-20 18:35:24 +01:00
|
|
|
bool support_prev_;
|
|
|
|
bool only_support_prefix_seek_;
|
2018-11-09 20:17:34 +01:00
|
|
|
std::shared_ptr<InternalKeyComparator> internal_comparator_;
|
2011-03-18 23:37:00 +01:00
|
|
|
};
|
|
|
|
|
2020-06-13 02:29:22 +02:00
|
|
|
class ParameterizedHarnessTest : public HarnessTest,
|
|
|
|
public testing::WithParamInterface<TestArgs> {
|
|
|
|
public:
|
|
|
|
ParameterizedHarnessTest() : HarnessTest(GetParam()) {}
|
|
|
|
};
|
|
|
|
|
|
|
|
INSTANTIATE_TEST_CASE_P(TableTest, ParameterizedHarnessTest,
|
|
|
|
::testing::ValuesIn(GenerateArgList()));
|
|
|
|
|
|
|
|
class DBHarnessTest : public HarnessTest {
|
|
|
|
public:
|
|
|
|
DBHarnessTest()
|
|
|
|
: HarnessTest(TestArgs{DB_TEST, /* reverse_compare */ false,
|
|
|
|
/* restart_interval */ 16, kNoCompression,
|
|
|
|
/* compression_parallel_threads */ 1,
|
|
|
|
/* format_version */ 0, /* use_mmap */ false}) {}
|
|
|
|
};
|
|
|
|
|
2011-03-18 23:37:00 +01:00
|
|
|
static bool Between(uint64_t val, uint64_t low, uint64_t high) {
|
|
|
|
bool result = (val >= low) && (val <= high);
|
|
|
|
if (!result) {
|
|
|
|
fprintf(stderr, "Value %llu is not in range [%llu, %llu]\n",
|
|
|
|
(unsigned long long)(val),
|
|
|
|
(unsigned long long)(low),
|
|
|
|
(unsigned long long)(high));
|
|
|
|
}
|
|
|
|
return result;
|
|
|
|
}
|
|
|
|
|
2014-01-24 21:14:08 +01:00
|
|
|
// Tests against all kinds of tables
|
2015-03-17 22:08:00 +01:00
|
|
|
class TableTest : public testing::Test {
|
2014-01-27 22:53:22 +01:00
|
|
|
public:
|
|
|
|
const InternalKeyComparator& GetPlainInternalComparator(
|
|
|
|
const Comparator* comp) {
|
|
|
|
if (!plain_internal_comparator) {
|
|
|
|
plain_internal_comparator.reset(
|
|
|
|
new test::PlainInternalKeyComparator(comp));
|
|
|
|
}
|
|
|
|
return *plain_internal_comparator;
|
|
|
|
}
|
2017-02-07 01:29:29 +01:00
|
|
|
void IndexTest(BlockBasedTableOptions table_options);
|
2014-01-27 22:53:22 +01:00
|
|
|
|
|
|
|
private:
|
|
|
|
std::unique_ptr<InternalKeyComparator> plain_internal_comparator;
|
|
|
|
};
|
|
|
|
|
|
|
|
class GeneralTableTest : public TableTest {};
|
2018-06-05 04:59:44 +02:00
|
|
|
class BlockBasedTableTest
|
|
|
|
: public TableTest,
|
|
|
|
virtual public ::testing::WithParamInterface<uint32_t> {
|
|
|
|
public:
|
2019-07-17 22:02:00 +02:00
|
|
|
BlockBasedTableTest() : format_(GetParam()) {
|
2020-02-20 21:07:53 +01:00
|
|
|
env_ = ROCKSDB_NAMESPACE::Env::Default();
|
2019-07-17 22:02:00 +02:00
|
|
|
}
|
2018-06-05 04:59:44 +02:00
|
|
|
|
|
|
|
BlockBasedTableOptions GetBlockBasedTableOptions() {
|
|
|
|
BlockBasedTableOptions options;
|
|
|
|
options.format_version = format_;
|
|
|
|
return options;
|
|
|
|
}
|
|
|
|
|
2019-07-17 22:02:00 +02:00
|
|
|
void SetupTracingTest(TableConstructor* c) {
|
|
|
|
test_path_ = test::PerThreadDBPath("block_based_table_tracing_test");
|
|
|
|
EXPECT_OK(env_->CreateDir(test_path_));
|
|
|
|
trace_file_path_ = test_path_ + "/block_cache_trace_file";
|
|
|
|
TraceOptions trace_opt;
|
|
|
|
std::unique_ptr<TraceWriter> trace_writer;
|
|
|
|
EXPECT_OK(NewFileTraceWriter(env_, EnvOptions(), trace_file_path_,
|
|
|
|
&trace_writer));
|
2020-10-08 20:11:52 +02:00
|
|
|
// Always return Status::OK().
|
|
|
|
assert(c->block_cache_tracer_
|
|
|
|
.StartTrace(env_, trace_opt, std::move(trace_writer))
|
|
|
|
.ok());
|
2019-07-17 22:02:00 +02:00
|
|
|
{
|
|
|
|
std::string user_key = "k01";
|
|
|
|
InternalKey internal_key(user_key, 0, kTypeValue);
|
|
|
|
std::string encoded_key = internal_key.Encode().ToString();
|
|
|
|
c->Add(encoded_key, kDummyValue);
|
|
|
|
}
|
|
|
|
{
|
|
|
|
std::string user_key = "k02";
|
|
|
|
InternalKey internal_key(user_key, 0, kTypeValue);
|
|
|
|
std::string encoded_key = internal_key.Encode().ToString();
|
|
|
|
c->Add(encoded_key, kDummyValue);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
void VerifyBlockAccessTrace(
|
|
|
|
TableConstructor* c,
|
|
|
|
const std::vector<BlockCacheTraceRecord>& expected_records) {
|
|
|
|
c->block_cache_tracer_.EndTrace();
|
|
|
|
|
2020-04-03 18:52:38 +02:00
|
|
|
{
|
|
|
|
std::unique_ptr<TraceReader> trace_reader;
|
|
|
|
Status s =
|
|
|
|
NewFileTraceReader(env_, EnvOptions(), trace_file_path_, &trace_reader);
|
|
|
|
EXPECT_OK(s);
|
|
|
|
BlockCacheTraceReader reader(std::move(trace_reader));
|
|
|
|
BlockCacheTraceHeader header;
|
|
|
|
EXPECT_OK(reader.ReadHeader(&header));
|
|
|
|
uint32_t index = 0;
|
|
|
|
while (s.ok()) {
|
|
|
|
BlockCacheTraceRecord access;
|
|
|
|
s = reader.ReadAccess(&access);
|
|
|
|
if (!s.ok()) {
|
|
|
|
break;
|
2019-07-17 22:02:00 +02:00
|
|
|
}
|
2020-04-03 18:52:38 +02:00
|
|
|
ASSERT_LT(index, expected_records.size());
|
|
|
|
EXPECT_NE("", access.block_key);
|
|
|
|
EXPECT_EQ(access.block_type, expected_records[index].block_type);
|
|
|
|
EXPECT_GT(access.block_size, 0);
|
|
|
|
EXPECT_EQ(access.caller, expected_records[index].caller);
|
|
|
|
EXPECT_EQ(access.no_insert, expected_records[index].no_insert);
|
|
|
|
EXPECT_EQ(access.is_cache_hit, expected_records[index].is_cache_hit);
|
|
|
|
// Get
|
|
|
|
if (access.caller == TableReaderCaller::kUserGet) {
|
|
|
|
EXPECT_EQ(access.referenced_key,
|
|
|
|
expected_records[index].referenced_key);
|
|
|
|
EXPECT_EQ(access.get_id, expected_records[index].get_id);
|
|
|
|
EXPECT_EQ(access.get_from_user_specified_snapshot,
|
|
|
|
expected_records[index].get_from_user_specified_snapshot);
|
|
|
|
if (access.block_type == TraceType::kBlockTraceDataBlock) {
|
|
|
|
EXPECT_GT(access.referenced_data_size, 0);
|
|
|
|
EXPECT_GT(access.num_keys_in_block, 0);
|
|
|
|
EXPECT_EQ(access.referenced_key_exist_in_block,
|
|
|
|
expected_records[index].referenced_key_exist_in_block);
|
|
|
|
}
|
|
|
|
} else {
|
|
|
|
EXPECT_EQ(access.referenced_key, "");
|
|
|
|
EXPECT_EQ(access.get_id, 0);
|
|
|
|
EXPECT_TRUE(access.get_from_user_specified_snapshot == Boolean::kFalse);
|
|
|
|
EXPECT_EQ(access.referenced_data_size, 0);
|
|
|
|
EXPECT_EQ(access.num_keys_in_block, 0);
|
|
|
|
EXPECT_TRUE(access.referenced_key_exist_in_block == Boolean::kFalse);
|
|
|
|
}
|
|
|
|
index++;
|
2019-07-17 22:02:00 +02:00
|
|
|
}
|
2020-04-03 18:52:38 +02:00
|
|
|
EXPECT_EQ(index, expected_records.size());
|
2019-07-17 22:02:00 +02:00
|
|
|
}
|
|
|
|
EXPECT_OK(env_->DeleteFile(trace_file_path_));
|
|
|
|
EXPECT_OK(env_->DeleteDir(test_path_));
|
|
|
|
}
|
|
|
|
|
2018-01-11 00:06:29 +01:00
|
|
|
protected:
|
|
|
|
uint64_t IndexUncompressedHelper(bool indexCompress);
|
2018-06-05 04:59:44 +02:00
|
|
|
|
|
|
|
private:
|
|
|
|
uint32_t format_;
|
2019-07-17 22:02:00 +02:00
|
|
|
Env* env_;
|
|
|
|
std::string trace_file_path_;
|
|
|
|
std::string test_path_;
|
2018-01-11 00:06:29 +01:00
|
|
|
};
|
2014-01-27 22:53:22 +01:00
|
|
|
class PlainTableTest : public TableTest {};
|
2015-03-17 22:08:00 +01:00
|
|
|
class TablePropertyTest : public testing::Test {};
|
2018-07-20 23:31:27 +02:00
|
|
|
class BBTTailPrefetchTest : public TableTest {};
|
2014-02-12 22:14:59 +01:00
|
|
|
|
2020-02-11 00:42:46 +01:00
|
|
|
// The helper class to test the file checksum
|
|
|
|
class FileChecksumTestHelper {
|
|
|
|
public:
|
|
|
|
FileChecksumTestHelper(bool convert_to_internal_key = false)
|
|
|
|
: convert_to_internal_key_(convert_to_internal_key) {
|
|
|
|
}
|
|
|
|
~FileChecksumTestHelper() {}
|
|
|
|
|
|
|
|
void CreateWriteableFile() {
|
2020-06-05 00:32:29 +02:00
|
|
|
sink_ = new test::StringSink();
|
2021-01-05 00:59:52 +01:00
|
|
|
std::unique_ptr<FSWritableFile> holder(sink_);
|
|
|
|
file_writer_.reset(new WritableFileWriter(
|
|
|
|
std::move(holder), "" /* don't care */, FileOptions()));
|
2020-02-11 00:42:46 +01:00
|
|
|
}
|
|
|
|
|
2020-03-30 00:57:02 +02:00
|
|
|
void SetFileChecksumGenerator(FileChecksumGenerator* checksum_generator) {
|
2020-02-11 00:42:46 +01:00
|
|
|
if (file_writer_ != nullptr) {
|
2020-03-30 00:57:02 +02:00
|
|
|
file_writer_->TEST_SetFileChecksumGenerator(checksum_generator);
|
2020-03-30 23:08:55 +02:00
|
|
|
} else {
|
|
|
|
delete checksum_generator;
|
2020-02-11 00:42:46 +01:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
WritableFileWriter* GetFileWriter() { return file_writer_.get(); }
|
|
|
|
|
|
|
|
Status ResetTableBuilder(std::unique_ptr<TableBuilder>&& builder) {
|
|
|
|
assert(builder != nullptr);
|
|
|
|
table_builder_ = std::move(builder);
|
|
|
|
return Status::OK();
|
|
|
|
}
|
|
|
|
|
|
|
|
void AddKVtoKVMap(int num_entries) {
|
|
|
|
Random rnd(test::RandomSeed());
|
|
|
|
for (int i = 0; i < num_entries; i++) {
|
2020-07-09 23:33:42 +02:00
|
|
|
std::string v = rnd.RandomString(100);
|
2020-02-11 00:42:46 +01:00
|
|
|
kv_map_[test::RandomKey(&rnd, 20)] = v;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
Status WriteKVAndFlushTable() {
|
2020-06-25 01:20:55 +02:00
|
|
|
for (const auto& kv : kv_map_) {
|
2020-02-11 00:42:46 +01:00
|
|
|
if (convert_to_internal_key_) {
|
|
|
|
ParsedInternalKey ikey(kv.first, kMaxSequenceNumber, kTypeValue);
|
|
|
|
std::string encoded;
|
|
|
|
AppendInternalKey(&encoded, ikey);
|
|
|
|
table_builder_->Add(encoded, kv.second);
|
|
|
|
} else {
|
|
|
|
table_builder_->Add(kv.first, kv.second);
|
|
|
|
}
|
|
|
|
EXPECT_TRUE(table_builder_->status().ok());
|
|
|
|
}
|
|
|
|
Status s = table_builder_->Finish();
|
2020-12-24 01:54:05 +01:00
|
|
|
EXPECT_OK(file_writer_->Flush());
|
|
|
|
EXPECT_OK(s);
|
2020-02-11 00:42:46 +01:00
|
|
|
|
|
|
|
EXPECT_EQ(sink_->contents().size(), table_builder_->FileSize());
|
|
|
|
return s;
|
|
|
|
}
|
|
|
|
|
2020-03-30 00:57:02 +02:00
|
|
|
std::string GetFileChecksum() {
|
2020-12-24 01:54:05 +01:00
|
|
|
EXPECT_OK(file_writer_->Close());
|
2020-03-30 00:57:02 +02:00
|
|
|
return table_builder_->GetFileChecksum();
|
|
|
|
}
|
2020-02-11 00:42:46 +01:00
|
|
|
|
|
|
|
const char* GetFileChecksumFuncName() {
|
|
|
|
return table_builder_->GetFileChecksumFuncName();
|
|
|
|
}
|
|
|
|
|
2020-03-30 00:57:02 +02:00
|
|
|
Status CalculateFileChecksum(FileChecksumGenerator* file_checksum_generator,
|
2020-02-11 00:42:46 +01:00
|
|
|
std::string* checksum) {
|
2020-03-30 00:57:02 +02:00
|
|
|
assert(file_checksum_generator != nullptr);
|
2020-02-11 00:42:46 +01:00
|
|
|
cur_uniq_id_ = checksum_uniq_id_++;
|
|
|
|
test::StringSink* ss_rw =
|
2021-01-05 00:59:52 +01:00
|
|
|
static_cast<test::StringSink*>(file_writer_->writable_file());
|
|
|
|
std::unique_ptr<FSRandomAccessFile> source(
|
|
|
|
new test::StringSource(ss_rw->contents()));
|
|
|
|
file_reader_.reset(new RandomAccessFileReader(std::move(source), "test"));
|
|
|
|
|
2020-02-11 00:42:46 +01:00
|
|
|
std::unique_ptr<char[]> scratch(new char[2048]);
|
|
|
|
Slice result;
|
|
|
|
uint64_t offset = 0;
|
|
|
|
Status s;
|
2020-04-30 23:48:51 +02:00
|
|
|
s = file_reader_->Read(IOOptions(), offset, 2048, &result, scratch.get(),
|
|
|
|
nullptr, false);
|
2020-02-11 00:42:46 +01:00
|
|
|
if (!s.ok()) {
|
|
|
|
return s;
|
|
|
|
}
|
|
|
|
while (result.size() != 0) {
|
2020-03-30 00:57:02 +02:00
|
|
|
file_checksum_generator->Update(scratch.get(), result.size());
|
2020-02-11 00:42:46 +01:00
|
|
|
offset += static_cast<uint64_t>(result.size());
|
2020-04-30 23:48:51 +02:00
|
|
|
s = file_reader_->Read(IOOptions(), offset, 2048, &result, scratch.get(),
|
|
|
|
nullptr, false);
|
2020-02-11 00:42:46 +01:00
|
|
|
if (!s.ok()) {
|
|
|
|
return s;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
EXPECT_EQ(offset, static_cast<uint64_t>(table_builder_->FileSize()));
|
2020-03-30 00:57:02 +02:00
|
|
|
file_checksum_generator->Finalize();
|
|
|
|
*checksum = file_checksum_generator->GetChecksum();
|
2020-02-11 00:42:46 +01:00
|
|
|
return Status::OK();
|
|
|
|
}
|
|
|
|
|
|
|
|
private:
|
|
|
|
bool convert_to_internal_key_;
|
|
|
|
uint64_t cur_uniq_id_;
|
|
|
|
std::unique_ptr<WritableFileWriter> file_writer_;
|
|
|
|
std::unique_ptr<RandomAccessFileReader> file_reader_;
|
|
|
|
std::unique_ptr<TableBuilder> table_builder_;
|
|
|
|
stl_wrappers::KVMap kv_map_;
|
2020-06-05 00:32:29 +02:00
|
|
|
test::StringSink* sink_ = nullptr;
|
2020-02-11 00:42:46 +01:00
|
|
|
|
|
|
|
static uint64_t checksum_uniq_id_;
|
|
|
|
};
|
|
|
|
|
|
|
|
uint64_t FileChecksumTestHelper::checksum_uniq_id_ = 1;
|
|
|
|
|
2020-06-04 00:53:09 +02:00
|
|
|
INSTANTIATE_TEST_CASE_P(FormatDef, BlockBasedTableTest,
|
|
|
|
testing::Values(test::kDefaultFormatVersion));
|
|
|
|
INSTANTIATE_TEST_CASE_P(FormatLatest, BlockBasedTableTest,
|
|
|
|
testing::Values(test::kLatestFormatVersion));
|
2018-06-05 04:59:44 +02:00
|
|
|
|
2014-02-12 22:14:59 +01:00
|
|
|
// This test serves as the living tutorial for the prefix scan of user collected
|
|
|
|
// properties.
|
2015-03-17 22:08:00 +01:00
|
|
|
TEST_F(TablePropertyTest, PrefixScanTest) {
|
2014-02-12 22:14:59 +01:00
|
|
|
UserCollectedProperties props{{"num.111.1", "1"},
|
|
|
|
{"num.111.2", "2"},
|
|
|
|
{"num.111.3", "3"},
|
|
|
|
{"num.333.1", "1"},
|
|
|
|
{"num.333.2", "2"},
|
|
|
|
{"num.333.3", "3"},
|
|
|
|
{"num.555.1", "1"},
|
|
|
|
{"num.555.2", "2"},
|
|
|
|
{"num.555.3", "3"}, };
|
|
|
|
|
|
|
|
// prefixes that exist
|
|
|
|
for (const std::string& prefix : {"num.111", "num.333", "num.555"}) {
|
|
|
|
int num = 0;
|
|
|
|
for (auto pos = props.lower_bound(prefix);
|
|
|
|
pos != props.end() &&
|
|
|
|
pos->first.compare(0, prefix.size(), prefix) == 0;
|
|
|
|
++pos) {
|
|
|
|
++num;
|
2014-11-25 05:44:49 +01:00
|
|
|
auto key = prefix + "." + ToString(num);
|
2014-02-12 22:14:59 +01:00
|
|
|
ASSERT_EQ(key, pos->first);
|
2014-11-25 05:44:49 +01:00
|
|
|
ASSERT_EQ(ToString(num), pos->second);
|
2014-02-12 22:14:59 +01:00
|
|
|
}
|
|
|
|
ASSERT_EQ(3, num);
|
|
|
|
}
|
|
|
|
|
|
|
|
// prefixes that don't exist
|
|
|
|
for (const std::string& prefix :
|
|
|
|
{"num.000", "num.222", "num.444", "num.666"}) {
|
|
|
|
auto pos = props.lower_bound(prefix);
|
|
|
|
ASSERT_TRUE(pos == props.end() ||
|
|
|
|
pos->first.compare(0, prefix.size(), prefix) != 0);
|
|
|
|
}
|
|
|
|
}
|
2011-03-18 23:37:00 +01:00
|
|
|
|
2013-10-10 20:43:24 +02:00
|
|
|
// This test include all the basic checks except those for index size and block
|
|
|
|
// size, which will be conducted in separated unit tests.
|
2018-06-05 04:59:44 +02:00
|
|
|
TEST_P(BlockBasedTableTest, BasicBlockBasedTableProperties) {
|
2016-08-20 00:10:31 +02:00
|
|
|
TableConstructor c(BytewiseComparator(), true /* convert_to_internal_key_ */);
|
2013-10-10 20:43:24 +02:00
|
|
|
|
|
|
|
c.Add("a1", "val1");
|
|
|
|
c.Add("b2", "val2");
|
|
|
|
c.Add("c3", "val3");
|
|
|
|
c.Add("d4", "val4");
|
|
|
|
c.Add("e5", "val5");
|
|
|
|
c.Add("f6", "val6");
|
|
|
|
c.Add("g7", "val7");
|
|
|
|
c.Add("h8", "val8");
|
|
|
|
c.Add("j9", "val9");
|
2016-08-20 00:10:31 +02:00
|
|
|
uint64_t diff_internal_user_bytes = 9 * 8; // 8 is seq size, 9 k-v totally
|
2013-10-10 20:43:24 +02:00
|
|
|
|
|
|
|
std::vector<std::string> keys;
|
2015-09-02 22:58:22 +02:00
|
|
|
stl_wrappers::KVMap kvmap;
|
2013-11-20 07:00:48 +01:00
|
|
|
Options options;
|
2013-10-10 20:43:24 +02:00
|
|
|
options.compression = kNoCompression;
|
2018-01-11 00:06:29 +01:00
|
|
|
options.statistics = CreateDBStatistics();
|
2019-03-01 19:39:00 +01:00
|
|
|
options.statistics->set_stats_level(StatsLevel::kAll);
|
2018-06-05 04:59:44 +02:00
|
|
|
BlockBasedTableOptions table_options = GetBlockBasedTableOptions();
|
2014-08-25 23:22:05 +02:00
|
|
|
table_options.block_restart_interval = 1;
|
|
|
|
options.table_factory.reset(NewBlockBasedTableFactory(table_options));
|
2013-10-10 20:43:24 +02:00
|
|
|
|
2018-01-11 00:06:29 +01:00
|
|
|
ImmutableCFOptions ioptions(options);
|
2018-05-21 23:33:55 +02:00
|
|
|
MutableCFOptions moptions(options);
|
2018-01-11 00:06:29 +01:00
|
|
|
ioptions.statistics = options.statistics.get();
|
2018-05-21 23:33:55 +02:00
|
|
|
c.Finish(options, ioptions, moptions, table_options,
|
2014-08-25 23:22:05 +02:00
|
|
|
GetPlainInternalComparator(options.comparator), &keys, &kvmap);
|
2018-01-11 00:06:29 +01:00
|
|
|
ASSERT_EQ(options.statistics->getTickerCount(NUMBER_BLOCK_NOT_COMPRESSED), 0);
|
2013-10-10 20:43:24 +02:00
|
|
|
|
2014-08-26 01:14:30 +02:00
|
|
|
auto& props = *c.GetTableReader()->GetTableProperties();
|
2013-11-20 01:29:42 +01:00
|
|
|
ASSERT_EQ(kvmap.size(), props.num_entries);
|
2013-10-10 20:43:24 +02:00
|
|
|
|
|
|
|
auto raw_key_size = kvmap.size() * 2ul;
|
|
|
|
auto raw_value_size = kvmap.size() * 4ul;
|
|
|
|
|
2016-08-20 00:10:31 +02:00
|
|
|
ASSERT_EQ(raw_key_size + diff_internal_user_bytes, props.raw_key_size);
|
2013-11-20 01:29:42 +01:00
|
|
|
ASSERT_EQ(raw_value_size, props.raw_value_size);
|
|
|
|
ASSERT_EQ(1ul, props.num_data_blocks);
|
|
|
|
ASSERT_EQ("", props.filter_policy_name); // no filter policy is used
|
2013-10-10 20:43:24 +02:00
|
|
|
|
|
|
|
// Verify data size.
|
2014-09-02 20:49:38 +02:00
|
|
|
BlockBuilder block_builder(1);
|
2013-10-10 20:43:24 +02:00
|
|
|
for (const auto& item : kvmap) {
|
|
|
|
block_builder.Add(item.first, item.second);
|
|
|
|
}
|
|
|
|
Slice content = block_builder.Finish();
|
2016-08-20 00:10:31 +02:00
|
|
|
ASSERT_EQ(content.size() + kBlockTrailerSize + diff_internal_user_bytes,
|
|
|
|
props.data_size);
|
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 19:42:39 +02:00
|
|
|
c.ResetTableReader();
|
2013-10-10 20:43:24 +02:00
|
|
|
}
|
|
|
|
|
2018-01-17 02:26:29 +01:00
|
|
|
#ifdef SNAPPY
|
2018-01-11 00:06:29 +01:00
|
|
|
uint64_t BlockBasedTableTest::IndexUncompressedHelper(bool compressed) {
|
|
|
|
TableConstructor c(BytewiseComparator(), true /* convert_to_internal_key_ */);
|
|
|
|
constexpr size_t kNumKeys = 10000;
|
|
|
|
|
|
|
|
for (size_t k = 0; k < kNumKeys; ++k) {
|
|
|
|
c.Add("key" + ToString(k), "val" + ToString(k));
|
|
|
|
}
|
|
|
|
|
|
|
|
std::vector<std::string> keys;
|
|
|
|
stl_wrappers::KVMap kvmap;
|
|
|
|
Options options;
|
|
|
|
options.compression = kSnappyCompression;
|
|
|
|
options.statistics = CreateDBStatistics();
|
2019-03-01 19:39:00 +01:00
|
|
|
options.statistics->set_stats_level(StatsLevel::kAll);
|
2018-06-05 04:59:44 +02:00
|
|
|
BlockBasedTableOptions table_options = GetBlockBasedTableOptions();
|
2018-01-11 00:06:29 +01:00
|
|
|
table_options.block_restart_interval = 1;
|
|
|
|
table_options.enable_index_compression = compressed;
|
|
|
|
options.table_factory.reset(NewBlockBasedTableFactory(table_options));
|
|
|
|
|
|
|
|
ImmutableCFOptions ioptions(options);
|
2018-05-21 23:33:55 +02:00
|
|
|
MutableCFOptions moptions(options);
|
2018-01-11 00:06:29 +01:00
|
|
|
ioptions.statistics = options.statistics.get();
|
2018-05-21 23:33:55 +02:00
|
|
|
c.Finish(options, ioptions, moptions, table_options,
|
2018-01-11 00:06:29 +01:00
|
|
|
GetPlainInternalComparator(options.comparator), &keys, &kvmap);
|
|
|
|
c.ResetTableReader();
|
|
|
|
return options.statistics->getTickerCount(NUMBER_BLOCK_COMPRESSED);
|
|
|
|
}
|
2018-06-05 04:59:44 +02:00
|
|
|
TEST_P(BlockBasedTableTest, IndexUncompressed) {
|
2018-01-11 00:06:29 +01:00
|
|
|
uint64_t tbl1_compressed_cnt = IndexUncompressedHelper(true);
|
|
|
|
uint64_t tbl2_compressed_cnt = IndexUncompressedHelper(false);
|
|
|
|
// tbl1_compressed_cnt should include 1 index block
|
|
|
|
EXPECT_EQ(tbl2_compressed_cnt + 1, tbl1_compressed_cnt);
|
|
|
|
}
|
2018-01-17 02:26:29 +01:00
|
|
|
#endif // SNAPPY
|
2018-01-11 00:06:29 +01:00
|
|
|
|
2018-06-05 04:59:44 +02:00
|
|
|
TEST_P(BlockBasedTableTest, BlockBasedTableProperties2) {
|
2016-04-21 19:16:28 +02:00
|
|
|
TableConstructor c(&reverse_key_comparator);
|
|
|
|
std::vector<std::string> keys;
|
|
|
|
stl_wrappers::KVMap kvmap;
|
|
|
|
|
|
|
|
{
|
|
|
|
Options options;
|
2016-05-12 18:47:16 +02:00
|
|
|
options.compression = CompressionType::kNoCompression;
|
2018-06-05 04:59:44 +02:00
|
|
|
BlockBasedTableOptions table_options = GetBlockBasedTableOptions();
|
2016-04-21 19:16:28 +02:00
|
|
|
options.table_factory.reset(NewBlockBasedTableFactory(table_options));
|
|
|
|
|
|
|
|
const ImmutableCFOptions ioptions(options);
|
2018-05-21 23:33:55 +02:00
|
|
|
const MutableCFOptions moptions(options);
|
|
|
|
c.Finish(options, ioptions, moptions, table_options,
|
2016-04-21 19:16:28 +02:00
|
|
|
GetPlainInternalComparator(options.comparator), &keys, &kvmap);
|
|
|
|
|
|
|
|
auto& props = *c.GetTableReader()->GetTableProperties();
|
|
|
|
|
|
|
|
// Default comparator
|
|
|
|
ASSERT_EQ("leveldb.BytewiseComparator", props.comparator_name);
|
|
|
|
// No merge operator
|
|
|
|
ASSERT_EQ("nullptr", props.merge_operator_name);
|
2016-08-26 20:46:32 +02:00
|
|
|
// No prefix extractor
|
|
|
|
ASSERT_EQ("nullptr", props.prefix_extractor_name);
|
2016-04-21 19:16:28 +02:00
|
|
|
// No property collectors
|
|
|
|
ASSERT_EQ("[]", props.property_collectors_names);
|
|
|
|
// No filter policy is used
|
|
|
|
ASSERT_EQ("", props.filter_policy_name);
|
2016-05-12 18:47:16 +02:00
|
|
|
// Compression type == that set:
|
|
|
|
ASSERT_EQ("NoCompression", props.compression_name);
|
2016-04-21 19:16:28 +02:00
|
|
|
c.ResetTableReader();
|
|
|
|
}
|
|
|
|
|
|
|
|
{
|
|
|
|
Options options;
|
2018-06-05 04:59:44 +02:00
|
|
|
BlockBasedTableOptions table_options = GetBlockBasedTableOptions();
|
2016-04-21 19:16:28 +02:00
|
|
|
options.table_factory.reset(NewBlockBasedTableFactory(table_options));
|
|
|
|
options.comparator = &reverse_key_comparator;
|
|
|
|
options.merge_operator = MergeOperators::CreateUInt64AddOperator();
|
2016-08-26 20:46:32 +02:00
|
|
|
options.prefix_extractor.reset(NewNoopTransform());
|
2016-04-21 19:16:28 +02:00
|
|
|
options.table_properties_collector_factories.emplace_back(
|
|
|
|
new DummyPropertiesCollectorFactory1());
|
|
|
|
options.table_properties_collector_factories.emplace_back(
|
|
|
|
new DummyPropertiesCollectorFactory2());
|
|
|
|
|
|
|
|
const ImmutableCFOptions ioptions(options);
|
2018-05-21 23:33:55 +02:00
|
|
|
const MutableCFOptions moptions(options);
|
|
|
|
c.Finish(options, ioptions, moptions, table_options,
|
2016-04-21 19:16:28 +02:00
|
|
|
GetPlainInternalComparator(options.comparator), &keys, &kvmap);
|
|
|
|
|
|
|
|
auto& props = *c.GetTableReader()->GetTableProperties();
|
|
|
|
|
|
|
|
ASSERT_EQ("rocksdb.ReverseBytewiseComparator", props.comparator_name);
|
|
|
|
ASSERT_EQ("UInt64AddOperator", props.merge_operator_name);
|
2016-08-26 20:46:32 +02:00
|
|
|
ASSERT_EQ("rocksdb.Noop", props.prefix_extractor_name);
|
2020-10-02 00:21:17 +02:00
|
|
|
ASSERT_EQ(
|
|
|
|
"[DummyPropertiesCollectorFactory1,DummyPropertiesCollectorFactory2]",
|
|
|
|
props.property_collectors_names);
|
2016-04-21 19:16:28 +02:00
|
|
|
ASSERT_EQ("", props.filter_policy_name); // no filter policy is used
|
|
|
|
c.ResetTableReader();
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2018-06-05 04:59:44 +02:00
|
|
|
TEST_P(BlockBasedTableTest, RangeDelBlock) {
|
2016-08-20 00:10:31 +02:00
|
|
|
TableConstructor c(BytewiseComparator());
|
|
|
|
std::vector<std::string> keys = {"1pika", "2chu"};
|
|
|
|
std::vector<std::string> vals = {"p", "c"};
|
|
|
|
|
Cache fragmented range tombstones in BlockBasedTableReader (#4493)
Summary:
This allows tombstone fragmenting to only be performed when the table is opened, and cached for subsequent accesses.
On the same DB used in #4449, running `readrandom` results in the following:
```
readrandom : 0.983 micros/op 1017076 ops/sec; 78.3 MB/s (63103 of 100000 found)
```
Now that Get performance in the presence of range tombstones is reasonable, I also compared the performance between a DB with range tombstones, "expanded" range tombstones (several point tombstones that cover the same keys the equivalent range tombstone would cover, a common workaround for DeleteRange), and no range tombstones. The created DBs had 5 million keys each, and DeleteRange was called at regular intervals (depending on the total number of range tombstones being written) after 4.5 million Puts. The table below summarizes the results of a `readwhilewriting` benchmark (in order to provide somewhat more realistic results):
```
Tombstones? | avg micros/op | stddev micros/op | avg ops/s | stddev ops/s
----------------- | ------------- | ---------------- | ------------ | ------------
None | 0.6186 | 0.04637 | 1,625,252.90 | 124,679.41
500 Expanded | 0.6019 | 0.03628 | 1,666,670.40 | 101,142.65
500 Unexpanded | 0.6435 | 0.03994 | 1,559,979.40 | 104,090.52
1k Expanded | 0.6034 | 0.04349 | 1,665,128.10 | 125,144.57
1k Unexpanded | 0.6261 | 0.03093 | 1,600,457.50 | 79,024.94
5k Expanded | 0.6163 | 0.05926 | 1,636,668.80 | 154,888.85
5k Unexpanded | 0.6402 | 0.04002 | 1,567,804.70 | 100,965.55
10k Expanded | 0.6036 | 0.05105 | 1,667,237.70 | 142,830.36
10k Unexpanded | 0.6128 | 0.02598 | 1,634,633.40 | 72,161.82
25k Expanded | 0.6198 | 0.04542 | 1,620,980.50 | 116,662.93
25k Unexpanded | 0.5478 | 0.0362 | 1,833,059.10 | 121,233.81
50k Expanded | 0.5104 | 0.04347 | 1,973,107.90 | 184,073.49
50k Unexpanded | 0.4528 | 0.03387 | 2,219,034.50 | 170,984.32
```
After a large enough quantity of range tombstones are written, range tombstone Gets can become faster than reading from an equivalent DB with several point tombstones.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4493
Differential Revision: D10842844
Pulled By: abhimadan
fbshipit-source-id: a7d44534f8120e6aabb65779d26c6b9df954c509
2018-10-26 04:25:00 +02:00
|
|
|
std::vector<RangeTombstone> expected_tombstones = {
|
|
|
|
{"1pika", "2chu", 0},
|
|
|
|
{"2chu", "c", 1},
|
|
|
|
{"2chu", "c", 0},
|
|
|
|
{"c", "p", 0},
|
|
|
|
};
|
|
|
|
|
2016-08-20 00:10:31 +02:00
|
|
|
for (int i = 0; i < 2; i++) {
|
|
|
|
RangeTombstone t(keys[i], vals[i], i);
|
|
|
|
std::pair<InternalKey, Slice> p = t.Serialize();
|
|
|
|
c.Add(p.first.Encode().ToString(), p.second);
|
|
|
|
}
|
|
|
|
|
|
|
|
std::vector<std::string> sorted_keys;
|
|
|
|
stl_wrappers::KVMap kvmap;
|
|
|
|
Options options;
|
|
|
|
options.compression = kNoCompression;
|
2018-06-05 04:59:44 +02:00
|
|
|
BlockBasedTableOptions table_options = GetBlockBasedTableOptions();
|
2016-08-20 00:10:31 +02:00
|
|
|
table_options.block_restart_interval = 1;
|
|
|
|
options.table_factory.reset(NewBlockBasedTableFactory(table_options));
|
|
|
|
|
|
|
|
const ImmutableCFOptions ioptions(options);
|
2018-05-21 23:33:55 +02:00
|
|
|
const MutableCFOptions moptions(options);
|
2016-08-20 00:10:31 +02:00
|
|
|
std::unique_ptr<InternalKeyComparator> internal_cmp(
|
|
|
|
new InternalKeyComparator(options.comparator));
|
2018-05-21 23:33:55 +02:00
|
|
|
c.Finish(options, ioptions, moptions, table_options, *internal_cmp,
|
|
|
|
&sorted_keys, &kvmap);
|
2016-08-20 00:10:31 +02:00
|
|
|
|
2016-11-05 17:10:51 +01:00
|
|
|
for (int j = 0; j < 2; ++j) {
|
|
|
|
std::unique_ptr<InternalIterator> iter(
|
|
|
|
c.GetTableReader()->NewRangeTombstoneIterator(ReadOptions()));
|
|
|
|
if (j > 0) {
|
|
|
|
// For second iteration, delete the table reader object and verify the
|
|
|
|
// iterator can still access its metablock's range tombstones.
|
|
|
|
c.ResetTableReader();
|
|
|
|
}
|
2017-03-09 07:16:45 +01:00
|
|
|
ASSERT_FALSE(iter->Valid());
|
2016-11-05 17:10:51 +01:00
|
|
|
iter->SeekToFirst();
|
2017-03-09 07:16:45 +01:00
|
|
|
ASSERT_TRUE(iter->Valid());
|
Cache fragmented range tombstones in BlockBasedTableReader (#4493)
Summary:
This allows tombstone fragmenting to only be performed when the table is opened, and cached for subsequent accesses.
On the same DB used in #4449, running `readrandom` results in the following:
```
readrandom : 0.983 micros/op 1017076 ops/sec; 78.3 MB/s (63103 of 100000 found)
```
Now that Get performance in the presence of range tombstones is reasonable, I also compared the performance between a DB with range tombstones, "expanded" range tombstones (several point tombstones that cover the same keys the equivalent range tombstone would cover, a common workaround for DeleteRange), and no range tombstones. The created DBs had 5 million keys each, and DeleteRange was called at regular intervals (depending on the total number of range tombstones being written) after 4.5 million Puts. The table below summarizes the results of a `readwhilewriting` benchmark (in order to provide somewhat more realistic results):
```
Tombstones? | avg micros/op | stddev micros/op | avg ops/s | stddev ops/s
----------------- | ------------- | ---------------- | ------------ | ------------
None | 0.6186 | 0.04637 | 1,625,252.90 | 124,679.41
500 Expanded | 0.6019 | 0.03628 | 1,666,670.40 | 101,142.65
500 Unexpanded | 0.6435 | 0.03994 | 1,559,979.40 | 104,090.52
1k Expanded | 0.6034 | 0.04349 | 1,665,128.10 | 125,144.57
1k Unexpanded | 0.6261 | 0.03093 | 1,600,457.50 | 79,024.94
5k Expanded | 0.6163 | 0.05926 | 1,636,668.80 | 154,888.85
5k Unexpanded | 0.6402 | 0.04002 | 1,567,804.70 | 100,965.55
10k Expanded | 0.6036 | 0.05105 | 1,667,237.70 | 142,830.36
10k Unexpanded | 0.6128 | 0.02598 | 1,634,633.40 | 72,161.82
25k Expanded | 0.6198 | 0.04542 | 1,620,980.50 | 116,662.93
25k Unexpanded | 0.5478 | 0.0362 | 1,833,059.10 | 121,233.81
50k Expanded | 0.5104 | 0.04347 | 1,973,107.90 | 184,073.49
50k Unexpanded | 0.4528 | 0.03387 | 2,219,034.50 | 170,984.32
```
After a large enough quantity of range tombstones are written, range tombstone Gets can become faster than reading from an equivalent DB with several point tombstones.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4493
Differential Revision: D10842844
Pulled By: abhimadan
fbshipit-source-id: a7d44534f8120e6aabb65779d26c6b9df954c509
2018-10-26 04:25:00 +02:00
|
|
|
for (size_t i = 0; i < expected_tombstones.size(); i++) {
|
2016-11-05 17:10:51 +01:00
|
|
|
ASSERT_TRUE(iter->Valid());
|
|
|
|
ParsedInternalKey parsed_key;
|
2020-10-28 18:11:13 +01:00
|
|
|
ASSERT_OK(
|
|
|
|
ParseInternalKey(iter->key(), &parsed_key, true /* log_err_key */));
|
2016-11-05 17:10:51 +01:00
|
|
|
RangeTombstone t(parsed_key, iter->value());
|
Cache fragmented range tombstones in BlockBasedTableReader (#4493)
Summary:
This allows tombstone fragmenting to only be performed when the table is opened, and cached for subsequent accesses.
On the same DB used in #4449, running `readrandom` results in the following:
```
readrandom : 0.983 micros/op 1017076 ops/sec; 78.3 MB/s (63103 of 100000 found)
```
Now that Get performance in the presence of range tombstones is reasonable, I also compared the performance between a DB with range tombstones, "expanded" range tombstones (several point tombstones that cover the same keys the equivalent range tombstone would cover, a common workaround for DeleteRange), and no range tombstones. The created DBs had 5 million keys each, and DeleteRange was called at regular intervals (depending on the total number of range tombstones being written) after 4.5 million Puts. The table below summarizes the results of a `readwhilewriting` benchmark (in order to provide somewhat more realistic results):
```
Tombstones? | avg micros/op | stddev micros/op | avg ops/s | stddev ops/s
----------------- | ------------- | ---------------- | ------------ | ------------
None | 0.6186 | 0.04637 | 1,625,252.90 | 124,679.41
500 Expanded | 0.6019 | 0.03628 | 1,666,670.40 | 101,142.65
500 Unexpanded | 0.6435 | 0.03994 | 1,559,979.40 | 104,090.52
1k Expanded | 0.6034 | 0.04349 | 1,665,128.10 | 125,144.57
1k Unexpanded | 0.6261 | 0.03093 | 1,600,457.50 | 79,024.94
5k Expanded | 0.6163 | 0.05926 | 1,636,668.80 | 154,888.85
5k Unexpanded | 0.6402 | 0.04002 | 1,567,804.70 | 100,965.55
10k Expanded | 0.6036 | 0.05105 | 1,667,237.70 | 142,830.36
10k Unexpanded | 0.6128 | 0.02598 | 1,634,633.40 | 72,161.82
25k Expanded | 0.6198 | 0.04542 | 1,620,980.50 | 116,662.93
25k Unexpanded | 0.5478 | 0.0362 | 1,833,059.10 | 121,233.81
50k Expanded | 0.5104 | 0.04347 | 1,973,107.90 | 184,073.49
50k Unexpanded | 0.4528 | 0.03387 | 2,219,034.50 | 170,984.32
```
After a large enough quantity of range tombstones are written, range tombstone Gets can become faster than reading from an equivalent DB with several point tombstones.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/4493
Differential Revision: D10842844
Pulled By: abhimadan
fbshipit-source-id: a7d44534f8120e6aabb65779d26c6b9df954c509
2018-10-26 04:25:00 +02:00
|
|
|
const auto& expected_t = expected_tombstones[i];
|
|
|
|
ASSERT_EQ(t.start_key_, expected_t.start_key_);
|
|
|
|
ASSERT_EQ(t.end_key_, expected_t.end_key_);
|
|
|
|
ASSERT_EQ(t.seq_, expected_t.seq_);
|
2016-11-05 17:10:51 +01:00
|
|
|
iter->Next();
|
|
|
|
}
|
|
|
|
ASSERT_TRUE(!iter->Valid());
|
|
|
|
}
|
2016-08-20 00:10:31 +02:00
|
|
|
}
|
|
|
|
|
2018-06-05 04:59:44 +02:00
|
|
|
TEST_P(BlockBasedTableTest, FilterPolicyNameProperties) {
|
2016-08-20 00:10:31 +02:00
|
|
|
TableConstructor c(BytewiseComparator(), true /* convert_to_internal_key_ */);
|
2013-10-17 01:57:20 +02:00
|
|
|
c.Add("a1", "val1");
|
|
|
|
std::vector<std::string> keys;
|
2015-09-02 22:58:22 +02:00
|
|
|
stl_wrappers::KVMap kvmap;
|
2018-06-05 04:59:44 +02:00
|
|
|
BlockBasedTableOptions table_options = GetBlockBasedTableOptions();
|
2014-08-25 23:22:05 +02:00
|
|
|
table_options.filter_policy.reset(NewBloomFilterPolicy(10));
|
2013-11-20 07:00:48 +01:00
|
|
|
Options options;
|
2014-08-25 23:22:05 +02:00
|
|
|
options.table_factory.reset(NewBlockBasedTableFactory(table_options));
|
2013-10-17 01:57:20 +02:00
|
|
|
|
2014-09-05 01:18:36 +02:00
|
|
|
const ImmutableCFOptions ioptions(options);
|
2018-05-21 23:33:55 +02:00
|
|
|
const MutableCFOptions moptions(options);
|
|
|
|
c.Finish(options, ioptions, moptions, table_options,
|
2014-08-25 23:22:05 +02:00
|
|
|
GetPlainInternalComparator(options.comparator), &keys, &kvmap);
|
2014-08-26 01:14:30 +02:00
|
|
|
auto& props = *c.GetTableReader()->GetTableProperties();
|
2013-11-20 01:29:42 +01:00
|
|
|
ASSERT_EQ("rocksdb.BuiltinBloomFilter", props.filter_policy_name);
|
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 19:42:39 +02:00
|
|
|
c.ResetTableReader();
|
2013-10-17 01:57:20 +02:00
|
|
|
}
|
|
|
|
|
2015-03-03 02:07:03 +01:00
|
|
|
//
|
|
|
|
// BlockBasedTableTest::PrefetchTest
|
|
|
|
//
|
|
|
|
void AssertKeysInCache(BlockBasedTable* table_reader,
|
2015-09-02 22:58:22 +02:00
|
|
|
const std::vector<std::string>& keys_in_cache,
|
2016-08-20 00:10:31 +02:00
|
|
|
const std::vector<std::string>& keys_not_in_cache,
|
|
|
|
bool convert = false) {
|
|
|
|
if (convert) {
|
|
|
|
for (auto key : keys_in_cache) {
|
|
|
|
InternalKey ikey(key, kMaxSequenceNumber, kTypeValue);
|
|
|
|
ASSERT_TRUE(table_reader->TEST_KeyInCache(ReadOptions(), ikey.Encode()));
|
|
|
|
}
|
|
|
|
for (auto key : keys_not_in_cache) {
|
|
|
|
InternalKey ikey(key, kMaxSequenceNumber, kTypeValue);
|
|
|
|
ASSERT_TRUE(!table_reader->TEST_KeyInCache(ReadOptions(), ikey.Encode()));
|
|
|
|
}
|
|
|
|
} else {
|
|
|
|
for (auto key : keys_in_cache) {
|
|
|
|
ASSERT_TRUE(table_reader->TEST_KeyInCache(ReadOptions(), key));
|
|
|
|
}
|
|
|
|
for (auto key : keys_not_in_cache) {
|
|
|
|
ASSERT_TRUE(!table_reader->TEST_KeyInCache(ReadOptions(), key));
|
|
|
|
}
|
2015-03-03 02:07:03 +01:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
void PrefetchRange(TableConstructor* c, Options* opt,
|
2016-08-20 00:10:31 +02:00
|
|
|
BlockBasedTableOptions* table_options, const char* key_begin,
|
2015-09-02 22:58:22 +02:00
|
|
|
const char* key_end,
|
|
|
|
const std::vector<std::string>& keys_in_cache,
|
|
|
|
const std::vector<std::string>& keys_not_in_cache,
|
2015-03-03 02:07:03 +01:00
|
|
|
const Status expected_status = Status::OK()) {
|
|
|
|
// reset the cache and reopen the table
|
2016-04-07 22:51:47 +02:00
|
|
|
table_options->block_cache = NewLRUCache(16 * 1024 * 1024, 4);
|
2015-03-03 02:07:03 +01:00
|
|
|
opt->table_factory.reset(NewBlockBasedTableFactory(*table_options));
|
|
|
|
const ImmutableCFOptions ioptions2(*opt);
|
2018-05-21 23:33:55 +02:00
|
|
|
const MutableCFOptions moptions(*opt);
|
|
|
|
ASSERT_OK(c->Reopen(ioptions2, moptions));
|
2015-03-03 02:07:03 +01:00
|
|
|
|
|
|
|
// prefetch
|
|
|
|
auto* table_reader = dynamic_cast<BlockBasedTable*>(c->GetTableReader());
|
2016-08-20 00:10:31 +02:00
|
|
|
Status s;
|
2018-11-09 20:17:34 +01:00
|
|
|
std::unique_ptr<Slice> begin, end;
|
|
|
|
std::unique_ptr<InternalKey> i_begin, i_end;
|
2016-08-20 00:10:31 +02:00
|
|
|
if (key_begin != nullptr) {
|
|
|
|
if (c->ConvertToInternalKey()) {
|
|
|
|
i_begin.reset(new InternalKey(key_begin, kMaxSequenceNumber, kTypeValue));
|
|
|
|
begin.reset(new Slice(i_begin->Encode()));
|
|
|
|
} else {
|
|
|
|
begin.reset(new Slice(key_begin));
|
|
|
|
}
|
|
|
|
}
|
|
|
|
if (key_end != nullptr) {
|
|
|
|
if (c->ConvertToInternalKey()) {
|
|
|
|
i_end.reset(new InternalKey(key_end, kMaxSequenceNumber, kTypeValue));
|
|
|
|
end.reset(new Slice(i_end->Encode()));
|
|
|
|
} else {
|
|
|
|
end.reset(new Slice(key_end));
|
|
|
|
}
|
|
|
|
}
|
|
|
|
s = table_reader->Prefetch(begin.get(), end.get());
|
|
|
|
|
2015-03-03 02:07:03 +01:00
|
|
|
ASSERT_TRUE(s.code() == expected_status.code());
|
|
|
|
|
|
|
|
// assert our expectation in cache warmup
|
2016-08-20 00:10:31 +02:00
|
|
|
AssertKeysInCache(table_reader, keys_in_cache, keys_not_in_cache,
|
|
|
|
c->ConvertToInternalKey());
|
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 19:42:39 +02:00
|
|
|
c->ResetTableReader();
|
2015-03-03 02:07:03 +01:00
|
|
|
}
|
|
|
|
|
2018-06-05 04:59:44 +02:00
|
|
|
TEST_P(BlockBasedTableTest, PrefetchTest) {
|
2015-03-03 02:07:03 +01:00
|
|
|
// The purpose of this test is to test the prefetching operation built into
|
|
|
|
// BlockBasedTable.
|
|
|
|
Options opt;
|
2018-11-09 20:17:34 +01:00
|
|
|
std::unique_ptr<InternalKeyComparator> ikc;
|
2015-03-03 02:07:03 +01:00
|
|
|
ikc.reset(new test::PlainInternalKeyComparator(opt.comparator));
|
|
|
|
opt.compression = kNoCompression;
|
2018-06-05 04:59:44 +02:00
|
|
|
BlockBasedTableOptions table_options = GetBlockBasedTableOptions();
|
2015-03-03 02:07:03 +01:00
|
|
|
table_options.block_size = 1024;
|
|
|
|
// big enough so we don't ever lose cached values.
|
2016-04-07 22:51:47 +02:00
|
|
|
table_options.block_cache = NewLRUCache(16 * 1024 * 1024, 4);
|
2015-03-03 02:07:03 +01:00
|
|
|
opt.table_factory.reset(NewBlockBasedTableFactory(table_options));
|
|
|
|
|
2016-08-20 00:10:31 +02:00
|
|
|
TableConstructor c(BytewiseComparator(), true /* convert_to_internal_key_ */);
|
2015-03-03 02:07:03 +01:00
|
|
|
c.Add("k01", "hello");
|
|
|
|
c.Add("k02", "hello2");
|
|
|
|
c.Add("k03", std::string(10000, 'x'));
|
|
|
|
c.Add("k04", std::string(200000, 'x'));
|
|
|
|
c.Add("k05", std::string(300000, 'x'));
|
|
|
|
c.Add("k06", "hello3");
|
|
|
|
c.Add("k07", std::string(100000, 'x'));
|
|
|
|
std::vector<std::string> keys;
|
2015-09-02 22:58:22 +02:00
|
|
|
stl_wrappers::KVMap kvmap;
|
2015-03-03 02:07:03 +01:00
|
|
|
const ImmutableCFOptions ioptions(opt);
|
2018-05-21 23:33:55 +02:00
|
|
|
const MutableCFOptions moptions(opt);
|
|
|
|
c.Finish(opt, ioptions, moptions, table_options, *ikc, &keys, &kvmap);
|
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 19:42:39 +02:00
|
|
|
c.ResetTableReader();
|
2015-03-03 02:07:03 +01:00
|
|
|
|
|
|
|
// We get the following data spread :
|
|
|
|
//
|
|
|
|
// Data block Index
|
|
|
|
// ========================
|
|
|
|
// [ k01 k02 k03 ] k03
|
|
|
|
// [ k04 ] k04
|
|
|
|
// [ k05 ] k05
|
|
|
|
// [ k06 k07 ] k07
|
|
|
|
|
|
|
|
|
|
|
|
// Simple
|
2016-08-20 00:10:31 +02:00
|
|
|
PrefetchRange(&c, &opt, &table_options,
|
|
|
|
/*key_range=*/"k01", "k05",
|
|
|
|
/*keys_in_cache=*/{"k01", "k02", "k03", "k04", "k05"},
|
|
|
|
/*keys_not_in_cache=*/{"k06", "k07"});
|
|
|
|
PrefetchRange(&c, &opt, &table_options, "k01", "k01", {"k01", "k02", "k03"},
|
2015-03-03 02:07:03 +01:00
|
|
|
{"k04", "k05", "k06", "k07"});
|
|
|
|
// odd
|
2016-08-20 00:10:31 +02:00
|
|
|
PrefetchRange(&c, &opt, &table_options, "a", "z",
|
|
|
|
{"k01", "k02", "k03", "k04", "k05", "k06", "k07"}, {});
|
|
|
|
PrefetchRange(&c, &opt, &table_options, "k00", "k00", {"k01", "k02", "k03"},
|
2015-03-03 02:07:03 +01:00
|
|
|
{"k04", "k05", "k06", "k07"});
|
|
|
|
// Edge cases
|
2016-08-20 00:10:31 +02:00
|
|
|
PrefetchRange(&c, &opt, &table_options, "k00", "k06",
|
|
|
|
{"k01", "k02", "k03", "k04", "k05", "k06", "k07"}, {});
|
|
|
|
PrefetchRange(&c, &opt, &table_options, "k00", "zzz",
|
|
|
|
{"k01", "k02", "k03", "k04", "k05", "k06", "k07"}, {});
|
2015-03-03 02:07:03 +01:00
|
|
|
// null keys
|
2016-08-20 00:10:31 +02:00
|
|
|
PrefetchRange(&c, &opt, &table_options, nullptr, nullptr,
|
|
|
|
{"k01", "k02", "k03", "k04", "k05", "k06", "k07"}, {});
|
|
|
|
PrefetchRange(&c, &opt, &table_options, "k04", nullptr,
|
|
|
|
{"k04", "k05", "k06", "k07"}, {"k01", "k02", "k03"});
|
|
|
|
PrefetchRange(&c, &opt, &table_options, nullptr, "k05",
|
|
|
|
{"k01", "k02", "k03", "k04", "k05"}, {"k06", "k07"});
|
2015-03-03 02:07:03 +01:00
|
|
|
// invalid
|
2016-08-20 00:10:31 +02:00
|
|
|
PrefetchRange(&c, &opt, &table_options, "k06", "k00", {}, {},
|
2015-03-03 02:07:03 +01:00
|
|
|
Status::InvalidArgument(Slice("k06 "), Slice("k07")));
|
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 19:42:39 +02:00
|
|
|
c.ResetTableReader();
|
2015-03-03 02:07:03 +01:00
|
|
|
}
|
|
|
|
|
2018-06-05 04:59:44 +02:00
|
|
|
TEST_P(BlockBasedTableTest, TotalOrderSeekOnHashIndex) {
|
|
|
|
BlockBasedTableOptions table_options = GetBlockBasedTableOptions();
|
Add an option to put first key of each sst block in the index (#5289)
Summary:
The first key is used to defer reading the data block until this file gets to the top of merging iterator's heap. For short range scans, most files never make it to the top of the heap, so this change can reduce read amplification by a lot sometimes.
Consider the following workload. There are a few data streams (we'll be calling them "logs"), each stream consisting of a sequence of blobs (we'll be calling them "records"). Each record is identified by log ID and a sequence number within the log. RocksDB key is concatenation of log ID and sequence number (big endian). Reads are mostly relatively short range scans, each within a single log. Writes are mostly sequential for each log, but writes to different logs are randomly interleaved. Compactions are disabled; instead, when we accumulate a few tens of sst files, we create a new column family and start writing to it.
So, a typical sst file consists of a few ranges of blocks, each range corresponding to one log ID (we use FlushBlockPolicy to cut blocks at log boundaries). A typical read would go like this. First, iterator Seek() reads one block from each sst file. Then a series of Next()s move through one sst file (since writes to each log are mostly sequential) until the subiterator reaches the end of this log in this sst file; then Next() switches to the next sst file and reads sequentially from that, and so on. Often a range scan will only return records from a small number of blocks in small number of sst files; in this case, the cost of initial Seek() reading one block from each file may be bigger than the cost of reading the actually useful blocks.
Neither iterate_upper_bound nor bloom filters can prevent reading one block from each file in Seek(). But this PR can: if the index contains first key from each block, we don't have to read the block until this block actually makes it to the top of merging iterator's heap, so for short range scans we won't read any blocks from most of the sst files.
This PR does the deferred block loading inside value() call. This is not ideal: there's no good way to report an IO error from inside value(). As discussed with siying offline, it would probably be better to change InternalIterator's interface to explicitly fetch deferred value and get status. I'll do it in a separate PR.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/5289
Differential Revision: D15256423
Pulled By: al13n321
fbshipit-source-id: 750e4c39ce88e8d41662f701cf6275d9388ba46a
2019-06-25 05:50:35 +02:00
|
|
|
for (int i = 0; i <= 5; ++i) {
|
2014-08-26 01:14:30 +02:00
|
|
|
Options options;
|
|
|
|
// Make each key/value an individual block
|
|
|
|
table_options.block_size = 64;
|
|
|
|
switch (i) {
|
|
|
|
case 0:
|
|
|
|
// Binary search index
|
|
|
|
table_options.index_type = BlockBasedTableOptions::kBinarySearch;
|
|
|
|
options.table_factory.reset(new BlockBasedTableFactory(table_options));
|
|
|
|
break;
|
|
|
|
case 1:
|
|
|
|
// Hash search index
|
|
|
|
table_options.index_type = BlockBasedTableOptions::kHashSearch;
|
|
|
|
options.table_factory.reset(new BlockBasedTableFactory(table_options));
|
|
|
|
options.prefix_extractor.reset(NewFixedPrefixTransform(4));
|
|
|
|
break;
|
|
|
|
case 2:
|
|
|
|
// Hash search index with hash_index_allow_collision
|
|
|
|
table_options.index_type = BlockBasedTableOptions::kHashSearch;
|
|
|
|
table_options.hash_index_allow_collision = true;
|
|
|
|
options.table_factory.reset(new BlockBasedTableFactory(table_options));
|
|
|
|
options.prefix_extractor.reset(NewFixedPrefixTransform(4));
|
|
|
|
break;
|
|
|
|
case 3:
|
|
|
|
// Hash search index with filter policy
|
|
|
|
table_options.index_type = BlockBasedTableOptions::kHashSearch;
|
|
|
|
table_options.filter_policy.reset(NewBloomFilterPolicy(10));
|
|
|
|
options.table_factory.reset(new BlockBasedTableFactory(table_options));
|
|
|
|
options.prefix_extractor.reset(NewFixedPrefixTransform(4));
|
|
|
|
break;
|
2017-02-07 01:29:29 +01:00
|
|
|
case 4:
|
Add an option to put first key of each sst block in the index (#5289)
Summary:
The first key is used to defer reading the data block until this file gets to the top of merging iterator's heap. For short range scans, most files never make it to the top of the heap, so this change can reduce read amplification by a lot sometimes.
Consider the following workload. There are a few data streams (we'll be calling them "logs"), each stream consisting of a sequence of blobs (we'll be calling them "records"). Each record is identified by log ID and a sequence number within the log. RocksDB key is concatenation of log ID and sequence number (big endian). Reads are mostly relatively short range scans, each within a single log. Writes are mostly sequential for each log, but writes to different logs are randomly interleaved. Compactions are disabled; instead, when we accumulate a few tens of sst files, we create a new column family and start writing to it.
So, a typical sst file consists of a few ranges of blocks, each range corresponding to one log ID (we use FlushBlockPolicy to cut blocks at log boundaries). A typical read would go like this. First, iterator Seek() reads one block from each sst file. Then a series of Next()s move through one sst file (since writes to each log are mostly sequential) until the subiterator reaches the end of this log in this sst file; then Next() switches to the next sst file and reads sequentially from that, and so on. Often a range scan will only return records from a small number of blocks in small number of sst files; in this case, the cost of initial Seek() reading one block from each file may be bigger than the cost of reading the actually useful blocks.
Neither iterate_upper_bound nor bloom filters can prevent reading one block from each file in Seek(). But this PR can: if the index contains first key from each block, we don't have to read the block until this block actually makes it to the top of merging iterator's heap, so for short range scans we won't read any blocks from most of the sst files.
This PR does the deferred block loading inside value() call. This is not ideal: there's no good way to report an IO error from inside value(). As discussed with siying offline, it would probably be better to change InternalIterator's interface to explicitly fetch deferred value and get status. I'll do it in a separate PR.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/5289
Differential Revision: D15256423
Pulled By: al13n321
fbshipit-source-id: 750e4c39ce88e8d41662f701cf6275d9388ba46a
2019-06-25 05:50:35 +02:00
|
|
|
// Two-level index
|
2017-02-07 01:29:29 +01:00
|
|
|
table_options.index_type = BlockBasedTableOptions::kTwoLevelIndexSearch;
|
|
|
|
options.table_factory.reset(new BlockBasedTableFactory(table_options));
|
|
|
|
break;
|
Add an option to put first key of each sst block in the index (#5289)
Summary:
The first key is used to defer reading the data block until this file gets to the top of merging iterator's heap. For short range scans, most files never make it to the top of the heap, so this change can reduce read amplification by a lot sometimes.
Consider the following workload. There are a few data streams (we'll be calling them "logs"), each stream consisting of a sequence of blobs (we'll be calling them "records"). Each record is identified by log ID and a sequence number within the log. RocksDB key is concatenation of log ID and sequence number (big endian). Reads are mostly relatively short range scans, each within a single log. Writes are mostly sequential for each log, but writes to different logs are randomly interleaved. Compactions are disabled; instead, when we accumulate a few tens of sst files, we create a new column family and start writing to it.
So, a typical sst file consists of a few ranges of blocks, each range corresponding to one log ID (we use FlushBlockPolicy to cut blocks at log boundaries). A typical read would go like this. First, iterator Seek() reads one block from each sst file. Then a series of Next()s move through one sst file (since writes to each log are mostly sequential) until the subiterator reaches the end of this log in this sst file; then Next() switches to the next sst file and reads sequentially from that, and so on. Often a range scan will only return records from a small number of blocks in small number of sst files; in this case, the cost of initial Seek() reading one block from each file may be bigger than the cost of reading the actually useful blocks.
Neither iterate_upper_bound nor bloom filters can prevent reading one block from each file in Seek(). But this PR can: if the index contains first key from each block, we don't have to read the block until this block actually makes it to the top of merging iterator's heap, so for short range scans we won't read any blocks from most of the sst files.
This PR does the deferred block loading inside value() call. This is not ideal: there's no good way to report an IO error from inside value(). As discussed with siying offline, it would probably be better to change InternalIterator's interface to explicitly fetch deferred value and get status. I'll do it in a separate PR.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/5289
Differential Revision: D15256423
Pulled By: al13n321
fbshipit-source-id: 750e4c39ce88e8d41662f701cf6275d9388ba46a
2019-06-25 05:50:35 +02:00
|
|
|
case 5:
|
|
|
|
// Binary search with first key
|
|
|
|
table_options.index_type =
|
|
|
|
BlockBasedTableOptions::kBinarySearchWithFirstKey;
|
|
|
|
options.table_factory.reset(new BlockBasedTableFactory(table_options));
|
|
|
|
break;
|
2014-08-26 01:14:30 +02:00
|
|
|
}
|
|
|
|
|
2016-08-20 00:10:31 +02:00
|
|
|
TableConstructor c(BytewiseComparator(),
|
|
|
|
true /* convert_to_internal_key_ */);
|
2014-08-26 01:14:30 +02:00
|
|
|
c.Add("aaaa1", std::string('a', 56));
|
|
|
|
c.Add("bbaa1", std::string('a', 56));
|
|
|
|
c.Add("cccc1", std::string('a', 56));
|
|
|
|
c.Add("bbbb1", std::string('a', 56));
|
|
|
|
c.Add("baaa1", std::string('a', 56));
|
|
|
|
c.Add("abbb1", std::string('a', 56));
|
|
|
|
c.Add("cccc2", std::string('a', 56));
|
|
|
|
std::vector<std::string> keys;
|
2015-09-02 22:58:22 +02:00
|
|
|
stl_wrappers::KVMap kvmap;
|
2014-09-05 01:18:36 +02:00
|
|
|
const ImmutableCFOptions ioptions(options);
|
2018-05-21 23:33:55 +02:00
|
|
|
const MutableCFOptions moptions(options);
|
|
|
|
c.Finish(options, ioptions, moptions, table_options,
|
2014-08-26 01:14:30 +02:00
|
|
|
GetPlainInternalComparator(options.comparator), &keys, &kvmap);
|
|
|
|
auto props = c.GetTableReader()->GetTableProperties();
|
|
|
|
ASSERT_EQ(7u, props->num_data_blocks);
|
|
|
|
auto* reader = c.GetTableReader();
|
|
|
|
ReadOptions ro;
|
|
|
|
ro.total_order_seek = true;
|
2019-06-20 23:28:22 +02:00
|
|
|
std::unique_ptr<InternalIterator> iter(reader->NewIterator(
|
|
|
|
ro, moptions.prefix_extractor.get(), /*arena=*/nullptr,
|
|
|
|
/*skip_filters=*/false, TableReaderCaller::kUncategorized));
|
2014-08-26 01:14:30 +02:00
|
|
|
|
|
|
|
iter->Seek(InternalKey("b", 0, kTypeValue).Encode());
|
|
|
|
ASSERT_OK(iter->status());
|
|
|
|
ASSERT_TRUE(iter->Valid());
|
|
|
|
ASSERT_EQ("baaa1", ExtractUserKey(iter->key()).ToString());
|
|
|
|
iter->Next();
|
|
|
|
ASSERT_OK(iter->status());
|
|
|
|
ASSERT_TRUE(iter->Valid());
|
|
|
|
ASSERT_EQ("bbaa1", ExtractUserKey(iter->key()).ToString());
|
|
|
|
|
|
|
|
iter->Seek(InternalKey("bb", 0, kTypeValue).Encode());
|
|
|
|
ASSERT_OK(iter->status());
|
|
|
|
ASSERT_TRUE(iter->Valid());
|
|
|
|
ASSERT_EQ("bbaa1", ExtractUserKey(iter->key()).ToString());
|
|
|
|
iter->Next();
|
|
|
|
ASSERT_OK(iter->status());
|
|
|
|
ASSERT_TRUE(iter->Valid());
|
|
|
|
ASSERT_EQ("bbbb1", ExtractUserKey(iter->key()).ToString());
|
|
|
|
|
|
|
|
iter->Seek(InternalKey("bbb", 0, kTypeValue).Encode());
|
|
|
|
ASSERT_OK(iter->status());
|
|
|
|
ASSERT_TRUE(iter->Valid());
|
|
|
|
ASSERT_EQ("bbbb1", ExtractUserKey(iter->key()).ToString());
|
|
|
|
iter->Next();
|
|
|
|
ASSERT_OK(iter->status());
|
|
|
|
ASSERT_TRUE(iter->Valid());
|
|
|
|
ASSERT_EQ("cccc1", ExtractUserKey(iter->key()).ToString());
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2018-06-05 04:59:44 +02:00
|
|
|
TEST_P(BlockBasedTableTest, NoopTransformSeek) {
|
|
|
|
BlockBasedTableOptions table_options = GetBlockBasedTableOptions();
|
2016-01-06 20:46:32 +01:00
|
|
|
table_options.filter_policy.reset(NewBloomFilterPolicy(10));
|
|
|
|
|
|
|
|
Options options;
|
|
|
|
options.comparator = BytewiseComparator();
|
|
|
|
options.table_factory.reset(new BlockBasedTableFactory(table_options));
|
|
|
|
options.prefix_extractor.reset(NewNoopTransform());
|
|
|
|
|
|
|
|
TableConstructor c(options.comparator);
|
|
|
|
// To tickle the PrefixMayMatch bug it is important that the
|
|
|
|
// user-key is a single byte so that the index key exactly matches
|
|
|
|
// the user-key.
|
|
|
|
InternalKey key("a", 1, kTypeValue);
|
|
|
|
c.Add(key.Encode().ToString(), "b");
|
|
|
|
std::vector<std::string> keys;
|
|
|
|
stl_wrappers::KVMap kvmap;
|
|
|
|
const ImmutableCFOptions ioptions(options);
|
2018-05-21 23:33:55 +02:00
|
|
|
const MutableCFOptions moptions(options);
|
2016-01-07 18:48:29 +01:00
|
|
|
const InternalKeyComparator internal_comparator(options.comparator);
|
2018-05-21 23:33:55 +02:00
|
|
|
c.Finish(options, ioptions, moptions, table_options, internal_comparator,
|
|
|
|
&keys, &kvmap);
|
2016-01-06 20:46:32 +01:00
|
|
|
|
|
|
|
auto* reader = c.GetTableReader();
|
|
|
|
for (int i = 0; i < 2; ++i) {
|
|
|
|
ReadOptions ro;
|
|
|
|
ro.total_order_seek = (i == 0);
|
2019-06-20 23:28:22 +02:00
|
|
|
std::unique_ptr<InternalIterator> iter(reader->NewIterator(
|
|
|
|
ro, moptions.prefix_extractor.get(), /*arena=*/nullptr,
|
|
|
|
/*skip_filters=*/false, TableReaderCaller::kUncategorized));
|
2016-01-06 20:46:32 +01:00
|
|
|
|
|
|
|
iter->Seek(key.Encode());
|
|
|
|
ASSERT_OK(iter->status());
|
|
|
|
ASSERT_TRUE(iter->Valid());
|
|
|
|
ASSERT_EQ("a", ExtractUserKey(iter->key()).ToString());
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2018-06-05 04:59:44 +02:00
|
|
|
TEST_P(BlockBasedTableTest, SkipPrefixBloomFilter) {
|
2016-08-26 20:46:32 +02:00
|
|
|
// if DB is opened with a prefix extractor of a different name,
|
|
|
|
// prefix bloom is skipped when read the file
|
2018-06-05 04:59:44 +02:00
|
|
|
BlockBasedTableOptions table_options = GetBlockBasedTableOptions();
|
2016-08-26 20:46:32 +02:00
|
|
|
table_options.filter_policy.reset(NewBloomFilterPolicy(2));
|
|
|
|
table_options.whole_key_filtering = false;
|
|
|
|
|
|
|
|
Options options;
|
|
|
|
options.comparator = BytewiseComparator();
|
|
|
|
options.table_factory.reset(new BlockBasedTableFactory(table_options));
|
|
|
|
options.prefix_extractor.reset(NewFixedPrefixTransform(1));
|
|
|
|
|
|
|
|
TableConstructor c(options.comparator);
|
|
|
|
InternalKey key("abcdefghijk", 1, kTypeValue);
|
|
|
|
c.Add(key.Encode().ToString(), "test");
|
|
|
|
std::vector<std::string> keys;
|
|
|
|
stl_wrappers::KVMap kvmap;
|
|
|
|
const ImmutableCFOptions ioptions(options);
|
2018-05-21 23:33:55 +02:00
|
|
|
const MutableCFOptions moptions(options);
|
2016-08-26 20:46:32 +02:00
|
|
|
const InternalKeyComparator internal_comparator(options.comparator);
|
2018-05-21 23:33:55 +02:00
|
|
|
c.Finish(options, ioptions, moptions, table_options, internal_comparator,
|
|
|
|
&keys, &kvmap);
|
|
|
|
// TODO(Zhongyi): update test to use MutableCFOptions
|
2016-08-26 20:46:32 +02:00
|
|
|
options.prefix_extractor.reset(NewFixedPrefixTransform(9));
|
|
|
|
const ImmutableCFOptions new_ioptions(options);
|
2018-05-21 23:33:55 +02:00
|
|
|
const MutableCFOptions new_moptions(options);
|
2020-10-08 20:11:52 +02:00
|
|
|
ASSERT_OK(c.Reopen(new_ioptions, new_moptions));
|
2016-08-26 20:46:32 +02:00
|
|
|
auto reader = c.GetTableReader();
|
2020-08-04 00:21:56 +02:00
|
|
|
ReadOptions read_options;
|
2019-06-20 23:28:22 +02:00
|
|
|
std::unique_ptr<InternalIterator> db_iter(reader->NewIterator(
|
2020-08-04 00:21:56 +02:00
|
|
|
read_options, new_moptions.prefix_extractor.get(), /*arena=*/nullptr,
|
2019-06-20 23:28:22 +02:00
|
|
|
/*skip_filters=*/false, TableReaderCaller::kUncategorized));
|
2016-08-26 20:46:32 +02:00
|
|
|
|
|
|
|
// Test point lookup
|
|
|
|
// only one kv
|
|
|
|
for (auto& kv : kvmap) {
|
|
|
|
db_iter->Seek(kv.first);
|
|
|
|
ASSERT_TRUE(db_iter->Valid());
|
|
|
|
ASSERT_OK(db_iter->status());
|
|
|
|
ASSERT_EQ(db_iter->key(), kv.first);
|
|
|
|
ASSERT_EQ(db_iter->value(), kv.second);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2014-09-26 18:14:05 +02:00
|
|
|
void AddInternalKey(TableConstructor* c, const std::string& prefix,
|
Add an option to put first key of each sst block in the index (#5289)
Summary:
The first key is used to defer reading the data block until this file gets to the top of merging iterator's heap. For short range scans, most files never make it to the top of the heap, so this change can reduce read amplification by a lot sometimes.
Consider the following workload. There are a few data streams (we'll be calling them "logs"), each stream consisting of a sequence of blobs (we'll be calling them "records"). Each record is identified by log ID and a sequence number within the log. RocksDB key is concatenation of log ID and sequence number (big endian). Reads are mostly relatively short range scans, each within a single log. Writes are mostly sequential for each log, but writes to different logs are randomly interleaved. Compactions are disabled; instead, when we accumulate a few tens of sst files, we create a new column family and start writing to it.
So, a typical sst file consists of a few ranges of blocks, each range corresponding to one log ID (we use FlushBlockPolicy to cut blocks at log boundaries). A typical read would go like this. First, iterator Seek() reads one block from each sst file. Then a series of Next()s move through one sst file (since writes to each log are mostly sequential) until the subiterator reaches the end of this log in this sst file; then Next() switches to the next sst file and reads sequentially from that, and so on. Often a range scan will only return records from a small number of blocks in small number of sst files; in this case, the cost of initial Seek() reading one block from each file may be bigger than the cost of reading the actually useful blocks.
Neither iterate_upper_bound nor bloom filters can prevent reading one block from each file in Seek(). But this PR can: if the index contains first key from each block, we don't have to read the block until this block actually makes it to the top of merging iterator's heap, so for short range scans we won't read any blocks from most of the sst files.
This PR does the deferred block loading inside value() call. This is not ideal: there's no good way to report an IO error from inside value(). As discussed with siying offline, it would probably be better to change InternalIterator's interface to explicitly fetch deferred value and get status. I'll do it in a separate PR.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/5289
Differential Revision: D15256423
Pulled By: al13n321
fbshipit-source-id: 750e4c39ce88e8d41662f701cf6275d9388ba46a
2019-06-25 05:50:35 +02:00
|
|
|
std::string value = "v", int /*suffix_len*/ = 800) {
|
2014-04-10 23:19:43 +02:00
|
|
|
static Random rnd(1023);
|
2020-07-09 23:33:42 +02:00
|
|
|
InternalKey k(prefix + rnd.RandomString(800), 0, kTypeValue);
|
Add an option to put first key of each sst block in the index (#5289)
Summary:
The first key is used to defer reading the data block until this file gets to the top of merging iterator's heap. For short range scans, most files never make it to the top of the heap, so this change can reduce read amplification by a lot sometimes.
Consider the following workload. There are a few data streams (we'll be calling them "logs"), each stream consisting of a sequence of blobs (we'll be calling them "records"). Each record is identified by log ID and a sequence number within the log. RocksDB key is concatenation of log ID and sequence number (big endian). Reads are mostly relatively short range scans, each within a single log. Writes are mostly sequential for each log, but writes to different logs are randomly interleaved. Compactions are disabled; instead, when we accumulate a few tens of sst files, we create a new column family and start writing to it.
So, a typical sst file consists of a few ranges of blocks, each range corresponding to one log ID (we use FlushBlockPolicy to cut blocks at log boundaries). A typical read would go like this. First, iterator Seek() reads one block from each sst file. Then a series of Next()s move through one sst file (since writes to each log are mostly sequential) until the subiterator reaches the end of this log in this sst file; then Next() switches to the next sst file and reads sequentially from that, and so on. Often a range scan will only return records from a small number of blocks in small number of sst files; in this case, the cost of initial Seek() reading one block from each file may be bigger than the cost of reading the actually useful blocks.
Neither iterate_upper_bound nor bloom filters can prevent reading one block from each file in Seek(). But this PR can: if the index contains first key from each block, we don't have to read the block until this block actually makes it to the top of merging iterator's heap, so for short range scans we won't read any blocks from most of the sst files.
This PR does the deferred block loading inside value() call. This is not ideal: there's no good way to report an IO error from inside value(). As discussed with siying offline, it would probably be better to change InternalIterator's interface to explicitly fetch deferred value and get status. I'll do it in a separate PR.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/5289
Differential Revision: D15256423
Pulled By: al13n321
fbshipit-source-id: 750e4c39ce88e8d41662f701cf6275d9388ba46a
2019-06-25 05:50:35 +02:00
|
|
|
c->Add(k.Encode().ToString(), value);
|
2014-04-10 23:19:43 +02:00
|
|
|
}
|
|
|
|
|
2017-02-07 01:29:29 +01:00
|
|
|
void TableTest::IndexTest(BlockBasedTableOptions table_options) {
|
2014-04-10 23:19:43 +02:00
|
|
|
TableConstructor c(BytewiseComparator());
|
|
|
|
|
|
|
|
// keys with prefix length 3, make sure the key/value is big enough to fill
|
|
|
|
// one block
|
|
|
|
AddInternalKey(&c, "0015");
|
|
|
|
AddInternalKey(&c, "0035");
|
|
|
|
|
|
|
|
AddInternalKey(&c, "0054");
|
|
|
|
AddInternalKey(&c, "0055");
|
|
|
|
|
|
|
|
AddInternalKey(&c, "0056");
|
|
|
|
AddInternalKey(&c, "0057");
|
|
|
|
|
|
|
|
AddInternalKey(&c, "0058");
|
|
|
|
AddInternalKey(&c, "0075");
|
|
|
|
|
|
|
|
AddInternalKey(&c, "0076");
|
|
|
|
AddInternalKey(&c, "0095");
|
|
|
|
|
|
|
|
std::vector<std::string> keys;
|
2015-09-02 22:58:22 +02:00
|
|
|
stl_wrappers::KVMap kvmap;
|
2014-04-10 23:19:43 +02:00
|
|
|
Options options;
|
2014-08-25 23:22:05 +02:00
|
|
|
options.prefix_extractor.reset(NewFixedPrefixTransform(3));
|
|
|
|
table_options.block_size = 1700;
|
2016-04-07 22:51:47 +02:00
|
|
|
table_options.block_cache = NewLRUCache(1024, 4);
|
2014-08-25 23:22:05 +02:00
|
|
|
options.table_factory.reset(NewBlockBasedTableFactory(table_options));
|
2014-04-10 23:19:43 +02:00
|
|
|
|
|
|
|
std::unique_ptr<InternalKeyComparator> comparator(
|
|
|
|
new InternalKeyComparator(BytewiseComparator()));
|
2014-09-05 01:18:36 +02:00
|
|
|
const ImmutableCFOptions ioptions(options);
|
2018-05-21 23:33:55 +02:00
|
|
|
const MutableCFOptions moptions(options);
|
|
|
|
c.Finish(options, ioptions, moptions, table_options, *comparator, &keys,
|
|
|
|
&kvmap);
|
2014-08-26 01:14:30 +02:00
|
|
|
auto reader = c.GetTableReader();
|
2014-04-10 23:19:43 +02:00
|
|
|
|
2014-08-26 01:14:30 +02:00
|
|
|
auto props = reader->GetTableProperties();
|
2014-04-10 23:19:43 +02:00
|
|
|
ASSERT_EQ(5u, props->num_data_blocks);
|
|
|
|
|
2018-05-21 23:33:55 +02:00
|
|
|
// TODO(Zhongyi): update test to use MutableCFOptions
|
2020-08-04 00:21:56 +02:00
|
|
|
ReadOptions read_options;
|
2019-06-20 23:28:22 +02:00
|
|
|
std::unique_ptr<InternalIterator> index_iter(reader->NewIterator(
|
2020-08-04 00:21:56 +02:00
|
|
|
read_options, moptions.prefix_extractor.get(), /*arena=*/nullptr,
|
2019-06-20 23:28:22 +02:00
|
|
|
/*skip_filters=*/false, TableReaderCaller::kUncategorized));
|
2014-04-10 23:19:43 +02:00
|
|
|
|
|
|
|
// -- Find keys do not exist, but have common prefix.
|
|
|
|
std::vector<std::string> prefixes = {"001", "003", "005", "007", "009"};
|
2020-03-17 20:28:46 +01:00
|
|
|
std::vector<std::string> lower_bound = {
|
|
|
|
keys[0], keys[1], keys[2], keys[7], keys[9],
|
|
|
|
};
|
2014-04-10 23:19:43 +02:00
|
|
|
|
|
|
|
// find the lower bound of the prefix
|
|
|
|
for (size_t i = 0; i < prefixes.size(); ++i) {
|
2017-02-07 01:29:29 +01:00
|
|
|
index_iter->Seek(InternalKey(prefixes[i], 0, kTypeValue).Encode());
|
|
|
|
ASSERT_OK(index_iter->status());
|
|
|
|
ASSERT_TRUE(index_iter->Valid());
|
2014-04-10 23:19:43 +02:00
|
|
|
|
|
|
|
// seek the first element in the block
|
2017-02-07 01:29:29 +01:00
|
|
|
ASSERT_EQ(lower_bound[i], index_iter->key().ToString());
|
|
|
|
ASSERT_EQ("v", index_iter->value().ToString());
|
2014-04-10 23:19:43 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
// find the upper bound of prefixes
|
|
|
|
std::vector<std::string> upper_bound = {keys[1], keys[2], keys[7], keys[9], };
|
|
|
|
|
|
|
|
// find existing keys
|
|
|
|
for (const auto& item : kvmap) {
|
|
|
|
auto ukey = ExtractUserKey(item.first).ToString();
|
2017-02-07 01:29:29 +01:00
|
|
|
index_iter->Seek(ukey);
|
2014-04-10 23:19:43 +02:00
|
|
|
|
|
|
|
// ASSERT_OK(regular_iter->status());
|
2017-02-07 01:29:29 +01:00
|
|
|
ASSERT_OK(index_iter->status());
|
2014-04-10 23:19:43 +02:00
|
|
|
|
|
|
|
// ASSERT_TRUE(regular_iter->Valid());
|
2017-02-07 01:29:29 +01:00
|
|
|
ASSERT_TRUE(index_iter->Valid());
|
2014-04-10 23:19:43 +02:00
|
|
|
|
2017-02-07 01:29:29 +01:00
|
|
|
ASSERT_EQ(item.first, index_iter->key().ToString());
|
|
|
|
ASSERT_EQ(item.second, index_iter->value().ToString());
|
2014-04-10 23:19:43 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
for (size_t i = 0; i < prefixes.size(); ++i) {
|
|
|
|
// the key is greater than any existing keys.
|
|
|
|
auto key = prefixes[i] + "9";
|
2017-02-07 01:29:29 +01:00
|
|
|
index_iter->Seek(InternalKey(key, 0, kTypeValue).Encode());
|
2014-04-10 23:19:43 +02:00
|
|
|
|
2020-01-15 23:03:18 +01:00
|
|
|
ASSERT_TRUE(index_iter->status().ok() || index_iter->status().IsNotFound());
|
|
|
|
ASSERT_TRUE(!index_iter->status().IsNotFound() || !index_iter->Valid());
|
2014-04-10 23:19:43 +02:00
|
|
|
if (i == prefixes.size() - 1) {
|
|
|
|
// last key
|
2017-02-07 01:29:29 +01:00
|
|
|
ASSERT_TRUE(!index_iter->Valid());
|
2014-04-10 23:19:43 +02:00
|
|
|
} else {
|
2017-02-07 01:29:29 +01:00
|
|
|
ASSERT_TRUE(index_iter->Valid());
|
2014-04-10 23:19:43 +02:00
|
|
|
// seek the first element in the block
|
2017-02-07 01:29:29 +01:00
|
|
|
ASSERT_EQ(upper_bound[i], index_iter->key().ToString());
|
|
|
|
ASSERT_EQ("v", index_iter->value().ToString());
|
2014-04-10 23:19:43 +02:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
// find keys with prefix that don't match any of the existing prefixes.
|
|
|
|
std::vector<std::string> non_exist_prefixes = {"002", "004", "006", "008"};
|
|
|
|
for (const auto& prefix : non_exist_prefixes) {
|
2017-02-07 01:29:29 +01:00
|
|
|
index_iter->Seek(InternalKey(prefix, 0, kTypeValue).Encode());
|
2014-04-10 23:19:43 +02:00
|
|
|
// regular_iter->Seek(prefix);
|
|
|
|
|
2017-02-07 01:29:29 +01:00
|
|
|
ASSERT_OK(index_iter->status());
|
2014-06-13 04:03:22 +02:00
|
|
|
// Seek to non-existing prefixes should yield either invalid, or a
|
|
|
|
// key with prefix greater than the target.
|
2017-02-07 01:29:29 +01:00
|
|
|
if (index_iter->Valid()) {
|
|
|
|
Slice ukey = ExtractUserKey(index_iter->key());
|
2014-06-13 04:03:22 +02:00
|
|
|
Slice ukey_prefix = options.prefix_extractor->Transform(ukey);
|
|
|
|
ASSERT_TRUE(BytewiseComparator()->Compare(prefix, ukey_prefix) < 0);
|
|
|
|
}
|
2014-04-10 23:19:43 +02:00
|
|
|
}
|
2020-01-15 23:03:18 +01:00
|
|
|
for (const auto& prefix : non_exist_prefixes) {
|
|
|
|
index_iter->SeekForPrev(InternalKey(prefix, 0, kTypeValue).Encode());
|
|
|
|
// regular_iter->Seek(prefix);
|
|
|
|
|
|
|
|
ASSERT_OK(index_iter->status());
|
|
|
|
// Seek to non-existing prefixes should yield either invalid, or a
|
|
|
|
// key with prefix greater than the target.
|
|
|
|
if (index_iter->Valid()) {
|
|
|
|
Slice ukey = ExtractUserKey(index_iter->key());
|
|
|
|
Slice ukey_prefix = options.prefix_extractor->Transform(ukey);
|
|
|
|
ASSERT_TRUE(BytewiseComparator()->Compare(prefix, ukey_prefix) > 0);
|
|
|
|
}
|
|
|
|
}
|
2020-03-17 20:28:46 +01:00
|
|
|
|
|
|
|
{
|
|
|
|
// Test reseek case. It should impact partitioned index more.
|
|
|
|
ReadOptions ro;
|
|
|
|
ro.total_order_seek = true;
|
|
|
|
std::unique_ptr<InternalIterator> index_iter2(reader->NewIterator(
|
|
|
|
ro, moptions.prefix_extractor.get(), /*arena=*/nullptr,
|
|
|
|
/*skip_filters=*/false, TableReaderCaller::kUncategorized));
|
|
|
|
|
|
|
|
// Things to cover in partitioned index:
|
|
|
|
// 1. Both of Seek() and SeekToLast() has optimization to prevent
|
|
|
|
// rereek leaf index block if it remains to the same one, and
|
|
|
|
// they reuse the same variable.
|
|
|
|
// 2. When Next() or Prev() is called, the block moves, so the
|
|
|
|
// optimization should kick in only with the current one.
|
|
|
|
index_iter2->Seek(InternalKey("0055", 0, kTypeValue).Encode());
|
|
|
|
ASSERT_TRUE(index_iter2->Valid());
|
|
|
|
ASSERT_EQ("0055", index_iter2->key().ToString().substr(0, 4));
|
|
|
|
|
|
|
|
index_iter2->SeekToLast();
|
|
|
|
ASSERT_TRUE(index_iter2->Valid());
|
|
|
|
ASSERT_EQ("0095", index_iter2->key().ToString().substr(0, 4));
|
|
|
|
|
|
|
|
index_iter2->Seek(InternalKey("0055", 0, kTypeValue).Encode());
|
|
|
|
ASSERT_TRUE(index_iter2->Valid());
|
|
|
|
ASSERT_EQ("0055", index_iter2->key().ToString().substr(0, 4));
|
|
|
|
|
|
|
|
index_iter2->SeekToLast();
|
|
|
|
ASSERT_TRUE(index_iter2->Valid());
|
|
|
|
ASSERT_EQ("0095", index_iter2->key().ToString().substr(0, 4));
|
|
|
|
index_iter2->Prev();
|
|
|
|
ASSERT_TRUE(index_iter2->Valid());
|
|
|
|
index_iter2->Prev();
|
|
|
|
ASSERT_TRUE(index_iter2->Valid());
|
|
|
|
ASSERT_EQ("0075", index_iter2->key().ToString().substr(0, 4));
|
|
|
|
|
|
|
|
index_iter2->Seek(InternalKey("0095", 0, kTypeValue).Encode());
|
|
|
|
ASSERT_TRUE(index_iter2->Valid());
|
|
|
|
ASSERT_EQ("0095", index_iter2->key().ToString().substr(0, 4));
|
|
|
|
index_iter2->Prev();
|
|
|
|
ASSERT_TRUE(index_iter2->Valid());
|
|
|
|
index_iter2->Prev();
|
|
|
|
ASSERT_TRUE(index_iter2->Valid());
|
|
|
|
ASSERT_EQ("0075", index_iter2->key().ToString().substr(0, 4));
|
|
|
|
|
|
|
|
index_iter2->SeekToLast();
|
|
|
|
ASSERT_TRUE(index_iter2->Valid());
|
|
|
|
ASSERT_EQ("0095", index_iter2->key().ToString().substr(0, 4));
|
|
|
|
|
|
|
|
index_iter2->Seek(InternalKey("0095", 0, kTypeValue).Encode());
|
|
|
|
ASSERT_TRUE(index_iter2->Valid());
|
|
|
|
ASSERT_EQ("0095", index_iter2->key().ToString().substr(0, 4));
|
|
|
|
|
|
|
|
index_iter2->Prev();
|
|
|
|
ASSERT_TRUE(index_iter2->Valid());
|
|
|
|
index_iter2->Prev();
|
|
|
|
ASSERT_TRUE(index_iter2->Valid());
|
|
|
|
ASSERT_EQ("0075", index_iter2->key().ToString().substr(0, 4));
|
|
|
|
|
|
|
|
index_iter2->Seek(InternalKey("0075", 0, kTypeValue).Encode());
|
|
|
|
ASSERT_TRUE(index_iter2->Valid());
|
|
|
|
ASSERT_EQ("0075", index_iter2->key().ToString().substr(0, 4));
|
|
|
|
|
|
|
|
index_iter2->Next();
|
|
|
|
ASSERT_TRUE(index_iter2->Valid());
|
|
|
|
index_iter2->Next();
|
|
|
|
ASSERT_TRUE(index_iter2->Valid());
|
|
|
|
ASSERT_EQ("0095", index_iter2->key().ToString().substr(0, 4));
|
|
|
|
|
|
|
|
index_iter2->SeekToLast();
|
|
|
|
ASSERT_TRUE(index_iter2->Valid());
|
|
|
|
ASSERT_EQ("0095", index_iter2->key().ToString().substr(0, 4));
|
|
|
|
}
|
|
|
|
|
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 19:42:39 +02:00
|
|
|
c.ResetTableReader();
|
2014-04-10 23:19:43 +02:00
|
|
|
}
|
|
|
|
|
2018-06-05 04:59:44 +02:00
|
|
|
TEST_P(BlockBasedTableTest, BinaryIndexTest) {
|
|
|
|
BlockBasedTableOptions table_options = GetBlockBasedTableOptions();
|
2017-02-07 01:29:29 +01:00
|
|
|
table_options.index_type = BlockBasedTableOptions::kBinarySearch;
|
|
|
|
IndexTest(table_options);
|
|
|
|
}
|
|
|
|
|
2018-06-05 04:59:44 +02:00
|
|
|
TEST_P(BlockBasedTableTest, HashIndexTest) {
|
|
|
|
BlockBasedTableOptions table_options = GetBlockBasedTableOptions();
|
2017-02-07 01:29:29 +01:00
|
|
|
table_options.index_type = BlockBasedTableOptions::kHashSearch;
|
|
|
|
IndexTest(table_options);
|
|
|
|
}
|
|
|
|
|
2018-06-05 04:59:44 +02:00
|
|
|
TEST_P(BlockBasedTableTest, PartitionIndexTest) {
|
2017-02-07 01:29:29 +01:00
|
|
|
const int max_index_keys = 5;
|
2017-03-28 20:56:56 +02:00
|
|
|
const int est_max_index_key_value_size = 32;
|
|
|
|
const int est_max_index_size = max_index_keys * est_max_index_key_value_size;
|
|
|
|
for (int i = 1; i <= est_max_index_size + 1; i++) {
|
2018-06-05 04:59:44 +02:00
|
|
|
BlockBasedTableOptions table_options = GetBlockBasedTableOptions();
|
2017-02-07 01:29:29 +01:00
|
|
|
table_options.index_type = BlockBasedTableOptions::kTwoLevelIndexSearch;
|
2017-03-28 20:56:56 +02:00
|
|
|
table_options.metadata_block_size = i;
|
2017-02-07 01:29:29 +01:00
|
|
|
IndexTest(table_options);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2019-05-10 21:36:40 +02:00
|
|
|
TEST_P(BlockBasedTableTest, IndexSeekOptimizationIncomplete) {
|
|
|
|
std::unique_ptr<InternalKeyComparator> comparator(
|
|
|
|
new InternalKeyComparator(BytewiseComparator()));
|
|
|
|
BlockBasedTableOptions table_options = GetBlockBasedTableOptions();
|
|
|
|
Options options;
|
|
|
|
options.table_factory.reset(NewBlockBasedTableFactory(table_options));
|
|
|
|
const ImmutableCFOptions ioptions(options);
|
|
|
|
const MutableCFOptions moptions(options);
|
|
|
|
|
|
|
|
TableConstructor c(BytewiseComparator());
|
|
|
|
AddInternalKey(&c, "pika");
|
|
|
|
|
|
|
|
std::vector<std::string> keys;
|
|
|
|
stl_wrappers::KVMap kvmap;
|
|
|
|
c.Finish(options, ioptions, moptions, table_options, *comparator, &keys,
|
|
|
|
&kvmap);
|
|
|
|
ASSERT_EQ(1, keys.size());
|
|
|
|
|
|
|
|
auto reader = c.GetTableReader();
|
|
|
|
ReadOptions ropt;
|
|
|
|
ropt.read_tier = ReadTier::kBlockCacheTier;
|
2019-06-20 23:28:22 +02:00
|
|
|
std::unique_ptr<InternalIterator> iter(reader->NewIterator(
|
|
|
|
ropt, /*prefix_extractor=*/nullptr, /*arena=*/nullptr,
|
|
|
|
/*skip_filters=*/false, TableReaderCaller::kUncategorized));
|
2019-05-10 21:36:40 +02:00
|
|
|
|
|
|
|
auto ikey = [](Slice user_key) {
|
|
|
|
return InternalKey(user_key, 0, kTypeValue).Encode().ToString();
|
|
|
|
};
|
|
|
|
|
|
|
|
iter->Seek(ikey("pika"));
|
|
|
|
ASSERT_FALSE(iter->Valid());
|
|
|
|
ASSERT_TRUE(iter->status().IsIncomplete());
|
|
|
|
|
|
|
|
// This used to crash at some point.
|
|
|
|
iter->Seek(ikey("pika"));
|
|
|
|
ASSERT_FALSE(iter->Valid());
|
|
|
|
ASSERT_TRUE(iter->status().IsIncomplete());
|
|
|
|
}
|
|
|
|
|
Add an option to put first key of each sst block in the index (#5289)
Summary:
The first key is used to defer reading the data block until this file gets to the top of merging iterator's heap. For short range scans, most files never make it to the top of the heap, so this change can reduce read amplification by a lot sometimes.
Consider the following workload. There are a few data streams (we'll be calling them "logs"), each stream consisting of a sequence of blobs (we'll be calling them "records"). Each record is identified by log ID and a sequence number within the log. RocksDB key is concatenation of log ID and sequence number (big endian). Reads are mostly relatively short range scans, each within a single log. Writes are mostly sequential for each log, but writes to different logs are randomly interleaved. Compactions are disabled; instead, when we accumulate a few tens of sst files, we create a new column family and start writing to it.
So, a typical sst file consists of a few ranges of blocks, each range corresponding to one log ID (we use FlushBlockPolicy to cut blocks at log boundaries). A typical read would go like this. First, iterator Seek() reads one block from each sst file. Then a series of Next()s move through one sst file (since writes to each log are mostly sequential) until the subiterator reaches the end of this log in this sst file; then Next() switches to the next sst file and reads sequentially from that, and so on. Often a range scan will only return records from a small number of blocks in small number of sst files; in this case, the cost of initial Seek() reading one block from each file may be bigger than the cost of reading the actually useful blocks.
Neither iterate_upper_bound nor bloom filters can prevent reading one block from each file in Seek(). But this PR can: if the index contains first key from each block, we don't have to read the block until this block actually makes it to the top of merging iterator's heap, so for short range scans we won't read any blocks from most of the sst files.
This PR does the deferred block loading inside value() call. This is not ideal: there's no good way to report an IO error from inside value(). As discussed with siying offline, it would probably be better to change InternalIterator's interface to explicitly fetch deferred value and get status. I'll do it in a separate PR.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/5289
Differential Revision: D15256423
Pulled By: al13n321
fbshipit-source-id: 750e4c39ce88e8d41662f701cf6275d9388ba46a
2019-06-25 05:50:35 +02:00
|
|
|
TEST_P(BlockBasedTableTest, BinaryIndexWithFirstKey1) {
|
|
|
|
BlockBasedTableOptions table_options = GetBlockBasedTableOptions();
|
|
|
|
table_options.index_type = BlockBasedTableOptions::kBinarySearchWithFirstKey;
|
|
|
|
IndexTest(table_options);
|
|
|
|
}
|
|
|
|
|
|
|
|
class CustomFlushBlockPolicy : public FlushBlockPolicyFactory,
|
|
|
|
public FlushBlockPolicy {
|
|
|
|
public:
|
|
|
|
explicit CustomFlushBlockPolicy(std::vector<int> keys_per_block)
|
|
|
|
: keys_per_block_(keys_per_block) {}
|
|
|
|
|
2020-10-02 00:21:17 +02:00
|
|
|
const char* Name() const override { return "CustomFlushBlockPolicy"; }
|
|
|
|
|
Add an option to put first key of each sst block in the index (#5289)
Summary:
The first key is used to defer reading the data block until this file gets to the top of merging iterator's heap. For short range scans, most files never make it to the top of the heap, so this change can reduce read amplification by a lot sometimes.
Consider the following workload. There are a few data streams (we'll be calling them "logs"), each stream consisting of a sequence of blobs (we'll be calling them "records"). Each record is identified by log ID and a sequence number within the log. RocksDB key is concatenation of log ID and sequence number (big endian). Reads are mostly relatively short range scans, each within a single log. Writes are mostly sequential for each log, but writes to different logs are randomly interleaved. Compactions are disabled; instead, when we accumulate a few tens of sst files, we create a new column family and start writing to it.
So, a typical sst file consists of a few ranges of blocks, each range corresponding to one log ID (we use FlushBlockPolicy to cut blocks at log boundaries). A typical read would go like this. First, iterator Seek() reads one block from each sst file. Then a series of Next()s move through one sst file (since writes to each log are mostly sequential) until the subiterator reaches the end of this log in this sst file; then Next() switches to the next sst file and reads sequentially from that, and so on. Often a range scan will only return records from a small number of blocks in small number of sst files; in this case, the cost of initial Seek() reading one block from each file may be bigger than the cost of reading the actually useful blocks.
Neither iterate_upper_bound nor bloom filters can prevent reading one block from each file in Seek(). But this PR can: if the index contains first key from each block, we don't have to read the block until this block actually makes it to the top of merging iterator's heap, so for short range scans we won't read any blocks from most of the sst files.
This PR does the deferred block loading inside value() call. This is not ideal: there's no good way to report an IO error from inside value(). As discussed with siying offline, it would probably be better to change InternalIterator's interface to explicitly fetch deferred value and get status. I'll do it in a separate PR.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/5289
Differential Revision: D15256423
Pulled By: al13n321
fbshipit-source-id: 750e4c39ce88e8d41662f701cf6275d9388ba46a
2019-06-25 05:50:35 +02:00
|
|
|
FlushBlockPolicy* NewFlushBlockPolicy(const BlockBasedTableOptions&,
|
|
|
|
const BlockBuilder&) const override {
|
|
|
|
return new CustomFlushBlockPolicy(keys_per_block_);
|
|
|
|
}
|
|
|
|
|
|
|
|
bool Update(const Slice&, const Slice&) override {
|
|
|
|
if (keys_in_current_block_ >= keys_per_block_.at(current_block_idx_)) {
|
|
|
|
++current_block_idx_;
|
|
|
|
keys_in_current_block_ = 1;
|
|
|
|
return true;
|
|
|
|
}
|
|
|
|
|
|
|
|
++keys_in_current_block_;
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
|
|
|
|
std::vector<int> keys_per_block_;
|
|
|
|
|
|
|
|
int current_block_idx_ = 0;
|
|
|
|
int keys_in_current_block_ = 0;
|
|
|
|
};
|
|
|
|
|
|
|
|
TEST_P(BlockBasedTableTest, BinaryIndexWithFirstKey2) {
|
|
|
|
for (int use_first_key = 0; use_first_key < 2; ++use_first_key) {
|
|
|
|
SCOPED_TRACE("use_first_key = " + std::to_string(use_first_key));
|
|
|
|
BlockBasedTableOptions table_options = GetBlockBasedTableOptions();
|
|
|
|
table_options.index_type =
|
|
|
|
use_first_key ? BlockBasedTableOptions::kBinarySearchWithFirstKey
|
|
|
|
: BlockBasedTableOptions::kBinarySearch;
|
|
|
|
table_options.block_cache = NewLRUCache(10000); // fits all blocks
|
|
|
|
table_options.index_shortening =
|
|
|
|
BlockBasedTableOptions::IndexShorteningMode::kNoShortening;
|
|
|
|
table_options.flush_block_policy_factory =
|
|
|
|
std::make_shared<CustomFlushBlockPolicy>(std::vector<int>{2, 1, 3, 2});
|
|
|
|
Options options;
|
|
|
|
options.table_factory.reset(NewBlockBasedTableFactory(table_options));
|
|
|
|
options.statistics = CreateDBStatistics();
|
|
|
|
Statistics* stats = options.statistics.get();
|
|
|
|
std::unique_ptr<InternalKeyComparator> comparator(
|
|
|
|
new InternalKeyComparator(BytewiseComparator()));
|
|
|
|
const ImmutableCFOptions ioptions(options);
|
|
|
|
const MutableCFOptions moptions(options);
|
|
|
|
|
|
|
|
TableConstructor c(BytewiseComparator());
|
|
|
|
|
|
|
|
// Block 0.
|
|
|
|
AddInternalKey(&c, "aaaa", "v0");
|
|
|
|
AddInternalKey(&c, "aaac", "v1");
|
|
|
|
|
|
|
|
// Block 1.
|
|
|
|
AddInternalKey(&c, "aaca", "v2");
|
|
|
|
|
|
|
|
// Block 2.
|
|
|
|
AddInternalKey(&c, "caaa", "v3");
|
|
|
|
AddInternalKey(&c, "caac", "v4");
|
|
|
|
AddInternalKey(&c, "caae", "v5");
|
|
|
|
|
|
|
|
// Block 3.
|
|
|
|
AddInternalKey(&c, "ccaa", "v6");
|
|
|
|
AddInternalKey(&c, "ccac", "v7");
|
|
|
|
|
|
|
|
// Write the file.
|
|
|
|
std::vector<std::string> keys;
|
|
|
|
stl_wrappers::KVMap kvmap;
|
|
|
|
c.Finish(options, ioptions, moptions, table_options, *comparator, &keys,
|
|
|
|
&kvmap);
|
|
|
|
ASSERT_EQ(8, keys.size());
|
|
|
|
|
|
|
|
auto reader = c.GetTableReader();
|
|
|
|
auto props = reader->GetTableProperties();
|
|
|
|
ASSERT_EQ(4u, props->num_data_blocks);
|
2020-08-04 00:21:56 +02:00
|
|
|
ReadOptions read_options;
|
Add an option to put first key of each sst block in the index (#5289)
Summary:
The first key is used to defer reading the data block until this file gets to the top of merging iterator's heap. For short range scans, most files never make it to the top of the heap, so this change can reduce read amplification by a lot sometimes.
Consider the following workload. There are a few data streams (we'll be calling them "logs"), each stream consisting of a sequence of blobs (we'll be calling them "records"). Each record is identified by log ID and a sequence number within the log. RocksDB key is concatenation of log ID and sequence number (big endian). Reads are mostly relatively short range scans, each within a single log. Writes are mostly sequential for each log, but writes to different logs are randomly interleaved. Compactions are disabled; instead, when we accumulate a few tens of sst files, we create a new column family and start writing to it.
So, a typical sst file consists of a few ranges of blocks, each range corresponding to one log ID (we use FlushBlockPolicy to cut blocks at log boundaries). A typical read would go like this. First, iterator Seek() reads one block from each sst file. Then a series of Next()s move through one sst file (since writes to each log are mostly sequential) until the subiterator reaches the end of this log in this sst file; then Next() switches to the next sst file and reads sequentially from that, and so on. Often a range scan will only return records from a small number of blocks in small number of sst files; in this case, the cost of initial Seek() reading one block from each file may be bigger than the cost of reading the actually useful blocks.
Neither iterate_upper_bound nor bloom filters can prevent reading one block from each file in Seek(). But this PR can: if the index contains first key from each block, we don't have to read the block until this block actually makes it to the top of merging iterator's heap, so for short range scans we won't read any blocks from most of the sst files.
This PR does the deferred block loading inside value() call. This is not ideal: there's no good way to report an IO error from inside value(). As discussed with siying offline, it would probably be better to change InternalIterator's interface to explicitly fetch deferred value and get status. I'll do it in a separate PR.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/5289
Differential Revision: D15256423
Pulled By: al13n321
fbshipit-source-id: 750e4c39ce88e8d41662f701cf6275d9388ba46a
2019-06-25 05:50:35 +02:00
|
|
|
std::unique_ptr<InternalIterator> iter(reader->NewIterator(
|
2020-08-04 00:21:56 +02:00
|
|
|
read_options, /*prefix_extractor=*/nullptr, /*arena=*/nullptr,
|
Properly report IO errors when IndexType::kBinarySearchWithFirstKey is used (#6621)
Summary:
Context: Index type `kBinarySearchWithFirstKey` added the ability for sst file iterator to sometimes report a key from index without reading the corresponding data block. This is useful when sst blocks are cut at some meaningful boundaries (e.g. one block per key prefix), and many seeks land between blocks (e.g. for each prefix, the ranges of keys in different sst files are nearly disjoint, so a typical seek needs to read a data block from only one file even if all files have the prefix). But this added a new error condition, which rocksdb code was really not equipped to deal with: `InternalIterator::value()` may fail with an IO error or Status::Incomplete, but it's just a method returning a Slice, with no way to report error instead. Before this PR, this type of error wasn't handled at all (an empty slice was returned), and kBinarySearchWithFirstKey implementation was considered a prototype.
Now that we (LogDevice) have experimented with kBinarySearchWithFirstKey for a while and confirmed that it's really useful, this PR is adding the missing error handling.
It's a pretty inconvenient situation implementation-wise. The error needs to be reported from InternalIterator when trying to access value. But there are ~700 call sites of `InternalIterator::value()`, most of which either can't hit the error condition (because the iterator is reading from memtable or from index or something) or wouldn't benefit from the deferred loading of the value (e.g. compaction iterator that reads all values anyway). Adding error handling to all these call sites would needlessly bloat the code. So instead I made the deferred value loading optional: only the call sites that may use deferred loading have to call the new method `PrepareValue()` before calling `value()`. The feature is enabled with a new bool argument `allow_unprepared_value` to a bunch of methods that create iterators (it wouldn't make sense to put it in ReadOptions because it's completely internal to iterators, with virtually no user-visible effect). Lmk if you have better ideas.
Note that the deferred value loading only happens for *internal* iterators. The user-visible iterator (DBIter) always prepares the value before returning from Seek/Next/etc. We could go further and add an API to defer that value loading too, but that's most likely not useful for LogDevice, so it doesn't seem worth the complexity for now.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6621
Test Plan: make -j5 check . Will also deploy to some logdevice test clusters and look at stats.
Reviewed By: siying
Differential Revision: D20786930
Pulled By: al13n321
fbshipit-source-id: 6da77d918bad3780522e918f17f4d5513d3e99ee
2020-04-16 02:37:23 +02:00
|
|
|
/*skip_filters=*/false, TableReaderCaller::kUncategorized,
|
|
|
|
/*compaction_readahead_size=*/0, /*allow_unprepared_value=*/true));
|
Add an option to put first key of each sst block in the index (#5289)
Summary:
The first key is used to defer reading the data block until this file gets to the top of merging iterator's heap. For short range scans, most files never make it to the top of the heap, so this change can reduce read amplification by a lot sometimes.
Consider the following workload. There are a few data streams (we'll be calling them "logs"), each stream consisting of a sequence of blobs (we'll be calling them "records"). Each record is identified by log ID and a sequence number within the log. RocksDB key is concatenation of log ID and sequence number (big endian). Reads are mostly relatively short range scans, each within a single log. Writes are mostly sequential for each log, but writes to different logs are randomly interleaved. Compactions are disabled; instead, when we accumulate a few tens of sst files, we create a new column family and start writing to it.
So, a typical sst file consists of a few ranges of blocks, each range corresponding to one log ID (we use FlushBlockPolicy to cut blocks at log boundaries). A typical read would go like this. First, iterator Seek() reads one block from each sst file. Then a series of Next()s move through one sst file (since writes to each log are mostly sequential) until the subiterator reaches the end of this log in this sst file; then Next() switches to the next sst file and reads sequentially from that, and so on. Often a range scan will only return records from a small number of blocks in small number of sst files; in this case, the cost of initial Seek() reading one block from each file may be bigger than the cost of reading the actually useful blocks.
Neither iterate_upper_bound nor bloom filters can prevent reading one block from each file in Seek(). But this PR can: if the index contains first key from each block, we don't have to read the block until this block actually makes it to the top of merging iterator's heap, so for short range scans we won't read any blocks from most of the sst files.
This PR does the deferred block loading inside value() call. This is not ideal: there's no good way to report an IO error from inside value(). As discussed with siying offline, it would probably be better to change InternalIterator's interface to explicitly fetch deferred value and get status. I'll do it in a separate PR.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/5289
Differential Revision: D15256423
Pulled By: al13n321
fbshipit-source-id: 750e4c39ce88e8d41662f701cf6275d9388ba46a
2019-06-25 05:50:35 +02:00
|
|
|
|
|
|
|
// Shouldn't have read data blocks before iterator is seeked.
|
|
|
|
EXPECT_EQ(0, stats->getTickerCount(BLOCK_CACHE_DATA_MISS));
|
|
|
|
EXPECT_EQ(0, stats->getTickerCount(BLOCK_CACHE_DATA_HIT));
|
|
|
|
|
|
|
|
auto ikey = [](Slice user_key) {
|
|
|
|
return InternalKey(user_key, 0, kTypeValue).Encode().ToString();
|
|
|
|
};
|
|
|
|
|
|
|
|
// Seek to a key between blocks. If index contains first key, we shouldn't
|
|
|
|
// read any data blocks until value is requested.
|
|
|
|
iter->Seek(ikey("aaba"));
|
|
|
|
ASSERT_TRUE(iter->Valid());
|
|
|
|
EXPECT_EQ(keys[2], iter->key().ToString());
|
|
|
|
EXPECT_EQ(use_first_key ? 0 : 1,
|
|
|
|
stats->getTickerCount(BLOCK_CACHE_DATA_MISS));
|
Properly report IO errors when IndexType::kBinarySearchWithFirstKey is used (#6621)
Summary:
Context: Index type `kBinarySearchWithFirstKey` added the ability for sst file iterator to sometimes report a key from index without reading the corresponding data block. This is useful when sst blocks are cut at some meaningful boundaries (e.g. one block per key prefix), and many seeks land between blocks (e.g. for each prefix, the ranges of keys in different sst files are nearly disjoint, so a typical seek needs to read a data block from only one file even if all files have the prefix). But this added a new error condition, which rocksdb code was really not equipped to deal with: `InternalIterator::value()` may fail with an IO error or Status::Incomplete, but it's just a method returning a Slice, with no way to report error instead. Before this PR, this type of error wasn't handled at all (an empty slice was returned), and kBinarySearchWithFirstKey implementation was considered a prototype.
Now that we (LogDevice) have experimented with kBinarySearchWithFirstKey for a while and confirmed that it's really useful, this PR is adding the missing error handling.
It's a pretty inconvenient situation implementation-wise. The error needs to be reported from InternalIterator when trying to access value. But there are ~700 call sites of `InternalIterator::value()`, most of which either can't hit the error condition (because the iterator is reading from memtable or from index or something) or wouldn't benefit from the deferred loading of the value (e.g. compaction iterator that reads all values anyway). Adding error handling to all these call sites would needlessly bloat the code. So instead I made the deferred value loading optional: only the call sites that may use deferred loading have to call the new method `PrepareValue()` before calling `value()`. The feature is enabled with a new bool argument `allow_unprepared_value` to a bunch of methods that create iterators (it wouldn't make sense to put it in ReadOptions because it's completely internal to iterators, with virtually no user-visible effect). Lmk if you have better ideas.
Note that the deferred value loading only happens for *internal* iterators. The user-visible iterator (DBIter) always prepares the value before returning from Seek/Next/etc. We could go further and add an API to defer that value loading too, but that's most likely not useful for LogDevice, so it doesn't seem worth the complexity for now.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6621
Test Plan: make -j5 check . Will also deploy to some logdevice test clusters and look at stats.
Reviewed By: siying
Differential Revision: D20786930
Pulled By: al13n321
fbshipit-source-id: 6da77d918bad3780522e918f17f4d5513d3e99ee
2020-04-16 02:37:23 +02:00
|
|
|
ASSERT_TRUE(iter->PrepareValue());
|
Add an option to put first key of each sst block in the index (#5289)
Summary:
The first key is used to defer reading the data block until this file gets to the top of merging iterator's heap. For short range scans, most files never make it to the top of the heap, so this change can reduce read amplification by a lot sometimes.
Consider the following workload. There are a few data streams (we'll be calling them "logs"), each stream consisting of a sequence of blobs (we'll be calling them "records"). Each record is identified by log ID and a sequence number within the log. RocksDB key is concatenation of log ID and sequence number (big endian). Reads are mostly relatively short range scans, each within a single log. Writes are mostly sequential for each log, but writes to different logs are randomly interleaved. Compactions are disabled; instead, when we accumulate a few tens of sst files, we create a new column family and start writing to it.
So, a typical sst file consists of a few ranges of blocks, each range corresponding to one log ID (we use FlushBlockPolicy to cut blocks at log boundaries). A typical read would go like this. First, iterator Seek() reads one block from each sst file. Then a series of Next()s move through one sst file (since writes to each log are mostly sequential) until the subiterator reaches the end of this log in this sst file; then Next() switches to the next sst file and reads sequentially from that, and so on. Often a range scan will only return records from a small number of blocks in small number of sst files; in this case, the cost of initial Seek() reading one block from each file may be bigger than the cost of reading the actually useful blocks.
Neither iterate_upper_bound nor bloom filters can prevent reading one block from each file in Seek(). But this PR can: if the index contains first key from each block, we don't have to read the block until this block actually makes it to the top of merging iterator's heap, so for short range scans we won't read any blocks from most of the sst files.
This PR does the deferred block loading inside value() call. This is not ideal: there's no good way to report an IO error from inside value(). As discussed with siying offline, it would probably be better to change InternalIterator's interface to explicitly fetch deferred value and get status. I'll do it in a separate PR.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/5289
Differential Revision: D15256423
Pulled By: al13n321
fbshipit-source-id: 750e4c39ce88e8d41662f701cf6275d9388ba46a
2019-06-25 05:50:35 +02:00
|
|
|
EXPECT_EQ("v2", iter->value().ToString());
|
|
|
|
EXPECT_EQ(1, stats->getTickerCount(BLOCK_CACHE_DATA_MISS));
|
|
|
|
EXPECT_EQ(0, stats->getTickerCount(BLOCK_CACHE_DATA_HIT));
|
|
|
|
|
|
|
|
// Seek to the middle of a block. The block should be read right away.
|
|
|
|
iter->Seek(ikey("caab"));
|
|
|
|
ASSERT_TRUE(iter->Valid());
|
|
|
|
EXPECT_EQ(keys[4], iter->key().ToString());
|
|
|
|
EXPECT_EQ(2, stats->getTickerCount(BLOCK_CACHE_DATA_MISS));
|
|
|
|
EXPECT_EQ(0, stats->getTickerCount(BLOCK_CACHE_DATA_HIT));
|
Properly report IO errors when IndexType::kBinarySearchWithFirstKey is used (#6621)
Summary:
Context: Index type `kBinarySearchWithFirstKey` added the ability for sst file iterator to sometimes report a key from index without reading the corresponding data block. This is useful when sst blocks are cut at some meaningful boundaries (e.g. one block per key prefix), and many seeks land between blocks (e.g. for each prefix, the ranges of keys in different sst files are nearly disjoint, so a typical seek needs to read a data block from only one file even if all files have the prefix). But this added a new error condition, which rocksdb code was really not equipped to deal with: `InternalIterator::value()` may fail with an IO error or Status::Incomplete, but it's just a method returning a Slice, with no way to report error instead. Before this PR, this type of error wasn't handled at all (an empty slice was returned), and kBinarySearchWithFirstKey implementation was considered a prototype.
Now that we (LogDevice) have experimented with kBinarySearchWithFirstKey for a while and confirmed that it's really useful, this PR is adding the missing error handling.
It's a pretty inconvenient situation implementation-wise. The error needs to be reported from InternalIterator when trying to access value. But there are ~700 call sites of `InternalIterator::value()`, most of which either can't hit the error condition (because the iterator is reading from memtable or from index or something) or wouldn't benefit from the deferred loading of the value (e.g. compaction iterator that reads all values anyway). Adding error handling to all these call sites would needlessly bloat the code. So instead I made the deferred value loading optional: only the call sites that may use deferred loading have to call the new method `PrepareValue()` before calling `value()`. The feature is enabled with a new bool argument `allow_unprepared_value` to a bunch of methods that create iterators (it wouldn't make sense to put it in ReadOptions because it's completely internal to iterators, with virtually no user-visible effect). Lmk if you have better ideas.
Note that the deferred value loading only happens for *internal* iterators. The user-visible iterator (DBIter) always prepares the value before returning from Seek/Next/etc. We could go further and add an API to defer that value loading too, but that's most likely not useful for LogDevice, so it doesn't seem worth the complexity for now.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6621
Test Plan: make -j5 check . Will also deploy to some logdevice test clusters and look at stats.
Reviewed By: siying
Differential Revision: D20786930
Pulled By: al13n321
fbshipit-source-id: 6da77d918bad3780522e918f17f4d5513d3e99ee
2020-04-16 02:37:23 +02:00
|
|
|
ASSERT_TRUE(iter->PrepareValue());
|
Add an option to put first key of each sst block in the index (#5289)
Summary:
The first key is used to defer reading the data block until this file gets to the top of merging iterator's heap. For short range scans, most files never make it to the top of the heap, so this change can reduce read amplification by a lot sometimes.
Consider the following workload. There are a few data streams (we'll be calling them "logs"), each stream consisting of a sequence of blobs (we'll be calling them "records"). Each record is identified by log ID and a sequence number within the log. RocksDB key is concatenation of log ID and sequence number (big endian). Reads are mostly relatively short range scans, each within a single log. Writes are mostly sequential for each log, but writes to different logs are randomly interleaved. Compactions are disabled; instead, when we accumulate a few tens of sst files, we create a new column family and start writing to it.
So, a typical sst file consists of a few ranges of blocks, each range corresponding to one log ID (we use FlushBlockPolicy to cut blocks at log boundaries). A typical read would go like this. First, iterator Seek() reads one block from each sst file. Then a series of Next()s move through one sst file (since writes to each log are mostly sequential) until the subiterator reaches the end of this log in this sst file; then Next() switches to the next sst file and reads sequentially from that, and so on. Often a range scan will only return records from a small number of blocks in small number of sst files; in this case, the cost of initial Seek() reading one block from each file may be bigger than the cost of reading the actually useful blocks.
Neither iterate_upper_bound nor bloom filters can prevent reading one block from each file in Seek(). But this PR can: if the index contains first key from each block, we don't have to read the block until this block actually makes it to the top of merging iterator's heap, so for short range scans we won't read any blocks from most of the sst files.
This PR does the deferred block loading inside value() call. This is not ideal: there's no good way to report an IO error from inside value(). As discussed with siying offline, it would probably be better to change InternalIterator's interface to explicitly fetch deferred value and get status. I'll do it in a separate PR.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/5289
Differential Revision: D15256423
Pulled By: al13n321
fbshipit-source-id: 750e4c39ce88e8d41662f701cf6275d9388ba46a
2019-06-25 05:50:35 +02:00
|
|
|
EXPECT_EQ("v4", iter->value().ToString());
|
|
|
|
EXPECT_EQ(0, stats->getTickerCount(BLOCK_CACHE_DATA_HIT));
|
|
|
|
|
|
|
|
// Seek to just before the same block and don't access value.
|
|
|
|
// The iterator should keep pinning the block contents.
|
|
|
|
iter->Seek(ikey("baaa"));
|
|
|
|
ASSERT_TRUE(iter->Valid());
|
|
|
|
EXPECT_EQ(keys[3], iter->key().ToString());
|
|
|
|
EXPECT_EQ(0, stats->getTickerCount(BLOCK_CACHE_DATA_HIT));
|
|
|
|
|
|
|
|
// Seek to the same block again to check that the block is still pinned.
|
|
|
|
iter->Seek(ikey("caae"));
|
|
|
|
ASSERT_TRUE(iter->Valid());
|
|
|
|
EXPECT_EQ(keys[5], iter->key().ToString());
|
|
|
|
EXPECT_EQ(0, stats->getTickerCount(BLOCK_CACHE_DATA_HIT));
|
Properly report IO errors when IndexType::kBinarySearchWithFirstKey is used (#6621)
Summary:
Context: Index type `kBinarySearchWithFirstKey` added the ability for sst file iterator to sometimes report a key from index without reading the corresponding data block. This is useful when sst blocks are cut at some meaningful boundaries (e.g. one block per key prefix), and many seeks land between blocks (e.g. for each prefix, the ranges of keys in different sst files are nearly disjoint, so a typical seek needs to read a data block from only one file even if all files have the prefix). But this added a new error condition, which rocksdb code was really not equipped to deal with: `InternalIterator::value()` may fail with an IO error or Status::Incomplete, but it's just a method returning a Slice, with no way to report error instead. Before this PR, this type of error wasn't handled at all (an empty slice was returned), and kBinarySearchWithFirstKey implementation was considered a prototype.
Now that we (LogDevice) have experimented with kBinarySearchWithFirstKey for a while and confirmed that it's really useful, this PR is adding the missing error handling.
It's a pretty inconvenient situation implementation-wise. The error needs to be reported from InternalIterator when trying to access value. But there are ~700 call sites of `InternalIterator::value()`, most of which either can't hit the error condition (because the iterator is reading from memtable or from index or something) or wouldn't benefit from the deferred loading of the value (e.g. compaction iterator that reads all values anyway). Adding error handling to all these call sites would needlessly bloat the code. So instead I made the deferred value loading optional: only the call sites that may use deferred loading have to call the new method `PrepareValue()` before calling `value()`. The feature is enabled with a new bool argument `allow_unprepared_value` to a bunch of methods that create iterators (it wouldn't make sense to put it in ReadOptions because it's completely internal to iterators, with virtually no user-visible effect). Lmk if you have better ideas.
Note that the deferred value loading only happens for *internal* iterators. The user-visible iterator (DBIter) always prepares the value before returning from Seek/Next/etc. We could go further and add an API to defer that value loading too, but that's most likely not useful for LogDevice, so it doesn't seem worth the complexity for now.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6621
Test Plan: make -j5 check . Will also deploy to some logdevice test clusters and look at stats.
Reviewed By: siying
Differential Revision: D20786930
Pulled By: al13n321
fbshipit-source-id: 6da77d918bad3780522e918f17f4d5513d3e99ee
2020-04-16 02:37:23 +02:00
|
|
|
ASSERT_TRUE(iter->PrepareValue());
|
Add an option to put first key of each sst block in the index (#5289)
Summary:
The first key is used to defer reading the data block until this file gets to the top of merging iterator's heap. For short range scans, most files never make it to the top of the heap, so this change can reduce read amplification by a lot sometimes.
Consider the following workload. There are a few data streams (we'll be calling them "logs"), each stream consisting of a sequence of blobs (we'll be calling them "records"). Each record is identified by log ID and a sequence number within the log. RocksDB key is concatenation of log ID and sequence number (big endian). Reads are mostly relatively short range scans, each within a single log. Writes are mostly sequential for each log, but writes to different logs are randomly interleaved. Compactions are disabled; instead, when we accumulate a few tens of sst files, we create a new column family and start writing to it.
So, a typical sst file consists of a few ranges of blocks, each range corresponding to one log ID (we use FlushBlockPolicy to cut blocks at log boundaries). A typical read would go like this. First, iterator Seek() reads one block from each sst file. Then a series of Next()s move through one sst file (since writes to each log are mostly sequential) until the subiterator reaches the end of this log in this sst file; then Next() switches to the next sst file and reads sequentially from that, and so on. Often a range scan will only return records from a small number of blocks in small number of sst files; in this case, the cost of initial Seek() reading one block from each file may be bigger than the cost of reading the actually useful blocks.
Neither iterate_upper_bound nor bloom filters can prevent reading one block from each file in Seek(). But this PR can: if the index contains first key from each block, we don't have to read the block until this block actually makes it to the top of merging iterator's heap, so for short range scans we won't read any blocks from most of the sst files.
This PR does the deferred block loading inside value() call. This is not ideal: there's no good way to report an IO error from inside value(). As discussed with siying offline, it would probably be better to change InternalIterator's interface to explicitly fetch deferred value and get status. I'll do it in a separate PR.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/5289
Differential Revision: D15256423
Pulled By: al13n321
fbshipit-source-id: 750e4c39ce88e8d41662f701cf6275d9388ba46a
2019-06-25 05:50:35 +02:00
|
|
|
EXPECT_EQ("v5", iter->value().ToString());
|
|
|
|
EXPECT_EQ(2, stats->getTickerCount(BLOCK_CACHE_DATA_MISS));
|
|
|
|
EXPECT_EQ(0, stats->getTickerCount(BLOCK_CACHE_DATA_HIT));
|
|
|
|
|
|
|
|
// Step forward and fall through to the next block. Don't access value.
|
|
|
|
iter->Next();
|
|
|
|
ASSERT_TRUE(iter->Valid());
|
|
|
|
EXPECT_EQ(keys[6], iter->key().ToString());
|
|
|
|
EXPECT_EQ(use_first_key ? 2 : 3,
|
|
|
|
stats->getTickerCount(BLOCK_CACHE_DATA_MISS));
|
|
|
|
EXPECT_EQ(0, stats->getTickerCount(BLOCK_CACHE_DATA_HIT));
|
|
|
|
|
|
|
|
// Step forward again. Block should be read.
|
|
|
|
iter->Next();
|
|
|
|
ASSERT_TRUE(iter->Valid());
|
|
|
|
EXPECT_EQ(keys[7], iter->key().ToString());
|
|
|
|
EXPECT_EQ(3, stats->getTickerCount(BLOCK_CACHE_DATA_MISS));
|
Properly report IO errors when IndexType::kBinarySearchWithFirstKey is used (#6621)
Summary:
Context: Index type `kBinarySearchWithFirstKey` added the ability for sst file iterator to sometimes report a key from index without reading the corresponding data block. This is useful when sst blocks are cut at some meaningful boundaries (e.g. one block per key prefix), and many seeks land between blocks (e.g. for each prefix, the ranges of keys in different sst files are nearly disjoint, so a typical seek needs to read a data block from only one file even if all files have the prefix). But this added a new error condition, which rocksdb code was really not equipped to deal with: `InternalIterator::value()` may fail with an IO error or Status::Incomplete, but it's just a method returning a Slice, with no way to report error instead. Before this PR, this type of error wasn't handled at all (an empty slice was returned), and kBinarySearchWithFirstKey implementation was considered a prototype.
Now that we (LogDevice) have experimented with kBinarySearchWithFirstKey for a while and confirmed that it's really useful, this PR is adding the missing error handling.
It's a pretty inconvenient situation implementation-wise. The error needs to be reported from InternalIterator when trying to access value. But there are ~700 call sites of `InternalIterator::value()`, most of which either can't hit the error condition (because the iterator is reading from memtable or from index or something) or wouldn't benefit from the deferred loading of the value (e.g. compaction iterator that reads all values anyway). Adding error handling to all these call sites would needlessly bloat the code. So instead I made the deferred value loading optional: only the call sites that may use deferred loading have to call the new method `PrepareValue()` before calling `value()`. The feature is enabled with a new bool argument `allow_unprepared_value` to a bunch of methods that create iterators (it wouldn't make sense to put it in ReadOptions because it's completely internal to iterators, with virtually no user-visible effect). Lmk if you have better ideas.
Note that the deferred value loading only happens for *internal* iterators. The user-visible iterator (DBIter) always prepares the value before returning from Seek/Next/etc. We could go further and add an API to defer that value loading too, but that's most likely not useful for LogDevice, so it doesn't seem worth the complexity for now.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6621
Test Plan: make -j5 check . Will also deploy to some logdevice test clusters and look at stats.
Reviewed By: siying
Differential Revision: D20786930
Pulled By: al13n321
fbshipit-source-id: 6da77d918bad3780522e918f17f4d5513d3e99ee
2020-04-16 02:37:23 +02:00
|
|
|
ASSERT_TRUE(iter->PrepareValue());
|
Add an option to put first key of each sst block in the index (#5289)
Summary:
The first key is used to defer reading the data block until this file gets to the top of merging iterator's heap. For short range scans, most files never make it to the top of the heap, so this change can reduce read amplification by a lot sometimes.
Consider the following workload. There are a few data streams (we'll be calling them "logs"), each stream consisting of a sequence of blobs (we'll be calling them "records"). Each record is identified by log ID and a sequence number within the log. RocksDB key is concatenation of log ID and sequence number (big endian). Reads are mostly relatively short range scans, each within a single log. Writes are mostly sequential for each log, but writes to different logs are randomly interleaved. Compactions are disabled; instead, when we accumulate a few tens of sst files, we create a new column family and start writing to it.
So, a typical sst file consists of a few ranges of blocks, each range corresponding to one log ID (we use FlushBlockPolicy to cut blocks at log boundaries). A typical read would go like this. First, iterator Seek() reads one block from each sst file. Then a series of Next()s move through one sst file (since writes to each log are mostly sequential) until the subiterator reaches the end of this log in this sst file; then Next() switches to the next sst file and reads sequentially from that, and so on. Often a range scan will only return records from a small number of blocks in small number of sst files; in this case, the cost of initial Seek() reading one block from each file may be bigger than the cost of reading the actually useful blocks.
Neither iterate_upper_bound nor bloom filters can prevent reading one block from each file in Seek(). But this PR can: if the index contains first key from each block, we don't have to read the block until this block actually makes it to the top of merging iterator's heap, so for short range scans we won't read any blocks from most of the sst files.
This PR does the deferred block loading inside value() call. This is not ideal: there's no good way to report an IO error from inside value(). As discussed with siying offline, it would probably be better to change InternalIterator's interface to explicitly fetch deferred value and get status. I'll do it in a separate PR.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/5289
Differential Revision: D15256423
Pulled By: al13n321
fbshipit-source-id: 750e4c39ce88e8d41662f701cf6275d9388ba46a
2019-06-25 05:50:35 +02:00
|
|
|
EXPECT_EQ("v7", iter->value().ToString());
|
|
|
|
EXPECT_EQ(0, stats->getTickerCount(BLOCK_CACHE_DATA_HIT));
|
|
|
|
|
|
|
|
// Step forward and reach the end.
|
|
|
|
iter->Next();
|
|
|
|
EXPECT_FALSE(iter->Valid());
|
|
|
|
EXPECT_EQ(3, stats->getTickerCount(BLOCK_CACHE_DATA_MISS));
|
|
|
|
EXPECT_EQ(0, stats->getTickerCount(BLOCK_CACHE_DATA_HIT));
|
|
|
|
|
|
|
|
// Seek to a single-key block and step forward without accessing value.
|
|
|
|
iter->Seek(ikey("aaca"));
|
|
|
|
ASSERT_TRUE(iter->Valid());
|
|
|
|
EXPECT_EQ(keys[2], iter->key().ToString());
|
|
|
|
EXPECT_EQ(use_first_key ? 0 : 1,
|
|
|
|
stats->getTickerCount(BLOCK_CACHE_DATA_HIT));
|
|
|
|
|
|
|
|
iter->Next();
|
|
|
|
ASSERT_TRUE(iter->Valid());
|
|
|
|
EXPECT_EQ(keys[3], iter->key().ToString());
|
|
|
|
EXPECT_EQ(use_first_key ? 1 : 2,
|
|
|
|
stats->getTickerCount(BLOCK_CACHE_DATA_HIT));
|
Properly report IO errors when IndexType::kBinarySearchWithFirstKey is used (#6621)
Summary:
Context: Index type `kBinarySearchWithFirstKey` added the ability for sst file iterator to sometimes report a key from index without reading the corresponding data block. This is useful when sst blocks are cut at some meaningful boundaries (e.g. one block per key prefix), and many seeks land between blocks (e.g. for each prefix, the ranges of keys in different sst files are nearly disjoint, so a typical seek needs to read a data block from only one file even if all files have the prefix). But this added a new error condition, which rocksdb code was really not equipped to deal with: `InternalIterator::value()` may fail with an IO error or Status::Incomplete, but it's just a method returning a Slice, with no way to report error instead. Before this PR, this type of error wasn't handled at all (an empty slice was returned), and kBinarySearchWithFirstKey implementation was considered a prototype.
Now that we (LogDevice) have experimented with kBinarySearchWithFirstKey for a while and confirmed that it's really useful, this PR is adding the missing error handling.
It's a pretty inconvenient situation implementation-wise. The error needs to be reported from InternalIterator when trying to access value. But there are ~700 call sites of `InternalIterator::value()`, most of which either can't hit the error condition (because the iterator is reading from memtable or from index or something) or wouldn't benefit from the deferred loading of the value (e.g. compaction iterator that reads all values anyway). Adding error handling to all these call sites would needlessly bloat the code. So instead I made the deferred value loading optional: only the call sites that may use deferred loading have to call the new method `PrepareValue()` before calling `value()`. The feature is enabled with a new bool argument `allow_unprepared_value` to a bunch of methods that create iterators (it wouldn't make sense to put it in ReadOptions because it's completely internal to iterators, with virtually no user-visible effect). Lmk if you have better ideas.
Note that the deferred value loading only happens for *internal* iterators. The user-visible iterator (DBIter) always prepares the value before returning from Seek/Next/etc. We could go further and add an API to defer that value loading too, but that's most likely not useful for LogDevice, so it doesn't seem worth the complexity for now.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6621
Test Plan: make -j5 check . Will also deploy to some logdevice test clusters and look at stats.
Reviewed By: siying
Differential Revision: D20786930
Pulled By: al13n321
fbshipit-source-id: 6da77d918bad3780522e918f17f4d5513d3e99ee
2020-04-16 02:37:23 +02:00
|
|
|
ASSERT_TRUE(iter->PrepareValue());
|
Add an option to put first key of each sst block in the index (#5289)
Summary:
The first key is used to defer reading the data block until this file gets to the top of merging iterator's heap. For short range scans, most files never make it to the top of the heap, so this change can reduce read amplification by a lot sometimes.
Consider the following workload. There are a few data streams (we'll be calling them "logs"), each stream consisting of a sequence of blobs (we'll be calling them "records"). Each record is identified by log ID and a sequence number within the log. RocksDB key is concatenation of log ID and sequence number (big endian). Reads are mostly relatively short range scans, each within a single log. Writes are mostly sequential for each log, but writes to different logs are randomly interleaved. Compactions are disabled; instead, when we accumulate a few tens of sst files, we create a new column family and start writing to it.
So, a typical sst file consists of a few ranges of blocks, each range corresponding to one log ID (we use FlushBlockPolicy to cut blocks at log boundaries). A typical read would go like this. First, iterator Seek() reads one block from each sst file. Then a series of Next()s move through one sst file (since writes to each log are mostly sequential) until the subiterator reaches the end of this log in this sst file; then Next() switches to the next sst file and reads sequentially from that, and so on. Often a range scan will only return records from a small number of blocks in small number of sst files; in this case, the cost of initial Seek() reading one block from each file may be bigger than the cost of reading the actually useful blocks.
Neither iterate_upper_bound nor bloom filters can prevent reading one block from each file in Seek(). But this PR can: if the index contains first key from each block, we don't have to read the block until this block actually makes it to the top of merging iterator's heap, so for short range scans we won't read any blocks from most of the sst files.
This PR does the deferred block loading inside value() call. This is not ideal: there's no good way to report an IO error from inside value(). As discussed with siying offline, it would probably be better to change InternalIterator's interface to explicitly fetch deferred value and get status. I'll do it in a separate PR.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/5289
Differential Revision: D15256423
Pulled By: al13n321
fbshipit-source-id: 750e4c39ce88e8d41662f701cf6275d9388ba46a
2019-06-25 05:50:35 +02:00
|
|
|
EXPECT_EQ("v3", iter->value().ToString());
|
|
|
|
EXPECT_EQ(2, stats->getTickerCount(BLOCK_CACHE_DATA_HIT));
|
|
|
|
EXPECT_EQ(3, stats->getTickerCount(BLOCK_CACHE_DATA_MISS));
|
|
|
|
|
|
|
|
// Seek between blocks and step back without accessing value.
|
|
|
|
iter->Seek(ikey("aaca"));
|
|
|
|
ASSERT_TRUE(iter->Valid());
|
|
|
|
EXPECT_EQ(keys[2], iter->key().ToString());
|
|
|
|
EXPECT_EQ(use_first_key ? 2 : 3,
|
|
|
|
stats->getTickerCount(BLOCK_CACHE_DATA_HIT));
|
|
|
|
EXPECT_EQ(3, stats->getTickerCount(BLOCK_CACHE_DATA_MISS));
|
|
|
|
|
|
|
|
iter->Prev();
|
|
|
|
ASSERT_TRUE(iter->Valid());
|
|
|
|
EXPECT_EQ(keys[1], iter->key().ToString());
|
|
|
|
EXPECT_EQ(use_first_key ? 2 : 3,
|
|
|
|
stats->getTickerCount(BLOCK_CACHE_DATA_HIT));
|
|
|
|
// All blocks are in cache now, there'll be no more misses ever.
|
|
|
|
EXPECT_EQ(4, stats->getTickerCount(BLOCK_CACHE_DATA_MISS));
|
Properly report IO errors when IndexType::kBinarySearchWithFirstKey is used (#6621)
Summary:
Context: Index type `kBinarySearchWithFirstKey` added the ability for sst file iterator to sometimes report a key from index without reading the corresponding data block. This is useful when sst blocks are cut at some meaningful boundaries (e.g. one block per key prefix), and many seeks land between blocks (e.g. for each prefix, the ranges of keys in different sst files are nearly disjoint, so a typical seek needs to read a data block from only one file even if all files have the prefix). But this added a new error condition, which rocksdb code was really not equipped to deal with: `InternalIterator::value()` may fail with an IO error or Status::Incomplete, but it's just a method returning a Slice, with no way to report error instead. Before this PR, this type of error wasn't handled at all (an empty slice was returned), and kBinarySearchWithFirstKey implementation was considered a prototype.
Now that we (LogDevice) have experimented with kBinarySearchWithFirstKey for a while and confirmed that it's really useful, this PR is adding the missing error handling.
It's a pretty inconvenient situation implementation-wise. The error needs to be reported from InternalIterator when trying to access value. But there are ~700 call sites of `InternalIterator::value()`, most of which either can't hit the error condition (because the iterator is reading from memtable or from index or something) or wouldn't benefit from the deferred loading of the value (e.g. compaction iterator that reads all values anyway). Adding error handling to all these call sites would needlessly bloat the code. So instead I made the deferred value loading optional: only the call sites that may use deferred loading have to call the new method `PrepareValue()` before calling `value()`. The feature is enabled with a new bool argument `allow_unprepared_value` to a bunch of methods that create iterators (it wouldn't make sense to put it in ReadOptions because it's completely internal to iterators, with virtually no user-visible effect). Lmk if you have better ideas.
Note that the deferred value loading only happens for *internal* iterators. The user-visible iterator (DBIter) always prepares the value before returning from Seek/Next/etc. We could go further and add an API to defer that value loading too, but that's most likely not useful for LogDevice, so it doesn't seem worth the complexity for now.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6621
Test Plan: make -j5 check . Will also deploy to some logdevice test clusters and look at stats.
Reviewed By: siying
Differential Revision: D20786930
Pulled By: al13n321
fbshipit-source-id: 6da77d918bad3780522e918f17f4d5513d3e99ee
2020-04-16 02:37:23 +02:00
|
|
|
ASSERT_TRUE(iter->PrepareValue());
|
Add an option to put first key of each sst block in the index (#5289)
Summary:
The first key is used to defer reading the data block until this file gets to the top of merging iterator's heap. For short range scans, most files never make it to the top of the heap, so this change can reduce read amplification by a lot sometimes.
Consider the following workload. There are a few data streams (we'll be calling them "logs"), each stream consisting of a sequence of blobs (we'll be calling them "records"). Each record is identified by log ID and a sequence number within the log. RocksDB key is concatenation of log ID and sequence number (big endian). Reads are mostly relatively short range scans, each within a single log. Writes are mostly sequential for each log, but writes to different logs are randomly interleaved. Compactions are disabled; instead, when we accumulate a few tens of sst files, we create a new column family and start writing to it.
So, a typical sst file consists of a few ranges of blocks, each range corresponding to one log ID (we use FlushBlockPolicy to cut blocks at log boundaries). A typical read would go like this. First, iterator Seek() reads one block from each sst file. Then a series of Next()s move through one sst file (since writes to each log are mostly sequential) until the subiterator reaches the end of this log in this sst file; then Next() switches to the next sst file and reads sequentially from that, and so on. Often a range scan will only return records from a small number of blocks in small number of sst files; in this case, the cost of initial Seek() reading one block from each file may be bigger than the cost of reading the actually useful blocks.
Neither iterate_upper_bound nor bloom filters can prevent reading one block from each file in Seek(). But this PR can: if the index contains first key from each block, we don't have to read the block until this block actually makes it to the top of merging iterator's heap, so for short range scans we won't read any blocks from most of the sst files.
This PR does the deferred block loading inside value() call. This is not ideal: there's no good way to report an IO error from inside value(). As discussed with siying offline, it would probably be better to change InternalIterator's interface to explicitly fetch deferred value and get status. I'll do it in a separate PR.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/5289
Differential Revision: D15256423
Pulled By: al13n321
fbshipit-source-id: 750e4c39ce88e8d41662f701cf6275d9388ba46a
2019-06-25 05:50:35 +02:00
|
|
|
EXPECT_EQ("v1", iter->value().ToString());
|
|
|
|
|
|
|
|
// Next into the next block again.
|
|
|
|
iter->Next();
|
|
|
|
ASSERT_TRUE(iter->Valid());
|
|
|
|
EXPECT_EQ(keys[2], iter->key().ToString());
|
|
|
|
EXPECT_EQ(use_first_key ? 2 : 4,
|
|
|
|
stats->getTickerCount(BLOCK_CACHE_DATA_HIT));
|
|
|
|
|
|
|
|
// Seek to first and step back without accessing value.
|
|
|
|
iter->SeekToFirst();
|
|
|
|
ASSERT_TRUE(iter->Valid());
|
|
|
|
EXPECT_EQ(keys[0], iter->key().ToString());
|
|
|
|
EXPECT_EQ(use_first_key ? 2 : 5,
|
|
|
|
stats->getTickerCount(BLOCK_CACHE_DATA_HIT));
|
|
|
|
|
|
|
|
iter->Prev();
|
|
|
|
EXPECT_FALSE(iter->Valid());
|
|
|
|
EXPECT_EQ(use_first_key ? 2 : 5,
|
|
|
|
stats->getTickerCount(BLOCK_CACHE_DATA_HIT));
|
|
|
|
|
|
|
|
// Do some SeekForPrev() and SeekToLast() just to cover all methods.
|
|
|
|
iter->SeekForPrev(ikey("caad"));
|
|
|
|
ASSERT_TRUE(iter->Valid());
|
|
|
|
EXPECT_EQ(keys[4], iter->key().ToString());
|
|
|
|
EXPECT_EQ(use_first_key ? 3 : 6,
|
|
|
|
stats->getTickerCount(BLOCK_CACHE_DATA_HIT));
|
Properly report IO errors when IndexType::kBinarySearchWithFirstKey is used (#6621)
Summary:
Context: Index type `kBinarySearchWithFirstKey` added the ability for sst file iterator to sometimes report a key from index without reading the corresponding data block. This is useful when sst blocks are cut at some meaningful boundaries (e.g. one block per key prefix), and many seeks land between blocks (e.g. for each prefix, the ranges of keys in different sst files are nearly disjoint, so a typical seek needs to read a data block from only one file even if all files have the prefix). But this added a new error condition, which rocksdb code was really not equipped to deal with: `InternalIterator::value()` may fail with an IO error or Status::Incomplete, but it's just a method returning a Slice, with no way to report error instead. Before this PR, this type of error wasn't handled at all (an empty slice was returned), and kBinarySearchWithFirstKey implementation was considered a prototype.
Now that we (LogDevice) have experimented with kBinarySearchWithFirstKey for a while and confirmed that it's really useful, this PR is adding the missing error handling.
It's a pretty inconvenient situation implementation-wise. The error needs to be reported from InternalIterator when trying to access value. But there are ~700 call sites of `InternalIterator::value()`, most of which either can't hit the error condition (because the iterator is reading from memtable or from index or something) or wouldn't benefit from the deferred loading of the value (e.g. compaction iterator that reads all values anyway). Adding error handling to all these call sites would needlessly bloat the code. So instead I made the deferred value loading optional: only the call sites that may use deferred loading have to call the new method `PrepareValue()` before calling `value()`. The feature is enabled with a new bool argument `allow_unprepared_value` to a bunch of methods that create iterators (it wouldn't make sense to put it in ReadOptions because it's completely internal to iterators, with virtually no user-visible effect). Lmk if you have better ideas.
Note that the deferred value loading only happens for *internal* iterators. The user-visible iterator (DBIter) always prepares the value before returning from Seek/Next/etc. We could go further and add an API to defer that value loading too, but that's most likely not useful for LogDevice, so it doesn't seem worth the complexity for now.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6621
Test Plan: make -j5 check . Will also deploy to some logdevice test clusters and look at stats.
Reviewed By: siying
Differential Revision: D20786930
Pulled By: al13n321
fbshipit-source-id: 6da77d918bad3780522e918f17f4d5513d3e99ee
2020-04-16 02:37:23 +02:00
|
|
|
ASSERT_TRUE(iter->PrepareValue());
|
Add an option to put first key of each sst block in the index (#5289)
Summary:
The first key is used to defer reading the data block until this file gets to the top of merging iterator's heap. For short range scans, most files never make it to the top of the heap, so this change can reduce read amplification by a lot sometimes.
Consider the following workload. There are a few data streams (we'll be calling them "logs"), each stream consisting of a sequence of blobs (we'll be calling them "records"). Each record is identified by log ID and a sequence number within the log. RocksDB key is concatenation of log ID and sequence number (big endian). Reads are mostly relatively short range scans, each within a single log. Writes are mostly sequential for each log, but writes to different logs are randomly interleaved. Compactions are disabled; instead, when we accumulate a few tens of sst files, we create a new column family and start writing to it.
So, a typical sst file consists of a few ranges of blocks, each range corresponding to one log ID (we use FlushBlockPolicy to cut blocks at log boundaries). A typical read would go like this. First, iterator Seek() reads one block from each sst file. Then a series of Next()s move through one sst file (since writes to each log are mostly sequential) until the subiterator reaches the end of this log in this sst file; then Next() switches to the next sst file and reads sequentially from that, and so on. Often a range scan will only return records from a small number of blocks in small number of sst files; in this case, the cost of initial Seek() reading one block from each file may be bigger than the cost of reading the actually useful blocks.
Neither iterate_upper_bound nor bloom filters can prevent reading one block from each file in Seek(). But this PR can: if the index contains first key from each block, we don't have to read the block until this block actually makes it to the top of merging iterator's heap, so for short range scans we won't read any blocks from most of the sst files.
This PR does the deferred block loading inside value() call. This is not ideal: there's no good way to report an IO error from inside value(). As discussed with siying offline, it would probably be better to change InternalIterator's interface to explicitly fetch deferred value and get status. I'll do it in a separate PR.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/5289
Differential Revision: D15256423
Pulled By: al13n321
fbshipit-source-id: 750e4c39ce88e8d41662f701cf6275d9388ba46a
2019-06-25 05:50:35 +02:00
|
|
|
EXPECT_EQ("v4", iter->value().ToString());
|
|
|
|
EXPECT_EQ(use_first_key ? 3 : 6,
|
|
|
|
stats->getTickerCount(BLOCK_CACHE_DATA_HIT));
|
|
|
|
|
|
|
|
iter->SeekToLast();
|
|
|
|
ASSERT_TRUE(iter->Valid());
|
|
|
|
EXPECT_EQ(keys[7], iter->key().ToString());
|
|
|
|
EXPECT_EQ(use_first_key ? 4 : 7,
|
|
|
|
stats->getTickerCount(BLOCK_CACHE_DATA_HIT));
|
Properly report IO errors when IndexType::kBinarySearchWithFirstKey is used (#6621)
Summary:
Context: Index type `kBinarySearchWithFirstKey` added the ability for sst file iterator to sometimes report a key from index without reading the corresponding data block. This is useful when sst blocks are cut at some meaningful boundaries (e.g. one block per key prefix), and many seeks land between blocks (e.g. for each prefix, the ranges of keys in different sst files are nearly disjoint, so a typical seek needs to read a data block from only one file even if all files have the prefix). But this added a new error condition, which rocksdb code was really not equipped to deal with: `InternalIterator::value()` may fail with an IO error or Status::Incomplete, but it's just a method returning a Slice, with no way to report error instead. Before this PR, this type of error wasn't handled at all (an empty slice was returned), and kBinarySearchWithFirstKey implementation was considered a prototype.
Now that we (LogDevice) have experimented with kBinarySearchWithFirstKey for a while and confirmed that it's really useful, this PR is adding the missing error handling.
It's a pretty inconvenient situation implementation-wise. The error needs to be reported from InternalIterator when trying to access value. But there are ~700 call sites of `InternalIterator::value()`, most of which either can't hit the error condition (because the iterator is reading from memtable or from index or something) or wouldn't benefit from the deferred loading of the value (e.g. compaction iterator that reads all values anyway). Adding error handling to all these call sites would needlessly bloat the code. So instead I made the deferred value loading optional: only the call sites that may use deferred loading have to call the new method `PrepareValue()` before calling `value()`. The feature is enabled with a new bool argument `allow_unprepared_value` to a bunch of methods that create iterators (it wouldn't make sense to put it in ReadOptions because it's completely internal to iterators, with virtually no user-visible effect). Lmk if you have better ideas.
Note that the deferred value loading only happens for *internal* iterators. The user-visible iterator (DBIter) always prepares the value before returning from Seek/Next/etc. We could go further and add an API to defer that value loading too, but that's most likely not useful for LogDevice, so it doesn't seem worth the complexity for now.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6621
Test Plan: make -j5 check . Will also deploy to some logdevice test clusters and look at stats.
Reviewed By: siying
Differential Revision: D20786930
Pulled By: al13n321
fbshipit-source-id: 6da77d918bad3780522e918f17f4d5513d3e99ee
2020-04-16 02:37:23 +02:00
|
|
|
ASSERT_TRUE(iter->PrepareValue());
|
Add an option to put first key of each sst block in the index (#5289)
Summary:
The first key is used to defer reading the data block until this file gets to the top of merging iterator's heap. For short range scans, most files never make it to the top of the heap, so this change can reduce read amplification by a lot sometimes.
Consider the following workload. There are a few data streams (we'll be calling them "logs"), each stream consisting of a sequence of blobs (we'll be calling them "records"). Each record is identified by log ID and a sequence number within the log. RocksDB key is concatenation of log ID and sequence number (big endian). Reads are mostly relatively short range scans, each within a single log. Writes are mostly sequential for each log, but writes to different logs are randomly interleaved. Compactions are disabled; instead, when we accumulate a few tens of sst files, we create a new column family and start writing to it.
So, a typical sst file consists of a few ranges of blocks, each range corresponding to one log ID (we use FlushBlockPolicy to cut blocks at log boundaries). A typical read would go like this. First, iterator Seek() reads one block from each sst file. Then a series of Next()s move through one sst file (since writes to each log are mostly sequential) until the subiterator reaches the end of this log in this sst file; then Next() switches to the next sst file and reads sequentially from that, and so on. Often a range scan will only return records from a small number of blocks in small number of sst files; in this case, the cost of initial Seek() reading one block from each file may be bigger than the cost of reading the actually useful blocks.
Neither iterate_upper_bound nor bloom filters can prevent reading one block from each file in Seek(). But this PR can: if the index contains first key from each block, we don't have to read the block until this block actually makes it to the top of merging iterator's heap, so for short range scans we won't read any blocks from most of the sst files.
This PR does the deferred block loading inside value() call. This is not ideal: there's no good way to report an IO error from inside value(). As discussed with siying offline, it would probably be better to change InternalIterator's interface to explicitly fetch deferred value and get status. I'll do it in a separate PR.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/5289
Differential Revision: D15256423
Pulled By: al13n321
fbshipit-source-id: 750e4c39ce88e8d41662f701cf6275d9388ba46a
2019-06-25 05:50:35 +02:00
|
|
|
EXPECT_EQ("v7", iter->value().ToString());
|
|
|
|
EXPECT_EQ(use_first_key ? 4 : 7,
|
|
|
|
stats->getTickerCount(BLOCK_CACHE_DATA_HIT));
|
|
|
|
|
|
|
|
EXPECT_EQ(4, stats->getTickerCount(BLOCK_CACHE_DATA_MISS));
|
|
|
|
|
|
|
|
c.ResetTableReader();
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
TEST_P(BlockBasedTableTest, BinaryIndexWithFirstKeyGlobalSeqno) {
|
|
|
|
BlockBasedTableOptions table_options = GetBlockBasedTableOptions();
|
|
|
|
table_options.index_type = BlockBasedTableOptions::kBinarySearchWithFirstKey;
|
|
|
|
table_options.block_cache = NewLRUCache(10000);
|
|
|
|
Options options;
|
|
|
|
options.statistics = CreateDBStatistics();
|
|
|
|
Statistics* stats = options.statistics.get();
|
|
|
|
options.table_factory.reset(NewBlockBasedTableFactory(table_options));
|
|
|
|
std::unique_ptr<InternalKeyComparator> comparator(
|
|
|
|
new InternalKeyComparator(BytewiseComparator()));
|
|
|
|
const ImmutableCFOptions ioptions(options);
|
|
|
|
const MutableCFOptions moptions(options);
|
|
|
|
|
|
|
|
TableConstructor c(BytewiseComparator(), /* convert_to_internal_key */ false,
|
|
|
|
/* level */ -1, /* largest_seqno */ 42);
|
|
|
|
|
|
|
|
c.Add(InternalKey("b", 0, kTypeValue).Encode().ToString(), "x");
|
|
|
|
c.Add(InternalKey("c", 0, kTypeValue).Encode().ToString(), "y");
|
|
|
|
|
|
|
|
std::vector<std::string> keys;
|
|
|
|
stl_wrappers::KVMap kvmap;
|
|
|
|
c.Finish(options, ioptions, moptions, table_options, *comparator, &keys,
|
|
|
|
&kvmap);
|
|
|
|
ASSERT_EQ(2, keys.size());
|
|
|
|
|
|
|
|
auto reader = c.GetTableReader();
|
|
|
|
auto props = reader->GetTableProperties();
|
|
|
|
ASSERT_EQ(1u, props->num_data_blocks);
|
2020-08-04 00:21:56 +02:00
|
|
|
ReadOptions read_options;
|
Add an option to put first key of each sst block in the index (#5289)
Summary:
The first key is used to defer reading the data block until this file gets to the top of merging iterator's heap. For short range scans, most files never make it to the top of the heap, so this change can reduce read amplification by a lot sometimes.
Consider the following workload. There are a few data streams (we'll be calling them "logs"), each stream consisting of a sequence of blobs (we'll be calling them "records"). Each record is identified by log ID and a sequence number within the log. RocksDB key is concatenation of log ID and sequence number (big endian). Reads are mostly relatively short range scans, each within a single log. Writes are mostly sequential for each log, but writes to different logs are randomly interleaved. Compactions are disabled; instead, when we accumulate a few tens of sst files, we create a new column family and start writing to it.
So, a typical sst file consists of a few ranges of blocks, each range corresponding to one log ID (we use FlushBlockPolicy to cut blocks at log boundaries). A typical read would go like this. First, iterator Seek() reads one block from each sst file. Then a series of Next()s move through one sst file (since writes to each log are mostly sequential) until the subiterator reaches the end of this log in this sst file; then Next() switches to the next sst file and reads sequentially from that, and so on. Often a range scan will only return records from a small number of blocks in small number of sst files; in this case, the cost of initial Seek() reading one block from each file may be bigger than the cost of reading the actually useful blocks.
Neither iterate_upper_bound nor bloom filters can prevent reading one block from each file in Seek(). But this PR can: if the index contains first key from each block, we don't have to read the block until this block actually makes it to the top of merging iterator's heap, so for short range scans we won't read any blocks from most of the sst files.
This PR does the deferred block loading inside value() call. This is not ideal: there's no good way to report an IO error from inside value(). As discussed with siying offline, it would probably be better to change InternalIterator's interface to explicitly fetch deferred value and get status. I'll do it in a separate PR.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/5289
Differential Revision: D15256423
Pulled By: al13n321
fbshipit-source-id: 750e4c39ce88e8d41662f701cf6275d9388ba46a
2019-06-25 05:50:35 +02:00
|
|
|
std::unique_ptr<InternalIterator> iter(reader->NewIterator(
|
2020-08-04 00:21:56 +02:00
|
|
|
read_options, /*prefix_extractor=*/nullptr, /*arena=*/nullptr,
|
Properly report IO errors when IndexType::kBinarySearchWithFirstKey is used (#6621)
Summary:
Context: Index type `kBinarySearchWithFirstKey` added the ability for sst file iterator to sometimes report a key from index without reading the corresponding data block. This is useful when sst blocks are cut at some meaningful boundaries (e.g. one block per key prefix), and many seeks land between blocks (e.g. for each prefix, the ranges of keys in different sst files are nearly disjoint, so a typical seek needs to read a data block from only one file even if all files have the prefix). But this added a new error condition, which rocksdb code was really not equipped to deal with: `InternalIterator::value()` may fail with an IO error or Status::Incomplete, but it's just a method returning a Slice, with no way to report error instead. Before this PR, this type of error wasn't handled at all (an empty slice was returned), and kBinarySearchWithFirstKey implementation was considered a prototype.
Now that we (LogDevice) have experimented with kBinarySearchWithFirstKey for a while and confirmed that it's really useful, this PR is adding the missing error handling.
It's a pretty inconvenient situation implementation-wise. The error needs to be reported from InternalIterator when trying to access value. But there are ~700 call sites of `InternalIterator::value()`, most of which either can't hit the error condition (because the iterator is reading from memtable or from index or something) or wouldn't benefit from the deferred loading of the value (e.g. compaction iterator that reads all values anyway). Adding error handling to all these call sites would needlessly bloat the code. So instead I made the deferred value loading optional: only the call sites that may use deferred loading have to call the new method `PrepareValue()` before calling `value()`. The feature is enabled with a new bool argument `allow_unprepared_value` to a bunch of methods that create iterators (it wouldn't make sense to put it in ReadOptions because it's completely internal to iterators, with virtually no user-visible effect). Lmk if you have better ideas.
Note that the deferred value loading only happens for *internal* iterators. The user-visible iterator (DBIter) always prepares the value before returning from Seek/Next/etc. We could go further and add an API to defer that value loading too, but that's most likely not useful for LogDevice, so it doesn't seem worth the complexity for now.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6621
Test Plan: make -j5 check . Will also deploy to some logdevice test clusters and look at stats.
Reviewed By: siying
Differential Revision: D20786930
Pulled By: al13n321
fbshipit-source-id: 6da77d918bad3780522e918f17f4d5513d3e99ee
2020-04-16 02:37:23 +02:00
|
|
|
/*skip_filters=*/false, TableReaderCaller::kUncategorized,
|
|
|
|
/*compaction_readahead_size=*/0, /*allow_unprepared_value=*/true));
|
Add an option to put first key of each sst block in the index (#5289)
Summary:
The first key is used to defer reading the data block until this file gets to the top of merging iterator's heap. For short range scans, most files never make it to the top of the heap, so this change can reduce read amplification by a lot sometimes.
Consider the following workload. There are a few data streams (we'll be calling them "logs"), each stream consisting of a sequence of blobs (we'll be calling them "records"). Each record is identified by log ID and a sequence number within the log. RocksDB key is concatenation of log ID and sequence number (big endian). Reads are mostly relatively short range scans, each within a single log. Writes are mostly sequential for each log, but writes to different logs are randomly interleaved. Compactions are disabled; instead, when we accumulate a few tens of sst files, we create a new column family and start writing to it.
So, a typical sst file consists of a few ranges of blocks, each range corresponding to one log ID (we use FlushBlockPolicy to cut blocks at log boundaries). A typical read would go like this. First, iterator Seek() reads one block from each sst file. Then a series of Next()s move through one sst file (since writes to each log are mostly sequential) until the subiterator reaches the end of this log in this sst file; then Next() switches to the next sst file and reads sequentially from that, and so on. Often a range scan will only return records from a small number of blocks in small number of sst files; in this case, the cost of initial Seek() reading one block from each file may be bigger than the cost of reading the actually useful blocks.
Neither iterate_upper_bound nor bloom filters can prevent reading one block from each file in Seek(). But this PR can: if the index contains first key from each block, we don't have to read the block until this block actually makes it to the top of merging iterator's heap, so for short range scans we won't read any blocks from most of the sst files.
This PR does the deferred block loading inside value() call. This is not ideal: there's no good way to report an IO error from inside value(). As discussed with siying offline, it would probably be better to change InternalIterator's interface to explicitly fetch deferred value and get status. I'll do it in a separate PR.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/5289
Differential Revision: D15256423
Pulled By: al13n321
fbshipit-source-id: 750e4c39ce88e8d41662f701cf6275d9388ba46a
2019-06-25 05:50:35 +02:00
|
|
|
|
|
|
|
iter->Seek(InternalKey("a", 0, kTypeValue).Encode().ToString());
|
|
|
|
ASSERT_TRUE(iter->Valid());
|
|
|
|
EXPECT_EQ(InternalKey("b", 42, kTypeValue).Encode().ToString(),
|
|
|
|
iter->key().ToString());
|
|
|
|
EXPECT_NE(keys[0], iter->key().ToString());
|
|
|
|
// Key should have been served from index, without reading data blocks.
|
|
|
|
EXPECT_EQ(0, stats->getTickerCount(BLOCK_CACHE_DATA_MISS));
|
|
|
|
|
Properly report IO errors when IndexType::kBinarySearchWithFirstKey is used (#6621)
Summary:
Context: Index type `kBinarySearchWithFirstKey` added the ability for sst file iterator to sometimes report a key from index without reading the corresponding data block. This is useful when sst blocks are cut at some meaningful boundaries (e.g. one block per key prefix), and many seeks land between blocks (e.g. for each prefix, the ranges of keys in different sst files are nearly disjoint, so a typical seek needs to read a data block from only one file even if all files have the prefix). But this added a new error condition, which rocksdb code was really not equipped to deal with: `InternalIterator::value()` may fail with an IO error or Status::Incomplete, but it's just a method returning a Slice, with no way to report error instead. Before this PR, this type of error wasn't handled at all (an empty slice was returned), and kBinarySearchWithFirstKey implementation was considered a prototype.
Now that we (LogDevice) have experimented with kBinarySearchWithFirstKey for a while and confirmed that it's really useful, this PR is adding the missing error handling.
It's a pretty inconvenient situation implementation-wise. The error needs to be reported from InternalIterator when trying to access value. But there are ~700 call sites of `InternalIterator::value()`, most of which either can't hit the error condition (because the iterator is reading from memtable or from index or something) or wouldn't benefit from the deferred loading of the value (e.g. compaction iterator that reads all values anyway). Adding error handling to all these call sites would needlessly bloat the code. So instead I made the deferred value loading optional: only the call sites that may use deferred loading have to call the new method `PrepareValue()` before calling `value()`. The feature is enabled with a new bool argument `allow_unprepared_value` to a bunch of methods that create iterators (it wouldn't make sense to put it in ReadOptions because it's completely internal to iterators, with virtually no user-visible effect). Lmk if you have better ideas.
Note that the deferred value loading only happens for *internal* iterators. The user-visible iterator (DBIter) always prepares the value before returning from Seek/Next/etc. We could go further and add an API to defer that value loading too, but that's most likely not useful for LogDevice, so it doesn't seem worth the complexity for now.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6621
Test Plan: make -j5 check . Will also deploy to some logdevice test clusters and look at stats.
Reviewed By: siying
Differential Revision: D20786930
Pulled By: al13n321
fbshipit-source-id: 6da77d918bad3780522e918f17f4d5513d3e99ee
2020-04-16 02:37:23 +02:00
|
|
|
ASSERT_TRUE(iter->PrepareValue());
|
Add an option to put first key of each sst block in the index (#5289)
Summary:
The first key is used to defer reading the data block until this file gets to the top of merging iterator's heap. For short range scans, most files never make it to the top of the heap, so this change can reduce read amplification by a lot sometimes.
Consider the following workload. There are a few data streams (we'll be calling them "logs"), each stream consisting of a sequence of blobs (we'll be calling them "records"). Each record is identified by log ID and a sequence number within the log. RocksDB key is concatenation of log ID and sequence number (big endian). Reads are mostly relatively short range scans, each within a single log. Writes are mostly sequential for each log, but writes to different logs are randomly interleaved. Compactions are disabled; instead, when we accumulate a few tens of sst files, we create a new column family and start writing to it.
So, a typical sst file consists of a few ranges of blocks, each range corresponding to one log ID (we use FlushBlockPolicy to cut blocks at log boundaries). A typical read would go like this. First, iterator Seek() reads one block from each sst file. Then a series of Next()s move through one sst file (since writes to each log are mostly sequential) until the subiterator reaches the end of this log in this sst file; then Next() switches to the next sst file and reads sequentially from that, and so on. Often a range scan will only return records from a small number of blocks in small number of sst files; in this case, the cost of initial Seek() reading one block from each file may be bigger than the cost of reading the actually useful blocks.
Neither iterate_upper_bound nor bloom filters can prevent reading one block from each file in Seek(). But this PR can: if the index contains first key from each block, we don't have to read the block until this block actually makes it to the top of merging iterator's heap, so for short range scans we won't read any blocks from most of the sst files.
This PR does the deferred block loading inside value() call. This is not ideal: there's no good way to report an IO error from inside value(). As discussed with siying offline, it would probably be better to change InternalIterator's interface to explicitly fetch deferred value and get status. I'll do it in a separate PR.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/5289
Differential Revision: D15256423
Pulled By: al13n321
fbshipit-source-id: 750e4c39ce88e8d41662f701cf6275d9388ba46a
2019-06-25 05:50:35 +02:00
|
|
|
EXPECT_EQ("x", iter->value().ToString());
|
|
|
|
EXPECT_EQ(1, stats->getTickerCount(BLOCK_CACHE_DATA_MISS));
|
|
|
|
EXPECT_EQ(0, stats->getTickerCount(BLOCK_CACHE_DATA_HIT));
|
|
|
|
EXPECT_EQ(InternalKey("b", 42, kTypeValue).Encode().ToString(),
|
|
|
|
iter->key().ToString());
|
|
|
|
|
|
|
|
c.ResetTableReader();
|
|
|
|
}
|
|
|
|
|
2013-10-10 20:43:24 +02:00
|
|
|
// It's very hard to figure out the index block size of a block accurately.
|
|
|
|
// To make sure we get the index size, we just make sure as key number
|
|
|
|
// grows, the filter block size also grows.
|
2018-06-05 04:59:44 +02:00
|
|
|
TEST_P(BlockBasedTableTest, IndexSizeStat) {
|
2013-10-10 20:43:24 +02:00
|
|
|
uint64_t last_index_size = 0;
|
|
|
|
|
|
|
|
// we need to use random keys since the pure human readable texts
|
|
|
|
// may be well compressed, resulting insignifcant change of index
|
|
|
|
// block size.
|
|
|
|
Random rnd(test::RandomSeed());
|
|
|
|
std::vector<std::string> keys;
|
|
|
|
|
|
|
|
for (int i = 0; i < 100; ++i) {
|
2020-07-09 23:33:42 +02:00
|
|
|
keys.push_back(rnd.RandomString(10000));
|
2013-10-10 20:43:24 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
// Each time we load one more key to the table. the table index block
|
|
|
|
// size is expected to be larger than last time's.
|
|
|
|
for (size_t i = 1; i < keys.size(); ++i) {
|
2016-08-20 00:10:31 +02:00
|
|
|
TableConstructor c(BytewiseComparator(),
|
|
|
|
true /* convert_to_internal_key_ */);
|
2013-10-10 20:43:24 +02:00
|
|
|
for (size_t j = 0; j < i; ++j) {
|
|
|
|
c.Add(keys[j], "val");
|
|
|
|
}
|
|
|
|
|
|
|
|
std::vector<std::string> ks;
|
2015-09-02 22:58:22 +02:00
|
|
|
stl_wrappers::KVMap kvmap;
|
2013-11-20 07:00:48 +01:00
|
|
|
Options options;
|
2013-10-10 20:43:24 +02:00
|
|
|
options.compression = kNoCompression;
|
2018-06-05 04:59:44 +02:00
|
|
|
BlockBasedTableOptions table_options = GetBlockBasedTableOptions();
|
2014-08-25 23:22:05 +02:00
|
|
|
table_options.block_restart_interval = 1;
|
|
|
|
options.table_factory.reset(NewBlockBasedTableFactory(table_options));
|
2013-10-10 20:43:24 +02:00
|
|
|
|
2014-09-05 01:18:36 +02:00
|
|
|
const ImmutableCFOptions ioptions(options);
|
2018-05-21 23:33:55 +02:00
|
|
|
const MutableCFOptions moptions(options);
|
|
|
|
c.Finish(options, ioptions, moptions, table_options,
|
2014-08-25 23:22:05 +02:00
|
|
|
GetPlainInternalComparator(options.comparator), &ks, &kvmap);
|
2014-08-26 01:14:30 +02:00
|
|
|
auto index_size = c.GetTableReader()->GetTableProperties()->index_size;
|
2013-10-10 20:43:24 +02:00
|
|
|
ASSERT_GT(index_size, last_index_size);
|
|
|
|
last_index_size = index_size;
|
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 19:42:39 +02:00
|
|
|
c.ResetTableReader();
|
2013-10-10 20:43:24 +02:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2018-06-05 04:59:44 +02:00
|
|
|
TEST_P(BlockBasedTableTest, NumBlockStat) {
|
2013-10-10 20:43:24 +02:00
|
|
|
Random rnd(test::RandomSeed());
|
2016-08-20 00:10:31 +02:00
|
|
|
TableConstructor c(BytewiseComparator(), true /* convert_to_internal_key_ */);
|
2013-10-10 20:43:24 +02:00
|
|
|
Options options;
|
|
|
|
options.compression = kNoCompression;
|
2018-06-05 04:59:44 +02:00
|
|
|
BlockBasedTableOptions table_options = GetBlockBasedTableOptions();
|
2014-08-25 23:22:05 +02:00
|
|
|
table_options.block_restart_interval = 1;
|
|
|
|
table_options.block_size = 1000;
|
|
|
|
options.table_factory.reset(NewBlockBasedTableFactory(table_options));
|
2013-10-10 20:43:24 +02:00
|
|
|
|
|
|
|
for (int i = 0; i < 10; ++i) {
|
|
|
|
// the key/val are slightly smaller than block size, so that each block
|
|
|
|
// holds roughly one key/value pair.
|
2020-07-09 23:33:42 +02:00
|
|
|
c.Add(rnd.RandomString(900), "val");
|
2013-10-10 20:43:24 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
std::vector<std::string> ks;
|
2015-09-02 22:58:22 +02:00
|
|
|
stl_wrappers::KVMap kvmap;
|
2014-09-05 01:18:36 +02:00
|
|
|
const ImmutableCFOptions ioptions(options);
|
2018-05-21 23:33:55 +02:00
|
|
|
const MutableCFOptions moptions(options);
|
|
|
|
c.Finish(options, ioptions, moptions, table_options,
|
2014-08-25 23:22:05 +02:00
|
|
|
GetPlainInternalComparator(options.comparator), &ks, &kvmap);
|
2014-02-05 01:21:47 +01:00
|
|
|
ASSERT_EQ(kvmap.size(),
|
2014-08-26 01:14:30 +02:00
|
|
|
c.GetTableReader()->GetTableProperties()->num_data_blocks);
|
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 19:42:39 +02:00
|
|
|
c.ResetTableReader();
|
2013-10-10 20:43:24 +02:00
|
|
|
}
|
|
|
|
|
2019-07-17 22:02:00 +02:00
|
|
|
TEST_P(BlockBasedTableTest, TracingGetTest) {
|
|
|
|
TableConstructor c(BytewiseComparator());
|
|
|
|
Options options;
|
|
|
|
BlockBasedTableOptions table_options = GetBlockBasedTableOptions();
|
|
|
|
options.create_if_missing = true;
|
|
|
|
table_options.block_cache = NewLRUCache(1024 * 1024, 0);
|
|
|
|
table_options.cache_index_and_filter_blocks = true;
|
|
|
|
table_options.filter_policy.reset(NewBloomFilterPolicy(10, true));
|
|
|
|
options.table_factory.reset(new BlockBasedTableFactory(table_options));
|
|
|
|
SetupTracingTest(&c);
|
|
|
|
std::vector<std::string> keys;
|
|
|
|
stl_wrappers::KVMap kvmap;
|
|
|
|
ImmutableCFOptions ioptions(options);
|
|
|
|
MutableCFOptions moptions(options);
|
|
|
|
c.Finish(options, ioptions, moptions, table_options,
|
|
|
|
GetPlainInternalComparator(options.comparator), &keys, &kvmap);
|
|
|
|
std::string user_key = "k01";
|
|
|
|
InternalKey internal_key(user_key, 0, kTypeValue);
|
|
|
|
std::string encoded_key = internal_key.Encode().ToString();
|
|
|
|
for (uint32_t i = 1; i <= 2; i++) {
|
|
|
|
PinnableSlice value;
|
|
|
|
GetContext get_context(options.comparator, nullptr, nullptr, nullptr,
|
|
|
|
GetContext::kNotFound, user_key, &value, nullptr,
|
New API to get all merge operands for a Key (#5604)
Summary:
This is a new API added to db.h to allow for fetching all merge operands associated with a Key. The main motivation for this API is to support use cases where doing a full online merge is not necessary as it is performance sensitive. Example use-cases:
1. Update subset of columns and read subset of columns -
Imagine a SQL Table, a row is encoded as a K/V pair (as it is done in MyRocks). If there are many columns and users only updated one of them, we can use merge operator to reduce write amplification. While users only read one or two columns in the read query, this feature can avoid a full merging of the whole row, and save some CPU.
2. Updating very few attributes in a value which is a JSON-like document -
Updating one attribute can be done efficiently using merge operator, while reading back one attribute can be done more efficiently if we don't need to do a full merge.
----------------------------------------------------------------------------------------------------
API :
Status GetMergeOperands(
const ReadOptions& options, ColumnFamilyHandle* column_family,
const Slice& key, PinnableSlice* merge_operands,
GetMergeOperandsOptions* get_merge_operands_options,
int* number_of_operands)
Example usage :
int size = 100;
int number_of_operands = 0;
std::vector<PinnableSlice> values(size);
GetMergeOperandsOptions merge_operands_info;
db_->GetMergeOperands(ReadOptions(), db_->DefaultColumnFamily(), "k1", values.data(), merge_operands_info, &number_of_operands);
Description :
Returns all the merge operands corresponding to the key. If the number of merge operands in DB is greater than merge_operands_options.expected_max_number_of_operands no merge operands are returned and status is Incomplete. Merge operands returned are in the order of insertion.
merge_operands-> Points to an array of at-least merge_operands_options.expected_max_number_of_operands and the caller is responsible for allocating it. If the status returned is Incomplete then number_of_operands will contain the total number of merge operands found in DB for key.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/5604
Test Plan:
Added unit test and perf test in db_bench that can be run using the command:
./db_bench -benchmarks=getmergeoperands --merge_operator=sortlist
Differential Revision: D16657366
Pulled By: vjnadimpalli
fbshipit-source-id: 0faadd752351745224ee12d4ae9ef3cb529951bf
2019-08-06 23:22:34 +02:00
|
|
|
nullptr, true, nullptr, nullptr, nullptr, nullptr,
|
|
|
|
nullptr, nullptr, /*tracing_get_id=*/i);
|
2019-07-17 22:02:00 +02:00
|
|
|
get_perf_context()->Reset();
|
|
|
|
ASSERT_OK(c.GetTableReader()->Get(ReadOptions(), encoded_key, &get_context,
|
|
|
|
moptions.prefix_extractor.get()));
|
|
|
|
ASSERT_EQ(get_context.State(), GetContext::kFound);
|
|
|
|
ASSERT_EQ(value.ToString(), kDummyValue);
|
|
|
|
}
|
|
|
|
|
|
|
|
// Verify traces.
|
|
|
|
std::vector<BlockCacheTraceRecord> expected_records;
|
|
|
|
// The first two records should be prefetching index and filter blocks.
|
|
|
|
BlockCacheTraceRecord record;
|
|
|
|
record.block_type = TraceType::kBlockTraceIndexBlock;
|
|
|
|
record.caller = TableReaderCaller::kPrefetch;
|
|
|
|
record.is_cache_hit = Boolean::kFalse;
|
|
|
|
record.no_insert = Boolean::kFalse;
|
|
|
|
expected_records.push_back(record);
|
|
|
|
record.block_type = TraceType::kBlockTraceFilterBlock;
|
|
|
|
expected_records.push_back(record);
|
|
|
|
// Then we should have three records for one index, one filter, and one data
|
|
|
|
// block access.
|
|
|
|
record.get_id = 1;
|
|
|
|
record.block_type = TraceType::kBlockTraceIndexBlock;
|
|
|
|
record.caller = TableReaderCaller::kUserGet;
|
|
|
|
record.get_from_user_specified_snapshot = Boolean::kFalse;
|
|
|
|
record.referenced_key = encoded_key;
|
|
|
|
record.referenced_key_exist_in_block = Boolean::kTrue;
|
|
|
|
record.is_cache_hit = Boolean::kTrue;
|
|
|
|
expected_records.push_back(record);
|
|
|
|
record.block_type = TraceType::kBlockTraceFilterBlock;
|
|
|
|
expected_records.push_back(record);
|
|
|
|
record.is_cache_hit = Boolean::kFalse;
|
|
|
|
record.block_type = TraceType::kBlockTraceDataBlock;
|
|
|
|
expected_records.push_back(record);
|
|
|
|
// The second get should all observe cache hits.
|
|
|
|
record.is_cache_hit = Boolean::kTrue;
|
|
|
|
record.get_id = 2;
|
|
|
|
record.block_type = TraceType::kBlockTraceIndexBlock;
|
|
|
|
record.caller = TableReaderCaller::kUserGet;
|
|
|
|
record.get_from_user_specified_snapshot = Boolean::kFalse;
|
|
|
|
record.referenced_key = encoded_key;
|
|
|
|
expected_records.push_back(record);
|
|
|
|
record.block_type = TraceType::kBlockTraceFilterBlock;
|
|
|
|
expected_records.push_back(record);
|
|
|
|
record.block_type = TraceType::kBlockTraceDataBlock;
|
|
|
|
expected_records.push_back(record);
|
|
|
|
VerifyBlockAccessTrace(&c, expected_records);
|
|
|
|
c.ResetTableReader();
|
|
|
|
}
|
|
|
|
|
|
|
|
TEST_P(BlockBasedTableTest, TracingApproximateOffsetOfTest) {
|
|
|
|
TableConstructor c(BytewiseComparator());
|
|
|
|
Options options;
|
|
|
|
BlockBasedTableOptions table_options = GetBlockBasedTableOptions();
|
|
|
|
options.create_if_missing = true;
|
|
|
|
table_options.block_cache = NewLRUCache(1024 * 1024, 0);
|
|
|
|
table_options.cache_index_and_filter_blocks = true;
|
|
|
|
table_options.filter_policy.reset(NewBloomFilterPolicy(10, true));
|
|
|
|
options.table_factory.reset(new BlockBasedTableFactory(table_options));
|
|
|
|
SetupTracingTest(&c);
|
|
|
|
std::vector<std::string> keys;
|
|
|
|
stl_wrappers::KVMap kvmap;
|
|
|
|
ImmutableCFOptions ioptions(options);
|
|
|
|
MutableCFOptions moptions(options);
|
|
|
|
c.Finish(options, ioptions, moptions, table_options,
|
|
|
|
GetPlainInternalComparator(options.comparator), &keys, &kvmap);
|
|
|
|
for (uint32_t i = 1; i <= 2; i++) {
|
|
|
|
std::string user_key = "k01";
|
|
|
|
InternalKey internal_key(user_key, 0, kTypeValue);
|
|
|
|
std::string encoded_key = internal_key.Encode().ToString();
|
|
|
|
c.GetTableReader()->ApproximateOffsetOf(
|
|
|
|
encoded_key, TableReaderCaller::kUserApproximateSize);
|
|
|
|
}
|
|
|
|
// Verify traces.
|
|
|
|
std::vector<BlockCacheTraceRecord> expected_records;
|
|
|
|
// The first two records should be prefetching index and filter blocks.
|
|
|
|
BlockCacheTraceRecord record;
|
|
|
|
record.block_type = TraceType::kBlockTraceIndexBlock;
|
|
|
|
record.caller = TableReaderCaller::kPrefetch;
|
|
|
|
record.is_cache_hit = Boolean::kFalse;
|
|
|
|
record.no_insert = Boolean::kFalse;
|
|
|
|
expected_records.push_back(record);
|
|
|
|
record.block_type = TraceType::kBlockTraceFilterBlock;
|
|
|
|
expected_records.push_back(record);
|
|
|
|
// Then we should have two records for only index blocks.
|
|
|
|
record.block_type = TraceType::kBlockTraceIndexBlock;
|
|
|
|
record.caller = TableReaderCaller::kUserApproximateSize;
|
|
|
|
record.is_cache_hit = Boolean::kTrue;
|
|
|
|
expected_records.push_back(record);
|
|
|
|
expected_records.push_back(record);
|
|
|
|
VerifyBlockAccessTrace(&c, expected_records);
|
|
|
|
c.ResetTableReader();
|
|
|
|
}
|
|
|
|
|
|
|
|
TEST_P(BlockBasedTableTest, TracingIterator) {
|
|
|
|
TableConstructor c(BytewiseComparator());
|
|
|
|
Options options;
|
|
|
|
BlockBasedTableOptions table_options = GetBlockBasedTableOptions();
|
|
|
|
options.create_if_missing = true;
|
|
|
|
table_options.block_cache = NewLRUCache(1024 * 1024, 0);
|
|
|
|
table_options.cache_index_and_filter_blocks = true;
|
|
|
|
table_options.filter_policy.reset(NewBloomFilterPolicy(10, true));
|
|
|
|
options.table_factory.reset(new BlockBasedTableFactory(table_options));
|
|
|
|
SetupTracingTest(&c);
|
|
|
|
std::vector<std::string> keys;
|
|
|
|
stl_wrappers::KVMap kvmap;
|
|
|
|
ImmutableCFOptions ioptions(options);
|
|
|
|
MutableCFOptions moptions(options);
|
|
|
|
c.Finish(options, ioptions, moptions, table_options,
|
|
|
|
GetPlainInternalComparator(options.comparator), &keys, &kvmap);
|
|
|
|
|
|
|
|
for (uint32_t i = 1; i <= 2; i++) {
|
2020-08-04 00:21:56 +02:00
|
|
|
ReadOptions read_options;
|
2019-07-17 22:02:00 +02:00
|
|
|
std::unique_ptr<InternalIterator> iter(c.GetTableReader()->NewIterator(
|
2020-08-04 00:21:56 +02:00
|
|
|
read_options, moptions.prefix_extractor.get(), /*arena=*/nullptr,
|
2019-07-17 22:02:00 +02:00
|
|
|
/*skip_filters=*/false, TableReaderCaller::kUserIterator));
|
|
|
|
iter->SeekToFirst();
|
|
|
|
while (iter->Valid()) {
|
|
|
|
iter->key();
|
|
|
|
iter->value();
|
|
|
|
iter->Next();
|
|
|
|
}
|
|
|
|
ASSERT_OK(iter->status());
|
|
|
|
iter.reset();
|
|
|
|
}
|
|
|
|
|
|
|
|
// Verify traces.
|
|
|
|
std::vector<BlockCacheTraceRecord> expected_records;
|
|
|
|
// The first two records should be prefetching index and filter blocks.
|
|
|
|
BlockCacheTraceRecord record;
|
|
|
|
record.block_type = TraceType::kBlockTraceIndexBlock;
|
|
|
|
record.caller = TableReaderCaller::kPrefetch;
|
|
|
|
record.is_cache_hit = Boolean::kFalse;
|
|
|
|
record.no_insert = Boolean::kFalse;
|
|
|
|
expected_records.push_back(record);
|
|
|
|
record.block_type = TraceType::kBlockTraceFilterBlock;
|
|
|
|
expected_records.push_back(record);
|
|
|
|
// Then we should have three records for index and two data block access.
|
|
|
|
record.block_type = TraceType::kBlockTraceIndexBlock;
|
|
|
|
record.caller = TableReaderCaller::kUserIterator;
|
|
|
|
record.is_cache_hit = Boolean::kTrue;
|
|
|
|
expected_records.push_back(record);
|
|
|
|
record.block_type = TraceType::kBlockTraceDataBlock;
|
|
|
|
record.is_cache_hit = Boolean::kFalse;
|
|
|
|
expected_records.push_back(record);
|
|
|
|
expected_records.push_back(record);
|
|
|
|
// When we iterate this file for the second time, we should observe all cache
|
|
|
|
// hits.
|
|
|
|
record.block_type = TraceType::kBlockTraceIndexBlock;
|
|
|
|
record.is_cache_hit = Boolean::kTrue;
|
|
|
|
expected_records.push_back(record);
|
|
|
|
record.block_type = TraceType::kBlockTraceDataBlock;
|
|
|
|
expected_records.push_back(record);
|
|
|
|
expected_records.push_back(record);
|
|
|
|
VerifyBlockAccessTrace(&c, expected_records);
|
|
|
|
c.ResetTableReader();
|
|
|
|
}
|
|
|
|
|
2014-02-20 00:38:57 +01:00
|
|
|
// A simple tool that takes the snapshot of block cache statistics.
|
|
|
|
class BlockCachePropertiesSnapshot {
|
2013-11-13 07:46:51 +01:00
|
|
|
public:
|
2014-02-20 00:38:57 +01:00
|
|
|
explicit BlockCachePropertiesSnapshot(Statistics* statistics) {
|
2014-01-17 21:46:06 +01:00
|
|
|
block_cache_miss = statistics->getTickerCount(BLOCK_CACHE_MISS);
|
|
|
|
block_cache_hit = statistics->getTickerCount(BLOCK_CACHE_HIT);
|
|
|
|
index_block_cache_miss = statistics->getTickerCount(BLOCK_CACHE_INDEX_MISS);
|
|
|
|
index_block_cache_hit = statistics->getTickerCount(BLOCK_CACHE_INDEX_HIT);
|
|
|
|
data_block_cache_miss = statistics->getTickerCount(BLOCK_CACHE_DATA_MISS);
|
|
|
|
data_block_cache_hit = statistics->getTickerCount(BLOCK_CACHE_DATA_HIT);
|
2014-02-20 00:38:57 +01:00
|
|
|
filter_block_cache_miss =
|
|
|
|
statistics->getTickerCount(BLOCK_CACHE_FILTER_MISS);
|
|
|
|
filter_block_cache_hit = statistics->getTickerCount(BLOCK_CACHE_FILTER_HIT);
|
2015-10-08 00:17:20 +02:00
|
|
|
block_cache_bytes_read = statistics->getTickerCount(BLOCK_CACHE_BYTES_READ);
|
|
|
|
block_cache_bytes_write =
|
|
|
|
statistics->getTickerCount(BLOCK_CACHE_BYTES_WRITE);
|
2014-02-20 00:38:57 +01:00
|
|
|
}
|
|
|
|
|
2014-10-31 19:59:54 +01:00
|
|
|
void AssertIndexBlockStat(int64_t expected_index_block_cache_miss,
|
|
|
|
int64_t expected_index_block_cache_hit) {
|
|
|
|
ASSERT_EQ(expected_index_block_cache_miss, index_block_cache_miss);
|
|
|
|
ASSERT_EQ(expected_index_block_cache_hit, index_block_cache_hit);
|
2014-02-20 00:38:57 +01:00
|
|
|
}
|
|
|
|
|
2014-10-31 19:59:54 +01:00
|
|
|
void AssertFilterBlockStat(int64_t expected_filter_block_cache_miss,
|
|
|
|
int64_t expected_filter_block_cache_hit) {
|
|
|
|
ASSERT_EQ(expected_filter_block_cache_miss, filter_block_cache_miss);
|
|
|
|
ASSERT_EQ(expected_filter_block_cache_hit, filter_block_cache_hit);
|
2013-11-13 07:46:51 +01:00
|
|
|
}
|
|
|
|
|
2013-11-20 01:29:42 +01:00
|
|
|
// Check if the fetched props matches the expected ones.
|
2014-02-20 00:38:57 +01:00
|
|
|
// TODO(kailiu) Use this only when you disabled filter policy!
|
2014-10-31 19:59:54 +01:00
|
|
|
void AssertEqual(int64_t expected_index_block_cache_miss,
|
|
|
|
int64_t expected_index_block_cache_hit,
|
|
|
|
int64_t expected_data_block_cache_miss,
|
|
|
|
int64_t expected_data_block_cache_hit) const {
|
|
|
|
ASSERT_EQ(expected_index_block_cache_miss, index_block_cache_miss);
|
|
|
|
ASSERT_EQ(expected_index_block_cache_hit, index_block_cache_hit);
|
|
|
|
ASSERT_EQ(expected_data_block_cache_miss, data_block_cache_miss);
|
|
|
|
ASSERT_EQ(expected_data_block_cache_hit, data_block_cache_hit);
|
|
|
|
ASSERT_EQ(expected_index_block_cache_miss + expected_data_block_cache_miss,
|
|
|
|
block_cache_miss);
|
|
|
|
ASSERT_EQ(expected_index_block_cache_hit + expected_data_block_cache_hit,
|
|
|
|
block_cache_hit);
|
2013-11-13 07:46:51 +01:00
|
|
|
}
|
|
|
|
|
2015-10-08 00:17:20 +02:00
|
|
|
int64_t GetCacheBytesRead() { return block_cache_bytes_read; }
|
|
|
|
|
|
|
|
int64_t GetCacheBytesWrite() { return block_cache_bytes_write; }
|
|
|
|
|
2013-11-13 07:46:51 +01:00
|
|
|
private:
|
2014-02-05 01:21:47 +01:00
|
|
|
int64_t block_cache_miss = 0;
|
|
|
|
int64_t block_cache_hit = 0;
|
|
|
|
int64_t index_block_cache_miss = 0;
|
|
|
|
int64_t index_block_cache_hit = 0;
|
|
|
|
int64_t data_block_cache_miss = 0;
|
|
|
|
int64_t data_block_cache_hit = 0;
|
2014-02-20 00:38:57 +01:00
|
|
|
int64_t filter_block_cache_miss = 0;
|
|
|
|
int64_t filter_block_cache_hit = 0;
|
2015-10-08 00:17:20 +02:00
|
|
|
int64_t block_cache_bytes_read = 0;
|
|
|
|
int64_t block_cache_bytes_write = 0;
|
2013-11-13 07:46:51 +01:00
|
|
|
};
|
|
|
|
|
2014-02-20 00:38:57 +01:00
|
|
|
// Make sure, by default, index/filter blocks were pre-loaded (meaning we won't
|
|
|
|
// use block cache to store them).
|
2018-06-05 04:59:44 +02:00
|
|
|
TEST_P(BlockBasedTableTest, BlockCacheDisabledTest) {
|
2014-02-20 00:38:57 +01:00
|
|
|
Options options;
|
|
|
|
options.create_if_missing = true;
|
|
|
|
options.statistics = CreateDBStatistics();
|
2018-06-05 04:59:44 +02:00
|
|
|
BlockBasedTableOptions table_options = GetBlockBasedTableOptions();
|
2016-04-07 22:51:47 +02:00
|
|
|
table_options.block_cache = NewLRUCache(1024, 4);
|
2014-08-25 23:22:05 +02:00
|
|
|
table_options.filter_policy.reset(NewBloomFilterPolicy(10));
|
2014-02-20 00:38:57 +01:00
|
|
|
options.table_factory.reset(new BlockBasedTableFactory(table_options));
|
|
|
|
std::vector<std::string> keys;
|
2015-09-02 22:58:22 +02:00
|
|
|
stl_wrappers::KVMap kvmap;
|
2014-02-20 00:38:57 +01:00
|
|
|
|
2016-08-20 00:10:31 +02:00
|
|
|
TableConstructor c(BytewiseComparator(), true /* convert_to_internal_key_ */);
|
2014-02-20 00:38:57 +01:00
|
|
|
c.Add("key", "value");
|
2014-09-05 01:18:36 +02:00
|
|
|
const ImmutableCFOptions ioptions(options);
|
2018-05-21 23:33:55 +02:00
|
|
|
const MutableCFOptions moptions(options);
|
|
|
|
c.Finish(options, ioptions, moptions, table_options,
|
2014-08-25 23:22:05 +02:00
|
|
|
GetPlainInternalComparator(options.comparator), &keys, &kvmap);
|
2014-02-20 00:38:57 +01:00
|
|
|
|
|
|
|
// preloading filter/index blocks is enabled.
|
2014-08-26 01:14:30 +02:00
|
|
|
auto reader = dynamic_cast<BlockBasedTable*>(c.GetTableReader());
|
Move the filter readers out of the block cache (#5504)
Summary:
Currently, when the block cache is used for the filter block, it is not
really the block itself that is stored in the cache but a FilterBlockReader
object. Since this object is not pure data (it has, for instance, pointers that
might dangle, including in one case a back pointer to the TableReader), it's not
really sharable. To avoid the issues around this, the current code erases the
cache entries when the TableReader is closed (which, BTW, is not sufficient
since a concurrent TableReader might have picked up the object in the meantime).
Instead of doing this, the patch moves the FilterBlockReader out of the cache
altogether, and decouples the filter reader object from the filter block.
In particular, instead of the TableReader owning, or caching/pinning the
FilterBlockReader (based on the customer's settings), with the change the
TableReader unconditionally owns the FilterBlockReader, which in turn
owns/caches/pins the filter block. This change also enables us to reuse the code
paths historically used for data blocks for filters as well.
Note:
Eviction statistics for filter blocks are temporarily broken. We plan to fix this in a
separate phase.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/5504
Test Plan: make asan_check
Differential Revision: D16036974
Pulled By: ltamasi
fbshipit-source-id: 770f543c5fb4ed126fd1e04bfd3809cf4ff9c091
2019-07-16 22:11:23 +02:00
|
|
|
ASSERT_FALSE(reader->TEST_FilterBlockInCache());
|
Move the index readers out of the block cache (#5298)
Summary:
Currently, when the block cache is used for index blocks as well, it is
not really the index block that is stored in the cache but an
IndexReader object. Since this object is not pure data (it has, for
instance, pointers that might dangle), it's not really sharable. To
avoid the issues around this, the current code uses a dummy unique cache
key for each TableReader to store the IndexReader, and erases the
IndexReader entry when the TableReader is closed. Instead of doing this,
the new code moves the IndexReader out of the cache altogether. In
particular, instead of the TableReader owning, or caching/pinning the
IndexReader based on the customer's settings, the TableReader
unconditionally owns the IndexReader, which in turn owns/caches/pins
the index block (which is itself sharable and thus can be safely put in
the cache without any hacks).
Note: the change has two side effects:
1) Partitions of partitioned indexes no longer affect the read
amplification statistics.
2) Eviction statistics for index blocks are temporarily broken. We plan to fix
this in a separate phase.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/5298
Differential Revision: D15303203
Pulled By: ltamasi
fbshipit-source-id: 935a69ba59d87d5e44f42e2310619b790c366e47
2019-05-30 20:49:36 +02:00
|
|
|
ASSERT_FALSE(reader->TEST_IndexBlockInCache());
|
2014-02-20 00:38:57 +01:00
|
|
|
|
|
|
|
{
|
|
|
|
// nothing happens in the beginning
|
|
|
|
BlockCachePropertiesSnapshot props(options.statistics.get());
|
|
|
|
props.AssertIndexBlockStat(0, 0);
|
|
|
|
props.AssertFilterBlockStat(0, 0);
|
|
|
|
}
|
|
|
|
|
|
|
|
{
|
2014-09-29 20:09:09 +02:00
|
|
|
GetContext get_context(options.comparator, nullptr, nullptr, nullptr,
|
2015-03-03 19:59:36 +01:00
|
|
|
GetContext::kNotFound, Slice(), nullptr, nullptr,
|
New API to get all merge operands for a Key (#5604)
Summary:
This is a new API added to db.h to allow for fetching all merge operands associated with a Key. The main motivation for this API is to support use cases where doing a full online merge is not necessary as it is performance sensitive. Example use-cases:
1. Update subset of columns and read subset of columns -
Imagine a SQL Table, a row is encoded as a K/V pair (as it is done in MyRocks). If there are many columns and users only updated one of them, we can use merge operator to reduce write amplification. While users only read one or two columns in the read query, this feature can avoid a full merging of the whole row, and save some CPU.
2. Updating very few attributes in a value which is a JSON-like document -
Updating one attribute can be done efficiently using merge operator, while reading back one attribute can be done more efficiently if we don't need to do a full merge.
----------------------------------------------------------------------------------------------------
API :
Status GetMergeOperands(
const ReadOptions& options, ColumnFamilyHandle* column_family,
const Slice& key, PinnableSlice* merge_operands,
GetMergeOperandsOptions* get_merge_operands_options,
int* number_of_operands)
Example usage :
int size = 100;
int number_of_operands = 0;
std::vector<PinnableSlice> values(size);
GetMergeOperandsOptions merge_operands_info;
db_->GetMergeOperands(ReadOptions(), db_->DefaultColumnFamily(), "k1", values.data(), merge_operands_info, &number_of_operands);
Description :
Returns all the merge operands corresponding to the key. If the number of merge operands in DB is greater than merge_operands_options.expected_max_number_of_operands no merge operands are returned and status is Incomplete. Merge operands returned are in the order of insertion.
merge_operands-> Points to an array of at-least merge_operands_options.expected_max_number_of_operands and the caller is responsible for allocating it. If the status returned is Incomplete then number_of_operands will contain the total number of merge operands found in DB for key.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/5604
Test Plan:
Added unit test and perf test in db_bench that can be run using the command:
./db_bench -benchmarks=getmergeoperands --merge_operator=sortlist
Differential Revision: D16657366
Pulled By: vjnadimpalli
fbshipit-source-id: 0faadd752351745224ee12d4ae9ef3cb529951bf
2019-08-06 23:22:34 +02:00
|
|
|
nullptr, true, nullptr, nullptr);
|
2014-02-20 00:38:57 +01:00
|
|
|
// a hack that just to trigger BlockBasedTable::GetFilter.
|
2020-10-08 20:11:52 +02:00
|
|
|
ASSERT_OK(reader->Get(ReadOptions(), "non-exist-key", &get_context,
|
|
|
|
moptions.prefix_extractor.get()));
|
2014-02-20 00:38:57 +01:00
|
|
|
BlockCachePropertiesSnapshot props(options.statistics.get());
|
|
|
|
props.AssertIndexBlockStat(0, 0);
|
|
|
|
props.AssertFilterBlockStat(0, 0);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
// Due to the difficulities of the intersaction between statistics, this test
|
|
|
|
// only tests the case when "index block is put to block cache"
|
2018-06-05 04:59:44 +02:00
|
|
|
TEST_P(BlockBasedTableTest, FilterBlockInBlockCache) {
|
2013-11-13 07:46:51 +01:00
|
|
|
// -- Table construction
|
2013-11-20 07:00:48 +01:00
|
|
|
Options options;
|
2013-11-13 07:46:51 +01:00
|
|
|
options.create_if_missing = true;
|
2017-05-02 22:39:09 +02:00
|
|
|
options.statistics = CreateDBStatistics();
|
2014-01-24 19:57:15 +01:00
|
|
|
|
|
|
|
// Enable the cache for index/filter blocks
|
2018-06-05 04:59:44 +02:00
|
|
|
BlockBasedTableOptions table_options = GetBlockBasedTableOptions();
|
2019-09-17 00:14:51 +02:00
|
|
|
LRUCacheOptions co;
|
|
|
|
co.capacity = 2048;
|
|
|
|
co.num_shard_bits = 2;
|
|
|
|
co.metadata_charge_policy = kDontChargeCacheMetadata;
|
|
|
|
table_options.block_cache = NewLRUCache(co);
|
2014-01-24 19:57:15 +01:00
|
|
|
table_options.cache_index_and_filter_blocks = true;
|
|
|
|
options.table_factory.reset(new BlockBasedTableFactory(table_options));
|
2013-11-13 07:46:51 +01:00
|
|
|
std::vector<std::string> keys;
|
2015-09-02 22:58:22 +02:00
|
|
|
stl_wrappers::KVMap kvmap;
|
2013-11-13 07:46:51 +01:00
|
|
|
|
2016-08-20 00:10:31 +02:00
|
|
|
TableConstructor c(BytewiseComparator(), true /* convert_to_internal_key_ */);
|
2013-11-13 07:46:51 +01:00
|
|
|
c.Add("key", "value");
|
2014-09-05 01:18:36 +02:00
|
|
|
const ImmutableCFOptions ioptions(options);
|
2018-05-21 23:33:55 +02:00
|
|
|
const MutableCFOptions moptions(options);
|
|
|
|
c.Finish(options, ioptions, moptions, table_options,
|
2014-08-25 23:22:05 +02:00
|
|
|
GetPlainInternalComparator(options.comparator), &keys, &kvmap);
|
2014-02-20 00:38:57 +01:00
|
|
|
// preloading filter/index blocks is prohibited.
|
2014-10-22 20:52:35 +02:00
|
|
|
auto* reader = dynamic_cast<BlockBasedTable*>(c.GetTableReader());
|
Move the filter readers out of the block cache (#5504)
Summary:
Currently, when the block cache is used for the filter block, it is not
really the block itself that is stored in the cache but a FilterBlockReader
object. Since this object is not pure data (it has, for instance, pointers that
might dangle, including in one case a back pointer to the TableReader), it's not
really sharable. To avoid the issues around this, the current code erases the
cache entries when the TableReader is closed (which, BTW, is not sufficient
since a concurrent TableReader might have picked up the object in the meantime).
Instead of doing this, the patch moves the FilterBlockReader out of the cache
altogether, and decouples the filter reader object from the filter block.
In particular, instead of the TableReader owning, or caching/pinning the
FilterBlockReader (based on the customer's settings), with the change the
TableReader unconditionally owns the FilterBlockReader, which in turn
owns/caches/pins the filter block. This change also enables us to reuse the code
paths historically used for data blocks for filters as well.
Note:
Eviction statistics for filter blocks are temporarily broken. We plan to fix this in a
separate phase.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/5504
Test Plan: make asan_check
Differential Revision: D16036974
Pulled By: ltamasi
fbshipit-source-id: 770f543c5fb4ed126fd1e04bfd3809cf4ff9c091
2019-07-16 22:11:23 +02:00
|
|
|
ASSERT_FALSE(reader->TEST_FilterBlockInCache());
|
Move the index readers out of the block cache (#5298)
Summary:
Currently, when the block cache is used for index blocks as well, it is
not really the index block that is stored in the cache but an
IndexReader object. Since this object is not pure data (it has, for
instance, pointers that might dangle), it's not really sharable. To
avoid the issues around this, the current code uses a dummy unique cache
key for each TableReader to store the IndexReader, and erases the
IndexReader entry when the TableReader is closed. Instead of doing this,
the new code moves the IndexReader out of the cache altogether. In
particular, instead of the TableReader owning, or caching/pinning the
IndexReader based on the customer's settings, the TableReader
unconditionally owns the IndexReader, which in turn owns/caches/pins
the index block (which is itself sharable and thus can be safely put in
the cache without any hacks).
Note: the change has two side effects:
1) Partitions of partitioned indexes no longer affect the read
amplification statistics.
2) Eviction statistics for index blocks are temporarily broken. We plan to fix
this in a separate phase.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/5298
Differential Revision: D15303203
Pulled By: ltamasi
fbshipit-source-id: 935a69ba59d87d5e44f42e2310619b790c366e47
2019-05-30 20:49:36 +02:00
|
|
|
ASSERT_TRUE(reader->TEST_IndexBlockInCache());
|
2013-11-13 07:46:51 +01:00
|
|
|
|
|
|
|
// -- PART 1: Open with regular block cache.
|
|
|
|
// Since block_cache is disabled, no cache activities will be involved.
|
2018-11-09 20:17:34 +01:00
|
|
|
std::unique_ptr<InternalIterator> iter;
|
2013-11-13 07:46:51 +01:00
|
|
|
|
2015-10-08 00:17:20 +02:00
|
|
|
int64_t last_cache_bytes_read = 0;
|
2013-11-13 07:46:51 +01:00
|
|
|
// At first, no block will be accessed.
|
|
|
|
{
|
2014-02-20 00:38:57 +01:00
|
|
|
BlockCachePropertiesSnapshot props(options.statistics.get());
|
2013-11-13 07:46:51 +01:00
|
|
|
// index will be added to block cache.
|
2014-02-05 01:21:47 +01:00
|
|
|
props.AssertEqual(1, // index block miss
|
|
|
|
0, 0, 0);
|
2015-10-08 00:17:20 +02:00
|
|
|
ASSERT_EQ(props.GetCacheBytesRead(), 0);
|
|
|
|
ASSERT_EQ(props.GetCacheBytesWrite(),
|
2019-09-09 20:22:28 +02:00
|
|
|
static_cast<int64_t>(table_options.block_cache->GetUsage()));
|
2015-10-08 00:17:20 +02:00
|
|
|
last_cache_bytes_read = props.GetCacheBytesRead();
|
2013-11-13 07:46:51 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
// Only index block will be accessed
|
|
|
|
{
|
2018-05-21 23:33:55 +02:00
|
|
|
iter.reset(c.NewIterator(moptions.prefix_extractor.get()));
|
2014-02-20 00:38:57 +01:00
|
|
|
BlockCachePropertiesSnapshot props(options.statistics.get());
|
2013-11-13 07:46:51 +01:00
|
|
|
// NOTE: to help better highlight the "detla" of each ticker, I use
|
|
|
|
// <last_value> + <added_value> to indicate the increment of changed
|
|
|
|
// value; other numbers remain the same.
|
2014-02-05 01:21:47 +01:00
|
|
|
props.AssertEqual(1, 0 + 1, // index block hit
|
|
|
|
0, 0);
|
2015-10-08 00:17:20 +02:00
|
|
|
// Cache hit, bytes read from cache should increase
|
|
|
|
ASSERT_GT(props.GetCacheBytesRead(), last_cache_bytes_read);
|
|
|
|
ASSERT_EQ(props.GetCacheBytesWrite(),
|
2019-09-09 20:22:28 +02:00
|
|
|
static_cast<int64_t>(table_options.block_cache->GetUsage()));
|
2015-10-08 00:17:20 +02:00
|
|
|
last_cache_bytes_read = props.GetCacheBytesRead();
|
2013-11-13 07:46:51 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
// Only data block will be accessed
|
|
|
|
{
|
|
|
|
iter->SeekToFirst();
|
2020-10-08 20:11:52 +02:00
|
|
|
ASSERT_OK(iter->status());
|
2014-02-20 00:38:57 +01:00
|
|
|
BlockCachePropertiesSnapshot props(options.statistics.get());
|
2014-02-05 01:21:47 +01:00
|
|
|
props.AssertEqual(1, 1, 0 + 1, // data block miss
|
|
|
|
0);
|
2015-10-08 00:17:20 +02:00
|
|
|
// Cache miss, Bytes read from cache should not change
|
|
|
|
ASSERT_EQ(props.GetCacheBytesRead(), last_cache_bytes_read);
|
|
|
|
ASSERT_EQ(props.GetCacheBytesWrite(),
|
2019-09-09 20:22:28 +02:00
|
|
|
static_cast<int64_t>(table_options.block_cache->GetUsage()));
|
2015-10-08 00:17:20 +02:00
|
|
|
last_cache_bytes_read = props.GetCacheBytesRead();
|
2013-11-13 07:46:51 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
// Data block will be in cache
|
|
|
|
{
|
2018-05-21 23:33:55 +02:00
|
|
|
iter.reset(c.NewIterator(moptions.prefix_extractor.get()));
|
2013-11-13 07:46:51 +01:00
|
|
|
iter->SeekToFirst();
|
2020-10-08 20:11:52 +02:00
|
|
|
ASSERT_OK(iter->status());
|
2014-02-20 00:38:57 +01:00
|
|
|
BlockCachePropertiesSnapshot props(options.statistics.get());
|
2014-02-05 01:21:47 +01:00
|
|
|
props.AssertEqual(1, 1 + 1, /* index block hit */
|
|
|
|
1, 0 + 1 /* data block hit */);
|
2015-10-08 00:17:20 +02:00
|
|
|
// Cache hit, bytes read from cache should increase
|
|
|
|
ASSERT_GT(props.GetCacheBytesRead(), last_cache_bytes_read);
|
|
|
|
ASSERT_EQ(props.GetCacheBytesWrite(),
|
2019-09-09 20:22:28 +02:00
|
|
|
static_cast<int64_t>(table_options.block_cache->GetUsage()));
|
2013-11-13 07:46:51 +01:00
|
|
|
}
|
|
|
|
// release the iterator so that the block cache can reset correctly.
|
|
|
|
iter.reset();
|
2017-05-02 22:39:09 +02:00
|
|
|
|
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 19:42:39 +02:00
|
|
|
c.ResetTableReader();
|
|
|
|
|
2014-10-22 20:52:35 +02:00
|
|
|
// -- PART 2: Open with very small block cache
|
2013-11-13 07:46:51 +01:00
|
|
|
// In this test, no block will ever get hit since the block cache is
|
|
|
|
// too small to fit even one entry.
|
2016-04-07 22:51:47 +02:00
|
|
|
table_options.block_cache = NewLRUCache(1, 4);
|
2017-05-02 22:39:09 +02:00
|
|
|
options.statistics = CreateDBStatistics();
|
2014-08-25 23:22:05 +02:00
|
|
|
options.table_factory.reset(new BlockBasedTableFactory(table_options));
|
2014-09-05 01:18:36 +02:00
|
|
|
const ImmutableCFOptions ioptions2(options);
|
2018-05-21 23:33:55 +02:00
|
|
|
const MutableCFOptions moptions2(options);
|
2020-10-08 20:11:52 +02:00
|
|
|
ASSERT_OK(c.Reopen(ioptions2, moptions2));
|
2013-11-13 07:46:51 +01:00
|
|
|
{
|
2014-02-20 00:38:57 +01:00
|
|
|
BlockCachePropertiesSnapshot props(options.statistics.get());
|
2014-02-05 01:21:47 +01:00
|
|
|
props.AssertEqual(1, // index block miss
|
|
|
|
0, 0, 0);
|
2015-10-08 00:17:20 +02:00
|
|
|
// Cache miss, Bytes read from cache should not change
|
|
|
|
ASSERT_EQ(props.GetCacheBytesRead(), 0);
|
2013-11-13 07:46:51 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
{
|
|
|
|
// Both index and data block get accessed.
|
|
|
|
// It first cache index block then data block. But since the cache size
|
|
|
|
// is only 1, index block will be purged after data block is inserted.
|
2018-05-21 23:33:55 +02:00
|
|
|
iter.reset(c.NewIterator(moptions2.prefix_extractor.get()));
|
2014-02-20 00:38:57 +01:00
|
|
|
BlockCachePropertiesSnapshot props(options.statistics.get());
|
2014-02-05 01:21:47 +01:00
|
|
|
props.AssertEqual(1 + 1, // index block miss
|
|
|
|
0, 0, // data block miss
|
|
|
|
0);
|
2015-10-08 00:17:20 +02:00
|
|
|
// Cache hit, bytes read from cache should increase
|
|
|
|
ASSERT_EQ(props.GetCacheBytesRead(), 0);
|
2013-11-13 07:46:51 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
{
|
|
|
|
// SeekToFirst() accesses data block. With similar reason, we expect data
|
|
|
|
// block's cache miss.
|
|
|
|
iter->SeekToFirst();
|
2020-10-08 20:11:52 +02:00
|
|
|
ASSERT_OK(iter->status());
|
2014-02-20 00:38:57 +01:00
|
|
|
BlockCachePropertiesSnapshot props(options.statistics.get());
|
2014-02-05 01:21:47 +01:00
|
|
|
props.AssertEqual(2, 0, 0 + 1, // data block miss
|
|
|
|
0);
|
2015-10-08 00:17:20 +02:00
|
|
|
// Cache miss, Bytes read from cache should not change
|
|
|
|
ASSERT_EQ(props.GetCacheBytesRead(), 0);
|
2013-11-13 07:46:51 +01:00
|
|
|
}
|
2014-10-22 20:52:35 +02:00
|
|
|
iter.reset();
|
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 19:42:39 +02:00
|
|
|
c.ResetTableReader();
|
2014-10-22 20:52:35 +02:00
|
|
|
|
|
|
|
// -- PART 3: Open table with bloom filter enabled but not in SST file
|
2016-04-07 22:51:47 +02:00
|
|
|
table_options.block_cache = NewLRUCache(4096, 4);
|
2014-10-22 20:52:35 +02:00
|
|
|
table_options.cache_index_and_filter_blocks = false;
|
|
|
|
options.table_factory.reset(NewBlockBasedTableFactory(table_options));
|
|
|
|
|
|
|
|
TableConstructor c3(BytewiseComparator());
|
2014-10-22 22:53:35 +02:00
|
|
|
std::string user_key = "k01";
|
|
|
|
InternalKey internal_key(user_key, 0, kTypeValue);
|
|
|
|
c3.Add(internal_key.Encode().ToString(), "hello");
|
2014-10-22 20:52:35 +02:00
|
|
|
ImmutableCFOptions ioptions3(options);
|
2018-05-21 23:33:55 +02:00
|
|
|
MutableCFOptions moptions3(options);
|
2014-10-22 20:52:35 +02:00
|
|
|
// Generate table without filter policy
|
2018-05-21 23:33:55 +02:00
|
|
|
c3.Finish(options, ioptions3, moptions3, table_options,
|
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 19:42:39 +02:00
|
|
|
GetPlainInternalComparator(options.comparator), &keys, &kvmap);
|
|
|
|
c3.ResetTableReader();
|
|
|
|
|
2014-10-22 20:52:35 +02:00
|
|
|
// Open table with filter policy
|
|
|
|
table_options.filter_policy.reset(NewBloomFilterPolicy(1));
|
|
|
|
options.table_factory.reset(new BlockBasedTableFactory(table_options));
|
2017-05-02 22:39:09 +02:00
|
|
|
options.statistics = CreateDBStatistics();
|
2014-10-22 20:52:35 +02:00
|
|
|
ImmutableCFOptions ioptions4(options);
|
2018-05-21 23:33:55 +02:00
|
|
|
MutableCFOptions moptions4(options);
|
|
|
|
ASSERT_OK(c3.Reopen(ioptions4, moptions4));
|
2014-10-22 20:52:35 +02:00
|
|
|
reader = dynamic_cast<BlockBasedTable*>(c3.GetTableReader());
|
Move the filter readers out of the block cache (#5504)
Summary:
Currently, when the block cache is used for the filter block, it is not
really the block itself that is stored in the cache but a FilterBlockReader
object. Since this object is not pure data (it has, for instance, pointers that
might dangle, including in one case a back pointer to the TableReader), it's not
really sharable. To avoid the issues around this, the current code erases the
cache entries when the TableReader is closed (which, BTW, is not sufficient
since a concurrent TableReader might have picked up the object in the meantime).
Instead of doing this, the patch moves the FilterBlockReader out of the cache
altogether, and decouples the filter reader object from the filter block.
In particular, instead of the TableReader owning, or caching/pinning the
FilterBlockReader (based on the customer's settings), with the change the
TableReader unconditionally owns the FilterBlockReader, which in turn
owns/caches/pins the filter block. This change also enables us to reuse the code
paths historically used for data blocks for filters as well.
Note:
Eviction statistics for filter blocks are temporarily broken. We plan to fix this in a
separate phase.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/5504
Test Plan: make asan_check
Differential Revision: D16036974
Pulled By: ltamasi
fbshipit-source-id: 770f543c5fb4ed126fd1e04bfd3809cf4ff9c091
2019-07-16 22:11:23 +02:00
|
|
|
ASSERT_FALSE(reader->TEST_FilterBlockInCache());
|
2017-03-13 19:44:50 +01:00
|
|
|
PinnableSlice value;
|
2014-10-22 20:52:35 +02:00
|
|
|
GetContext get_context(options.comparator, nullptr, nullptr, nullptr,
|
2015-03-03 19:59:36 +01:00
|
|
|
GetContext::kNotFound, user_key, &value, nullptr,
|
New API to get all merge operands for a Key (#5604)
Summary:
This is a new API added to db.h to allow for fetching all merge operands associated with a Key. The main motivation for this API is to support use cases where doing a full online merge is not necessary as it is performance sensitive. Example use-cases:
1. Update subset of columns and read subset of columns -
Imagine a SQL Table, a row is encoded as a K/V pair (as it is done in MyRocks). If there are many columns and users only updated one of them, we can use merge operator to reduce write amplification. While users only read one or two columns in the read query, this feature can avoid a full merging of the whole row, and save some CPU.
2. Updating very few attributes in a value which is a JSON-like document -
Updating one attribute can be done efficiently using merge operator, while reading back one attribute can be done more efficiently if we don't need to do a full merge.
----------------------------------------------------------------------------------------------------
API :
Status GetMergeOperands(
const ReadOptions& options, ColumnFamilyHandle* column_family,
const Slice& key, PinnableSlice* merge_operands,
GetMergeOperandsOptions* get_merge_operands_options,
int* number_of_operands)
Example usage :
int size = 100;
int number_of_operands = 0;
std::vector<PinnableSlice> values(size);
GetMergeOperandsOptions merge_operands_info;
db_->GetMergeOperands(ReadOptions(), db_->DefaultColumnFamily(), "k1", values.data(), merge_operands_info, &number_of_operands);
Description :
Returns all the merge operands corresponding to the key. If the number of merge operands in DB is greater than merge_operands_options.expected_max_number_of_operands no merge operands are returned and status is Incomplete. Merge operands returned are in the order of insertion.
merge_operands-> Points to an array of at-least merge_operands_options.expected_max_number_of_operands and the caller is responsible for allocating it. If the status returned is Incomplete then number_of_operands will contain the total number of merge operands found in DB for key.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/5604
Test Plan:
Added unit test and perf test in db_bench that can be run using the command:
./db_bench -benchmarks=getmergeoperands --merge_operator=sortlist
Differential Revision: D16657366
Pulled By: vjnadimpalli
fbshipit-source-id: 0faadd752351745224ee12d4ae9ef3cb529951bf
2019-08-06 23:22:34 +02:00
|
|
|
nullptr, true, nullptr, nullptr);
|
2018-05-26 03:41:31 +02:00
|
|
|
ASSERT_OK(reader->Get(ReadOptions(), internal_key.Encode(), &get_context,
|
2018-05-21 23:33:55 +02:00
|
|
|
moptions4.prefix_extractor.get()));
|
2017-03-13 19:44:50 +01:00
|
|
|
ASSERT_STREQ(value.data(), "hello");
|
2014-10-22 20:52:35 +02:00
|
|
|
BlockCachePropertiesSnapshot props(options.statistics.get());
|
|
|
|
props.AssertFilterBlockStat(0, 0);
|
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 19:42:39 +02:00
|
|
|
c3.ResetTableReader();
|
2013-11-13 07:46:51 +01:00
|
|
|
}
|
|
|
|
|
2016-01-04 19:51:00 +01:00
|
|
|
void ValidateBlockSizeDeviation(int value, int expected) {
|
|
|
|
BlockBasedTableOptions table_options;
|
|
|
|
table_options.block_size_deviation = value;
|
|
|
|
BlockBasedTableFactory* factory = new BlockBasedTableFactory(table_options);
|
|
|
|
|
|
|
|
const BlockBasedTableOptions* normalized_table_options =
|
2020-09-15 01:59:00 +02:00
|
|
|
factory->GetOptions<BlockBasedTableOptions>();
|
2016-01-04 19:51:00 +01:00
|
|
|
ASSERT_EQ(normalized_table_options->block_size_deviation, expected);
|
|
|
|
|
|
|
|
delete factory;
|
|
|
|
}
|
|
|
|
|
|
|
|
void ValidateBlockRestartInterval(int value, int expected) {
|
|
|
|
BlockBasedTableOptions table_options;
|
|
|
|
table_options.block_restart_interval = value;
|
|
|
|
BlockBasedTableFactory* factory = new BlockBasedTableFactory(table_options);
|
|
|
|
|
|
|
|
const BlockBasedTableOptions* normalized_table_options =
|
2020-09-15 01:59:00 +02:00
|
|
|
factory->GetOptions<BlockBasedTableOptions>();
|
2016-01-04 19:51:00 +01:00
|
|
|
ASSERT_EQ(normalized_table_options->block_restart_interval, expected);
|
|
|
|
|
|
|
|
delete factory;
|
|
|
|
}
|
|
|
|
|
2018-06-05 04:59:44 +02:00
|
|
|
TEST_P(BlockBasedTableTest, InvalidOptions) {
|
2016-01-04 19:51:00 +01:00
|
|
|
// invalid values for block_size_deviation (<0 or >100) are silently set to 0
|
|
|
|
ValidateBlockSizeDeviation(-10, 0);
|
|
|
|
ValidateBlockSizeDeviation(-1, 0);
|
|
|
|
ValidateBlockSizeDeviation(0, 0);
|
|
|
|
ValidateBlockSizeDeviation(1, 1);
|
|
|
|
ValidateBlockSizeDeviation(99, 99);
|
|
|
|
ValidateBlockSizeDeviation(100, 100);
|
|
|
|
ValidateBlockSizeDeviation(101, 0);
|
|
|
|
ValidateBlockSizeDeviation(1000, 0);
|
|
|
|
|
|
|
|
// invalid values for block_restart_interval (<1) are silently set to 1
|
|
|
|
ValidateBlockRestartInterval(-10, 1);
|
|
|
|
ValidateBlockRestartInterval(-1, 1);
|
|
|
|
ValidateBlockRestartInterval(0, 1);
|
|
|
|
ValidateBlockRestartInterval(1, 1);
|
|
|
|
ValidateBlockRestartInterval(2, 2);
|
|
|
|
ValidateBlockRestartInterval(1000, 1000);
|
|
|
|
}
|
|
|
|
|
2018-06-05 04:59:44 +02:00
|
|
|
TEST_P(BlockBasedTableTest, BlockReadCountTest) {
|
2015-09-03 00:36:47 +02:00
|
|
|
// bloom_filter_type = 0 -- block-based filter
|
|
|
|
// bloom_filter_type = 0 -- full filter
|
|
|
|
for (int bloom_filter_type = 0; bloom_filter_type < 2; ++bloom_filter_type) {
|
|
|
|
for (int index_and_filter_in_cache = 0; index_and_filter_in_cache < 2;
|
|
|
|
++index_and_filter_in_cache) {
|
|
|
|
Options options;
|
|
|
|
options.create_if_missing = true;
|
|
|
|
|
2018-06-05 04:59:44 +02:00
|
|
|
BlockBasedTableOptions table_options = GetBlockBasedTableOptions();
|
2015-09-03 00:36:47 +02:00
|
|
|
table_options.block_cache = NewLRUCache(1, 0);
|
|
|
|
table_options.cache_index_and_filter_blocks = index_and_filter_in_cache;
|
|
|
|
table_options.filter_policy.reset(
|
|
|
|
NewBloomFilterPolicy(10, bloom_filter_type == 0));
|
|
|
|
options.table_factory.reset(new BlockBasedTableFactory(table_options));
|
|
|
|
std::vector<std::string> keys;
|
2015-09-03 01:05:53 +02:00
|
|
|
stl_wrappers::KVMap kvmap;
|
2015-09-03 00:36:47 +02:00
|
|
|
|
|
|
|
TableConstructor c(BytewiseComparator());
|
|
|
|
std::string user_key = "k04";
|
|
|
|
InternalKey internal_key(user_key, 0, kTypeValue);
|
|
|
|
std::string encoded_key = internal_key.Encode().ToString();
|
|
|
|
c.Add(encoded_key, "hello");
|
|
|
|
ImmutableCFOptions ioptions(options);
|
2018-05-21 23:33:55 +02:00
|
|
|
MutableCFOptions moptions(options);
|
2015-09-03 00:36:47 +02:00
|
|
|
// Generate table with filter policy
|
2018-05-21 23:33:55 +02:00
|
|
|
c.Finish(options, ioptions, moptions, table_options,
|
2015-09-03 00:36:47 +02:00
|
|
|
GetPlainInternalComparator(options.comparator), &keys, &kvmap);
|
|
|
|
auto reader = c.GetTableReader();
|
2017-03-13 19:44:50 +01:00
|
|
|
PinnableSlice value;
|
2019-07-04 03:45:36 +02:00
|
|
|
{
|
|
|
|
GetContext get_context(options.comparator, nullptr, nullptr, nullptr,
|
|
|
|
GetContext::kNotFound, user_key, &value, nullptr,
|
New API to get all merge operands for a Key (#5604)
Summary:
This is a new API added to db.h to allow for fetching all merge operands associated with a Key. The main motivation for this API is to support use cases where doing a full online merge is not necessary as it is performance sensitive. Example use-cases:
1. Update subset of columns and read subset of columns -
Imagine a SQL Table, a row is encoded as a K/V pair (as it is done in MyRocks). If there are many columns and users only updated one of them, we can use merge operator to reduce write amplification. While users only read one or two columns in the read query, this feature can avoid a full merging of the whole row, and save some CPU.
2. Updating very few attributes in a value which is a JSON-like document -
Updating one attribute can be done efficiently using merge operator, while reading back one attribute can be done more efficiently if we don't need to do a full merge.
----------------------------------------------------------------------------------------------------
API :
Status GetMergeOperands(
const ReadOptions& options, ColumnFamilyHandle* column_family,
const Slice& key, PinnableSlice* merge_operands,
GetMergeOperandsOptions* get_merge_operands_options,
int* number_of_operands)
Example usage :
int size = 100;
int number_of_operands = 0;
std::vector<PinnableSlice> values(size);
GetMergeOperandsOptions merge_operands_info;
db_->GetMergeOperands(ReadOptions(), db_->DefaultColumnFamily(), "k1", values.data(), merge_operands_info, &number_of_operands);
Description :
Returns all the merge operands corresponding to the key. If the number of merge operands in DB is greater than merge_operands_options.expected_max_number_of_operands no merge operands are returned and status is Incomplete. Merge operands returned are in the order of insertion.
merge_operands-> Points to an array of at-least merge_operands_options.expected_max_number_of_operands and the caller is responsible for allocating it. If the status returned is Incomplete then number_of_operands will contain the total number of merge operands found in DB for key.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/5604
Test Plan:
Added unit test and perf test in db_bench that can be run using the command:
./db_bench -benchmarks=getmergeoperands --merge_operator=sortlist
Differential Revision: D16657366
Pulled By: vjnadimpalli
fbshipit-source-id: 0faadd752351745224ee12d4ae9ef3cb529951bf
2019-08-06 23:22:34 +02:00
|
|
|
nullptr, true, nullptr, nullptr);
|
2019-07-04 03:45:36 +02:00
|
|
|
get_perf_context()->Reset();
|
|
|
|
ASSERT_OK(reader->Get(ReadOptions(), encoded_key, &get_context,
|
|
|
|
moptions.prefix_extractor.get()));
|
|
|
|
if (index_and_filter_in_cache) {
|
|
|
|
// data, index and filter block
|
|
|
|
ASSERT_EQ(get_perf_context()->block_read_count, 3);
|
|
|
|
ASSERT_EQ(get_perf_context()->index_block_read_count, 1);
|
|
|
|
ASSERT_EQ(get_perf_context()->filter_block_read_count, 1);
|
|
|
|
} else {
|
|
|
|
// just the data block
|
|
|
|
ASSERT_EQ(get_perf_context()->block_read_count, 1);
|
|
|
|
}
|
|
|
|
ASSERT_EQ(get_context.State(), GetContext::kFound);
|
|
|
|
ASSERT_STREQ(value.data(), "hello");
|
2015-09-03 00:36:47 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
// Get non-existing key
|
|
|
|
user_key = "does-not-exist";
|
|
|
|
internal_key = InternalKey(user_key, 0, kTypeValue);
|
|
|
|
encoded_key = internal_key.Encode().ToString();
|
|
|
|
|
2017-03-13 19:44:50 +01:00
|
|
|
value.Reset();
|
2019-07-04 03:45:36 +02:00
|
|
|
{
|
|
|
|
GetContext get_context(options.comparator, nullptr, nullptr, nullptr,
|
2015-09-03 00:36:47 +02:00
|
|
|
GetContext::kNotFound, user_key, &value, nullptr,
|
New API to get all merge operands for a Key (#5604)
Summary:
This is a new API added to db.h to allow for fetching all merge operands associated with a Key. The main motivation for this API is to support use cases where doing a full online merge is not necessary as it is performance sensitive. Example use-cases:
1. Update subset of columns and read subset of columns -
Imagine a SQL Table, a row is encoded as a K/V pair (as it is done in MyRocks). If there are many columns and users only updated one of them, we can use merge operator to reduce write amplification. While users only read one or two columns in the read query, this feature can avoid a full merging of the whole row, and save some CPU.
2. Updating very few attributes in a value which is a JSON-like document -
Updating one attribute can be done efficiently using merge operator, while reading back one attribute can be done more efficiently if we don't need to do a full merge.
----------------------------------------------------------------------------------------------------
API :
Status GetMergeOperands(
const ReadOptions& options, ColumnFamilyHandle* column_family,
const Slice& key, PinnableSlice* merge_operands,
GetMergeOperandsOptions* get_merge_operands_options,
int* number_of_operands)
Example usage :
int size = 100;
int number_of_operands = 0;
std::vector<PinnableSlice> values(size);
GetMergeOperandsOptions merge_operands_info;
db_->GetMergeOperands(ReadOptions(), db_->DefaultColumnFamily(), "k1", values.data(), merge_operands_info, &number_of_operands);
Description :
Returns all the merge operands corresponding to the key. If the number of merge operands in DB is greater than merge_operands_options.expected_max_number_of_operands no merge operands are returned and status is Incomplete. Merge operands returned are in the order of insertion.
merge_operands-> Points to an array of at-least merge_operands_options.expected_max_number_of_operands and the caller is responsible for allocating it. If the status returned is Incomplete then number_of_operands will contain the total number of merge operands found in DB for key.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/5604
Test Plan:
Added unit test and perf test in db_bench that can be run using the command:
./db_bench -benchmarks=getmergeoperands --merge_operator=sortlist
Differential Revision: D16657366
Pulled By: vjnadimpalli
fbshipit-source-id: 0faadd752351745224ee12d4ae9ef3cb529951bf
2019-08-06 23:22:34 +02:00
|
|
|
nullptr, true, nullptr, nullptr);
|
2019-07-04 03:45:36 +02:00
|
|
|
get_perf_context()->Reset();
|
|
|
|
ASSERT_OK(reader->Get(ReadOptions(), encoded_key, &get_context,
|
|
|
|
moptions.prefix_extractor.get()));
|
|
|
|
ASSERT_EQ(get_context.State(), GetContext::kNotFound);
|
|
|
|
}
|
2015-09-03 00:36:47 +02:00
|
|
|
|
|
|
|
if (index_and_filter_in_cache) {
|
|
|
|
if (bloom_filter_type == 0) {
|
|
|
|
// with block-based, we read index and then the filter
|
2017-06-03 02:12:39 +02:00
|
|
|
ASSERT_EQ(get_perf_context()->block_read_count, 2);
|
2019-06-19 04:00:03 +02:00
|
|
|
ASSERT_EQ(get_perf_context()->index_block_read_count, 1);
|
|
|
|
ASSERT_EQ(get_perf_context()->filter_block_read_count, 1);
|
2015-09-03 00:36:47 +02:00
|
|
|
} else {
|
|
|
|
// with full-filter, we read filter first and then we stop
|
2017-06-03 02:12:39 +02:00
|
|
|
ASSERT_EQ(get_perf_context()->block_read_count, 1);
|
2019-06-19 04:00:03 +02:00
|
|
|
ASSERT_EQ(get_perf_context()->filter_block_read_count, 1);
|
2015-09-03 00:36:47 +02:00
|
|
|
}
|
|
|
|
} else {
|
|
|
|
// filter is already in memory and it figures out that the key doesn't
|
|
|
|
// exist
|
2017-06-03 02:12:39 +02:00
|
|
|
ASSERT_EQ(get_perf_context()->block_read_count, 0);
|
2015-09-03 00:36:47 +02:00
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2018-06-05 04:59:44 +02:00
|
|
|
TEST_P(BlockBasedTableTest, BlockCacheLeak) {
|
2014-01-24 21:14:08 +01:00
|
|
|
// Check that when we reopen a table we don't lose access to blocks already
|
|
|
|
// in the cache. This test checks whether the Table actually makes use of the
|
|
|
|
// unique ID from the file.
|
|
|
|
|
|
|
|
Options opt;
|
2018-11-09 20:17:34 +01:00
|
|
|
std::unique_ptr<InternalKeyComparator> ikc;
|
2014-01-27 22:53:22 +01:00
|
|
|
ikc.reset(new test::PlainInternalKeyComparator(opt.comparator));
|
2014-01-24 21:14:08 +01:00
|
|
|
opt.compression = kNoCompression;
|
2018-06-05 04:59:44 +02:00
|
|
|
BlockBasedTableOptions table_options = GetBlockBasedTableOptions();
|
2014-08-25 23:22:05 +02:00
|
|
|
table_options.block_size = 1024;
|
|
|
|
// big enough so we don't ever lose cached values.
|
2016-04-07 22:51:47 +02:00
|
|
|
table_options.block_cache = NewLRUCache(16 * 1024 * 1024, 4);
|
2014-08-25 23:22:05 +02:00
|
|
|
opt.table_factory.reset(NewBlockBasedTableFactory(table_options));
|
2014-01-24 21:14:08 +01:00
|
|
|
|
2016-08-20 00:10:31 +02:00
|
|
|
TableConstructor c(BytewiseComparator(), true /* convert_to_internal_key_ */);
|
2014-01-24 21:14:08 +01:00
|
|
|
c.Add("k01", "hello");
|
|
|
|
c.Add("k02", "hello2");
|
|
|
|
c.Add("k03", std::string(10000, 'x'));
|
|
|
|
c.Add("k04", std::string(200000, 'x'));
|
|
|
|
c.Add("k05", std::string(300000, 'x'));
|
|
|
|
c.Add("k06", "hello3");
|
|
|
|
c.Add("k07", std::string(100000, 'x'));
|
|
|
|
std::vector<std::string> keys;
|
2015-09-02 22:58:22 +02:00
|
|
|
stl_wrappers::KVMap kvmap;
|
2014-09-05 01:18:36 +02:00
|
|
|
const ImmutableCFOptions ioptions(opt);
|
2018-05-21 23:33:55 +02:00
|
|
|
const MutableCFOptions moptions(opt);
|
|
|
|
c.Finish(opt, ioptions, moptions, table_options, *ikc, &keys, &kvmap);
|
2014-01-24 21:14:08 +01:00
|
|
|
|
2018-11-09 20:17:34 +01:00
|
|
|
std::unique_ptr<InternalIterator> iter(
|
2018-05-21 23:33:55 +02:00
|
|
|
c.NewIterator(moptions.prefix_extractor.get()));
|
2014-01-24 21:14:08 +01:00
|
|
|
iter->SeekToFirst();
|
|
|
|
while (iter->Valid()) {
|
|
|
|
iter->key();
|
|
|
|
iter->value();
|
|
|
|
iter->Next();
|
|
|
|
}
|
|
|
|
ASSERT_OK(iter->status());
|
Change and clarify the relationship between Valid(), status() and Seek*() for all iterators. Also fix some bugs
Summary:
Before this PR, Iterator/InternalIterator may simultaneously have non-ok status() and Valid() = true. That state means that the last operation failed, but the iterator is nevertheless positioned on some unspecified record. Likely intended uses of that are:
* If some sst files are corrupted, a normal iterator can be used to read the data from files that are not corrupted.
* When using read_tier = kBlockCacheTier, read the data that's in block cache, skipping over the data that is not.
However, this behavior wasn't documented well (and until recently the wiki on github had misleading incorrect information). In the code there's a lot of confusion about the relationship between status() and Valid(), and about whether Seek()/SeekToLast()/etc reset the status or not. There were a number of bugs caused by this confusion, both inside rocksdb and in the code that uses rocksdb (including ours).
This PR changes the convention to:
* If status() is not ok, Valid() always returns false.
* Any seek operation resets status. (Before the PR, it depended on iterator type and on particular error.)
This does sacrifice the two use cases listed above, but siying said it's ok.
Overview of the changes:
* A commit that adds missing status checks in MergingIterator. This fixes a bug that actually affects us, and we need it fixed. `DBIteratorTest.NonBlockingIterationBugRepro` explains the scenario.
* Changes to lots of iterator types to make all of them conform to the new convention. Some bug fixes along the way. By far the biggest changes are in DBIter, which is a big messy piece of code; I tried to make it less big and messy but mostly failed.
* A stress-test for DBIter, to gain some confidence that I didn't break it. It does a few million random operations on the iterator, while occasionally modifying the underlying data (like ForwardIterator does) and occasionally returning non-ok status from internal iterator.
To find the iterator types that needed changes I searched for "public .*Iterator" in the code. Here's an overview of all 27 iterator types:
Iterators that didn't need changes:
* status() is always ok(), or Valid() is always false: MemTableIterator, ModelIter, TestIterator, KVIter (2 classes with this name anonymous namespaces), LoggingForwardVectorIterator, VectorIterator, MockTableIterator, EmptyIterator, EmptyInternalIterator.
* Thin wrappers that always pass through Valid() and status(): ArenaWrappedDBIter, TtlIterator, InternalIteratorFromIterator.
Iterators with changes (see inline comments for details):
* DBIter - an overhaul:
- It used to silently skip corrupted keys (`FindParseableKey()`), which seems dangerous. This PR makes it just stop immediately after encountering a corrupted key, just like it would for other kinds of corruption. Let me know if there was actually some deeper meaning in this behavior and I should put it back.
- It had a few code paths silently discarding subiterator's status. The stress test caught a few.
- The backwards iteration code path was expecting the internal iterator's set of keys to be immutable. It's probably always true in practice at the moment, since ForwardIterator doesn't support backwards iteration, but this PR fixes it anyway. See added DBIteratorTest.ReverseToForwardBug for an example.
- Some parts of backwards iteration code path even did things like `assert(iter_->Valid())` after a seek, which is never a safe assumption.
- It used to not reset status on seek for some types of errors.
- Some simplifications and better comments.
- Some things got more complicated from the added error handling. I'm open to ideas for how to make it nicer.
* MergingIterator - check status after every operation on every subiterator, and in some places assert that valid subiterators have ok status.
* ForwardIterator - changed to the new convention, also slightly simplified.
* ForwardLevelIterator - fixed some bugs and simplified.
* LevelIterator - simplified.
* TwoLevelIterator - changed to the new convention. Also fixed a bug that would make SeekForPrev() sometimes silently ignore errors from first_level_iter_.
* BlockBasedTableIterator - minor changes.
* BlockIter - replaced `SetStatus()` with `Invalidate()` to make sure non-ok BlockIter is always invalid.
* PlainTableIterator - some seeks used to not reset status.
* CuckooTableIterator - tiny code cleanup.
* ManagedIterator - fixed some bugs.
* BaseDeltaIterator - changed to the new convention and fixed a bug.
* BlobDBIterator - seeks used to not reset status.
* KeyConvertingIterator - some small change.
Closes https://github.com/facebook/rocksdb/pull/3810
Differential Revision: D7888019
Pulled By: al13n321
fbshipit-source-id: 4aaf6d3421c545d16722a815b2fa2e7912bc851d
2018-05-17 11:44:14 +02:00
|
|
|
iter.reset();
|
2014-01-24 21:14:08 +01:00
|
|
|
|
2014-09-05 01:18:36 +02:00
|
|
|
const ImmutableCFOptions ioptions1(opt);
|
2018-05-21 23:33:55 +02:00
|
|
|
const MutableCFOptions moptions1(opt);
|
|
|
|
ASSERT_OK(c.Reopen(ioptions1, moptions1));
|
2014-08-26 01:14:30 +02:00
|
|
|
auto table_reader = dynamic_cast<BlockBasedTable*>(c.GetTableReader());
|
2014-01-25 06:10:19 +01:00
|
|
|
for (const std::string& key : keys) {
|
2018-05-26 03:41:31 +02:00
|
|
|
InternalKey ikey(key, kMaxSequenceNumber, kTypeValue);
|
|
|
|
ASSERT_TRUE(table_reader->TEST_KeyInCache(ReadOptions(), ikey.Encode()));
|
2014-01-24 21:14:08 +01:00
|
|
|
}
|
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 19:42:39 +02:00
|
|
|
c.ResetTableReader();
|
2014-06-20 10:23:02 +02:00
|
|
|
|
|
|
|
// rerun with different block cache
|
2016-04-07 22:51:47 +02:00
|
|
|
table_options.block_cache = NewLRUCache(16 * 1024 * 1024, 4);
|
2014-08-25 23:22:05 +02:00
|
|
|
opt.table_factory.reset(NewBlockBasedTableFactory(table_options));
|
2014-09-05 01:18:36 +02:00
|
|
|
const ImmutableCFOptions ioptions2(opt);
|
2018-05-21 23:33:55 +02:00
|
|
|
const MutableCFOptions moptions2(opt);
|
|
|
|
ASSERT_OK(c.Reopen(ioptions2, moptions2));
|
2014-08-26 01:14:30 +02:00
|
|
|
table_reader = dynamic_cast<BlockBasedTable*>(c.GetTableReader());
|
2014-06-20 10:23:02 +02:00
|
|
|
for (const std::string& key : keys) {
|
2018-05-26 03:41:31 +02:00
|
|
|
InternalKey ikey(key, kMaxSequenceNumber, kTypeValue);
|
|
|
|
ASSERT_TRUE(!table_reader->TEST_KeyInCache(ReadOptions(), ikey.Encode()));
|
2014-06-20 10:23:02 +02:00
|
|
|
}
|
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 19:42:39 +02:00
|
|
|
c.ResetTableReader();
|
2014-01-24 21:14:08 +01:00
|
|
|
}
|
|
|
|
|
2018-10-03 02:21:54 +02:00
|
|
|
namespace {
|
2018-10-26 23:27:09 +02:00
|
|
|
class CustomMemoryAllocator : public MemoryAllocator {
|
2018-10-03 02:21:54 +02:00
|
|
|
public:
|
2019-02-14 22:52:47 +01:00
|
|
|
const char* Name() const override { return "CustomMemoryAllocator"; }
|
2018-10-03 02:21:54 +02:00
|
|
|
|
|
|
|
void* Allocate(size_t size) override {
|
|
|
|
++numAllocations;
|
|
|
|
auto ptr = new char[size + 16];
|
2018-10-26 23:27:09 +02:00
|
|
|
memcpy(ptr, "memory_allocator_", 16); // mangle first 16 bytes
|
2018-10-03 02:21:54 +02:00
|
|
|
return reinterpret_cast<void*>(ptr + 16);
|
|
|
|
}
|
|
|
|
void Deallocate(void* p) override {
|
|
|
|
++numDeallocations;
|
|
|
|
char* ptr = reinterpret_cast<char*>(p) - 16;
|
|
|
|
delete[] ptr;
|
|
|
|
}
|
|
|
|
|
|
|
|
std::atomic<int> numAllocations;
|
|
|
|
std::atomic<int> numDeallocations;
|
|
|
|
};
|
|
|
|
} // namespace
|
|
|
|
|
2018-10-26 23:27:09 +02:00
|
|
|
TEST_P(BlockBasedTableTest, MemoryAllocator) {
|
|
|
|
auto custom_memory_allocator = std::make_shared<CustomMemoryAllocator>();
|
2018-10-03 02:21:54 +02:00
|
|
|
{
|
|
|
|
Options opt;
|
2018-11-09 20:17:34 +01:00
|
|
|
std::unique_ptr<InternalKeyComparator> ikc;
|
2018-10-03 02:21:54 +02:00
|
|
|
ikc.reset(new test::PlainInternalKeyComparator(opt.comparator));
|
|
|
|
opt.compression = kNoCompression;
|
|
|
|
BlockBasedTableOptions table_options;
|
|
|
|
table_options.block_size = 1024;
|
|
|
|
LRUCacheOptions lruOptions;
|
2018-11-21 20:28:02 +01:00
|
|
|
lruOptions.memory_allocator = custom_memory_allocator;
|
2018-10-03 02:21:54 +02:00
|
|
|
lruOptions.capacity = 16 * 1024 * 1024;
|
|
|
|
lruOptions.num_shard_bits = 4;
|
|
|
|
table_options.block_cache = NewLRUCache(std::move(lruOptions));
|
|
|
|
opt.table_factory.reset(NewBlockBasedTableFactory(table_options));
|
|
|
|
|
|
|
|
TableConstructor c(BytewiseComparator(),
|
|
|
|
true /* convert_to_internal_key_ */);
|
|
|
|
c.Add("k01", "hello");
|
|
|
|
c.Add("k02", "hello2");
|
|
|
|
c.Add("k03", std::string(10000, 'x'));
|
|
|
|
c.Add("k04", std::string(200000, 'x'));
|
|
|
|
c.Add("k05", std::string(300000, 'x'));
|
|
|
|
c.Add("k06", "hello3");
|
|
|
|
c.Add("k07", std::string(100000, 'x'));
|
|
|
|
std::vector<std::string> keys;
|
|
|
|
stl_wrappers::KVMap kvmap;
|
|
|
|
const ImmutableCFOptions ioptions(opt);
|
|
|
|
const MutableCFOptions moptions(opt);
|
|
|
|
c.Finish(opt, ioptions, moptions, table_options, *ikc, &keys, &kvmap);
|
|
|
|
|
2018-11-09 20:17:34 +01:00
|
|
|
std::unique_ptr<InternalIterator> iter(
|
2018-10-03 02:21:54 +02:00
|
|
|
c.NewIterator(moptions.prefix_extractor.get()));
|
|
|
|
iter->SeekToFirst();
|
|
|
|
while (iter->Valid()) {
|
|
|
|
iter->key();
|
|
|
|
iter->value();
|
|
|
|
iter->Next();
|
|
|
|
}
|
|
|
|
ASSERT_OK(iter->status());
|
|
|
|
}
|
|
|
|
|
|
|
|
// out of scope, block cache should have been deleted, all allocations
|
|
|
|
// deallocated
|
2018-10-26 23:27:09 +02:00
|
|
|
EXPECT_EQ(custom_memory_allocator->numAllocations.load(),
|
|
|
|
custom_memory_allocator->numDeallocations.load());
|
2018-10-03 02:21:54 +02:00
|
|
|
// make sure that allocations actually happened through the cache allocator
|
2018-10-26 23:27:09 +02:00
|
|
|
EXPECT_GT(custom_memory_allocator->numAllocations.load(), 0);
|
2018-10-03 02:21:54 +02:00
|
|
|
}
|
|
|
|
|
2020-02-11 00:42:46 +01:00
|
|
|
// Test the file checksum of block based table
|
|
|
|
TEST_P(BlockBasedTableTest, NoFileChecksum) {
|
|
|
|
Options options;
|
|
|
|
ImmutableCFOptions ioptions(options);
|
|
|
|
MutableCFOptions moptions(options);
|
|
|
|
BlockBasedTableOptions table_options = GetBlockBasedTableOptions();
|
|
|
|
std::unique_ptr<InternalKeyComparator> comparator(
|
|
|
|
new InternalKeyComparator(BytewiseComparator()));
|
|
|
|
int level = 0;
|
|
|
|
std::vector<std::unique_ptr<IntTblPropCollectorFactory>>
|
|
|
|
int_tbl_prop_collector_factories;
|
|
|
|
std::string column_family_name;
|
|
|
|
|
|
|
|
FileChecksumTestHelper f(true);
|
|
|
|
f.CreateWriteableFile();
|
|
|
|
std::unique_ptr<TableBuilder> builder;
|
|
|
|
builder.reset(ioptions.table_factory->NewTableBuilder(
|
|
|
|
TableBuilderOptions(ioptions, moptions, *comparator,
|
|
|
|
&int_tbl_prop_collector_factories,
|
|
|
|
options.compression, options.sample_for_compression,
|
|
|
|
options.compression_opts, false /* skip_filters */,
|
|
|
|
column_family_name, level),
|
|
|
|
TablePropertiesCollectorFactory::Context::kUnknownColumnFamily,
|
|
|
|
f.GetFileWriter()));
|
2020-10-08 20:11:52 +02:00
|
|
|
ASSERT_OK(f.ResetTableBuilder(std::move(builder)));
|
2020-02-11 00:42:46 +01:00
|
|
|
f.AddKVtoKVMap(1000);
|
2020-12-24 01:54:05 +01:00
|
|
|
ASSERT_OK(f.WriteKVAndFlushTable());
|
2020-06-08 06:54:54 +02:00
|
|
|
ASSERT_STREQ(f.GetFileChecksumFuncName(), kUnknownFileChecksumFuncName);
|
|
|
|
ASSERT_STREQ(f.GetFileChecksum().c_str(), kUnknownFileChecksum);
|
2020-02-11 00:42:46 +01:00
|
|
|
}
|
|
|
|
|
2020-05-21 17:10:43 +02:00
|
|
|
TEST_P(BlockBasedTableTest, Crc32cFileChecksum) {
|
2020-03-30 00:57:02 +02:00
|
|
|
FileChecksumGenCrc32cFactory* file_checksum_gen_factory =
|
|
|
|
new FileChecksumGenCrc32cFactory();
|
2020-02-11 00:42:46 +01:00
|
|
|
Options options;
|
2020-03-30 00:57:02 +02:00
|
|
|
options.file_checksum_gen_factory.reset(file_checksum_gen_factory);
|
2020-02-11 00:42:46 +01:00
|
|
|
ImmutableCFOptions ioptions(options);
|
|
|
|
MutableCFOptions moptions(options);
|
|
|
|
BlockBasedTableOptions table_options = GetBlockBasedTableOptions();
|
|
|
|
std::unique_ptr<InternalKeyComparator> comparator(
|
|
|
|
new InternalKeyComparator(BytewiseComparator()));
|
|
|
|
int level = 0;
|
|
|
|
std::vector<std::unique_ptr<IntTblPropCollectorFactory>>
|
|
|
|
int_tbl_prop_collector_factories;
|
|
|
|
std::string column_family_name;
|
|
|
|
|
2020-03-30 00:57:02 +02:00
|
|
|
FileChecksumGenContext gen_context;
|
|
|
|
gen_context.file_name = "db/tmp";
|
2020-05-21 17:10:43 +02:00
|
|
|
std::unique_ptr<FileChecksumGenerator> checksum_crc32c_gen1 =
|
2020-03-30 00:57:02 +02:00
|
|
|
options.file_checksum_gen_factory->CreateFileChecksumGenerator(
|
|
|
|
gen_context);
|
2020-02-11 00:42:46 +01:00
|
|
|
FileChecksumTestHelper f(true);
|
|
|
|
f.CreateWriteableFile();
|
2020-05-21 17:10:43 +02:00
|
|
|
f.SetFileChecksumGenerator(checksum_crc32c_gen1.release());
|
2020-02-11 00:42:46 +01:00
|
|
|
std::unique_ptr<TableBuilder> builder;
|
|
|
|
builder.reset(ioptions.table_factory->NewTableBuilder(
|
|
|
|
TableBuilderOptions(ioptions, moptions, *comparator,
|
|
|
|
&int_tbl_prop_collector_factories,
|
|
|
|
options.compression, options.sample_for_compression,
|
|
|
|
options.compression_opts, false /* skip_filters */,
|
|
|
|
column_family_name, level),
|
|
|
|
TablePropertiesCollectorFactory::Context::kUnknownColumnFamily,
|
|
|
|
f.GetFileWriter()));
|
2020-10-08 20:11:52 +02:00
|
|
|
ASSERT_OK(f.ResetTableBuilder(std::move(builder)));
|
2020-02-11 00:42:46 +01:00
|
|
|
f.AddKVtoKVMap(1000);
|
2020-12-24 01:54:05 +01:00
|
|
|
ASSERT_OK(f.WriteKVAndFlushTable());
|
2020-02-11 00:42:46 +01:00
|
|
|
ASSERT_STREQ(f.GetFileChecksumFuncName(), "FileChecksumCrc32c");
|
2020-03-30 00:57:02 +02:00
|
|
|
|
2020-05-21 17:10:43 +02:00
|
|
|
std::unique_ptr<FileChecksumGenerator> checksum_crc32c_gen2 =
|
2020-03-30 00:57:02 +02:00
|
|
|
options.file_checksum_gen_factory->CreateFileChecksumGenerator(
|
|
|
|
gen_context);
|
2020-02-11 00:42:46 +01:00
|
|
|
std::string checksum;
|
2020-05-21 17:10:43 +02:00
|
|
|
ASSERT_OK(f.CalculateFileChecksum(checksum_crc32c_gen2.get(), &checksum));
|
2020-02-11 00:42:46 +01:00
|
|
|
ASSERT_STREQ(f.GetFileChecksum().c_str(), checksum.c_str());
|
2020-05-21 17:10:43 +02:00
|
|
|
|
|
|
|
// Unit test the generator itself for schema stability
|
|
|
|
std::unique_ptr<FileChecksumGenerator> checksum_crc32c_gen3 =
|
|
|
|
options.file_checksum_gen_factory->CreateFileChecksumGenerator(
|
|
|
|
gen_context);
|
|
|
|
const char data[] = "here is some data";
|
|
|
|
checksum_crc32c_gen3->Update(data, sizeof(data));
|
|
|
|
checksum_crc32c_gen3->Finalize();
|
|
|
|
checksum = checksum_crc32c_gen3->GetChecksum();
|
|
|
|
ASSERT_STREQ(checksum.c_str(), "\345\245\277\110");
|
2020-02-11 00:42:46 +01:00
|
|
|
}
|
|
|
|
|
2015-07-20 20:09:14 +02:00
|
|
|
// Plain table is not supported in ROCKSDB_LITE
|
|
|
|
#ifndef ROCKSDB_LITE
|
2015-03-17 22:08:00 +01:00
|
|
|
TEST_F(PlainTableTest, BasicPlainTableProperties) {
|
2014-07-18 09:08:38 +02:00
|
|
|
PlainTableOptions plain_table_options;
|
|
|
|
plain_table_options.user_key_len = 8;
|
|
|
|
plain_table_options.bloom_bits_per_key = 8;
|
|
|
|
plain_table_options.hash_table_ratio = 0;
|
|
|
|
|
|
|
|
PlainTableFactory factory(plain_table_options);
|
2021-01-05 00:59:52 +01:00
|
|
|
std::unique_ptr<FSWritableFile> sink(new test::StringSink());
|
|
|
|
std::unique_ptr<WritableFileWriter> file_writer(new WritableFileWriter(
|
|
|
|
std::move(sink), "" /* don't care */, FileOptions()));
|
2014-01-27 22:53:22 +01:00
|
|
|
Options options;
|
2014-09-05 01:18:36 +02:00
|
|
|
const ImmutableCFOptions ioptions(options);
|
2018-05-21 23:33:55 +02:00
|
|
|
const MutableCFOptions moptions(options);
|
2014-01-27 22:53:22 +01:00
|
|
|
InternalKeyComparator ikc(options.comparator);
|
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 19:04:30 +02:00
|
|
|
std::vector<std::unique_ptr<IntTblPropCollectorFactory>>
|
|
|
|
int_tbl_prop_collector_factories;
|
2016-04-07 08:10:32 +02:00
|
|
|
std::string column_family_name;
|
2016-09-18 07:30:43 +02:00
|
|
|
int unknown_level = -1;
|
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 19:04:30 +02:00
|
|
|
std::unique_ptr<TableBuilder> builder(factory.NewTableBuilder(
|
2019-03-18 20:07:35 +01:00
|
|
|
TableBuilderOptions(
|
|
|
|
ioptions, moptions, ikc, &int_tbl_prop_collector_factories,
|
|
|
|
kNoCompression, 0 /* sample_for_compression */, CompressionOptions(),
|
|
|
|
false /* skip_filters */, column_family_name, unknown_level),
|
2015-10-09 01:57:35 +02:00
|
|
|
TablePropertiesCollectorFactory::Context::kUnknownColumnFamily,
|
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
2015-07-18 01:16:11 +02:00
|
|
|
file_writer.get()));
|
2014-01-24 21:14:08 +01:00
|
|
|
|
|
|
|
for (char c = 'a'; c <= 'z'; ++c) {
|
2014-01-27 22:53:22 +01:00
|
|
|
std::string key(8, c);
|
|
|
|
key.append("\1 "); // PlainTable expects internal key structure
|
2014-01-24 21:14:08 +01:00
|
|
|
std::string value(28, c + 42);
|
|
|
|
builder->Add(key, value);
|
|
|
|
}
|
|
|
|
ASSERT_OK(builder->Finish());
|
2020-12-24 01:54:05 +01:00
|
|
|
ASSERT_OK(file_writer->Flush());
|
2014-01-24 21:14:08 +01:00
|
|
|
|
2015-08-05 16:33:27 +02:00
|
|
|
test::StringSink* ss =
|
2021-01-05 00:59:52 +01:00
|
|
|
static_cast<test::StringSink*>(file_writer->writable_file());
|
|
|
|
std::unique_ptr<FSRandomAccessFile> source(
|
|
|
|
new test::StringSource(ss->contents(), 72242, true));
|
2018-11-09 20:17:34 +01:00
|
|
|
std::unique_ptr<RandomAccessFileReader> file_reader(
|
2021-01-05 00:59:52 +01:00
|
|
|
new RandomAccessFileReader(std::move(source), "test"));
|
2014-01-24 21:14:08 +01:00
|
|
|
|
2014-02-08 04:26:49 +01:00
|
|
|
TableProperties* props = nullptr;
|
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
2015-07-18 01:16:11 +02:00
|
|
|
auto s = ReadTableProperties(file_reader.get(), ss->contents().size(),
|
2016-07-19 18:44:03 +02:00
|
|
|
kPlainTableMagicNumber, ioptions,
|
2018-06-16 04:24:21 +02:00
|
|
|
&props, true /* compression_type_missing */);
|
2014-02-08 07:43:58 +01:00
|
|
|
std::unique_ptr<TableProperties> props_guard(props);
|
2014-01-24 21:14:08 +01:00
|
|
|
ASSERT_OK(s);
|
|
|
|
|
2014-02-08 04:26:49 +01:00
|
|
|
ASSERT_EQ(0ul, props->index_size);
|
|
|
|
ASSERT_EQ(0ul, props->filter_size);
|
|
|
|
ASSERT_EQ(16ul * 26, props->raw_key_size);
|
|
|
|
ASSERT_EQ(28ul * 26, props->raw_value_size);
|
|
|
|
ASSERT_EQ(26ul, props->num_entries);
|
|
|
|
ASSERT_EQ(1ul, props->num_data_blocks);
|
2014-01-24 21:14:08 +01:00
|
|
|
}
|
2020-02-11 00:42:46 +01:00
|
|
|
|
|
|
|
TEST_F(PlainTableTest, NoFileChecksum) {
|
|
|
|
PlainTableOptions plain_table_options;
|
|
|
|
plain_table_options.user_key_len = 20;
|
|
|
|
plain_table_options.bloom_bits_per_key = 8;
|
|
|
|
plain_table_options.hash_table_ratio = 0;
|
|
|
|
PlainTableFactory factory(plain_table_options);
|
|
|
|
|
|
|
|
Options options;
|
|
|
|
const ImmutableCFOptions ioptions(options);
|
|
|
|
const MutableCFOptions moptions(options);
|
|
|
|
InternalKeyComparator ikc(options.comparator);
|
|
|
|
std::vector<std::unique_ptr<IntTblPropCollectorFactory>>
|
|
|
|
int_tbl_prop_collector_factories;
|
|
|
|
std::string column_family_name;
|
|
|
|
int unknown_level = -1;
|
|
|
|
FileChecksumTestHelper f(true);
|
|
|
|
f.CreateWriteableFile();
|
|
|
|
|
|
|
|
std::unique_ptr<TableBuilder> builder(factory.NewTableBuilder(
|
|
|
|
TableBuilderOptions(
|
|
|
|
ioptions, moptions, ikc, &int_tbl_prop_collector_factories,
|
|
|
|
kNoCompression, 0 /* sample_for_compression */, CompressionOptions(),
|
|
|
|
false /* skip_filters */, column_family_name, unknown_level),
|
|
|
|
TablePropertiesCollectorFactory::Context::kUnknownColumnFamily,
|
|
|
|
f.GetFileWriter()));
|
2020-10-08 20:11:52 +02:00
|
|
|
ASSERT_OK(f.ResetTableBuilder(std::move(builder)));
|
2020-02-11 00:42:46 +01:00
|
|
|
f.AddKVtoKVMap(1000);
|
2020-12-24 01:54:05 +01:00
|
|
|
ASSERT_OK(f.WriteKVAndFlushTable());
|
2020-06-08 06:54:54 +02:00
|
|
|
ASSERT_STREQ(f.GetFileChecksumFuncName(), kUnknownFileChecksumFuncName);
|
|
|
|
EXPECT_EQ(f.GetFileChecksum(), kUnknownFileChecksum);
|
2020-02-11 00:42:46 +01:00
|
|
|
}
|
|
|
|
|
2020-05-21 17:10:43 +02:00
|
|
|
TEST_F(PlainTableTest, Crc32cFileChecksum) {
|
2020-02-11 00:42:46 +01:00
|
|
|
PlainTableOptions plain_table_options;
|
|
|
|
plain_table_options.user_key_len = 20;
|
|
|
|
plain_table_options.bloom_bits_per_key = 8;
|
|
|
|
plain_table_options.hash_table_ratio = 0;
|
|
|
|
PlainTableFactory factory(plain_table_options);
|
|
|
|
|
2020-03-30 00:57:02 +02:00
|
|
|
FileChecksumGenCrc32cFactory* file_checksum_gen_factory =
|
|
|
|
new FileChecksumGenCrc32cFactory();
|
2020-02-11 00:42:46 +01:00
|
|
|
Options options;
|
2020-03-30 00:57:02 +02:00
|
|
|
options.file_checksum_gen_factory.reset(file_checksum_gen_factory);
|
2020-02-11 00:42:46 +01:00
|
|
|
const ImmutableCFOptions ioptions(options);
|
|
|
|
const MutableCFOptions moptions(options);
|
|
|
|
InternalKeyComparator ikc(options.comparator);
|
|
|
|
std::vector<std::unique_ptr<IntTblPropCollectorFactory>>
|
|
|
|
int_tbl_prop_collector_factories;
|
|
|
|
std::string column_family_name;
|
|
|
|
int unknown_level = -1;
|
2020-03-30 00:57:02 +02:00
|
|
|
|
|
|
|
FileChecksumGenContext gen_context;
|
|
|
|
gen_context.file_name = "db/tmp";
|
2020-05-21 17:10:43 +02:00
|
|
|
std::unique_ptr<FileChecksumGenerator> checksum_crc32c_gen1 =
|
2020-03-30 00:57:02 +02:00
|
|
|
options.file_checksum_gen_factory->CreateFileChecksumGenerator(
|
|
|
|
gen_context);
|
2020-02-11 00:42:46 +01:00
|
|
|
FileChecksumTestHelper f(true);
|
|
|
|
f.CreateWriteableFile();
|
2020-05-21 17:10:43 +02:00
|
|
|
f.SetFileChecksumGenerator(checksum_crc32c_gen1.release());
|
2020-02-11 00:42:46 +01:00
|
|
|
|
|
|
|
std::unique_ptr<TableBuilder> builder(factory.NewTableBuilder(
|
|
|
|
TableBuilderOptions(
|
|
|
|
ioptions, moptions, ikc, &int_tbl_prop_collector_factories,
|
|
|
|
kNoCompression, 0 /* sample_for_compression */, CompressionOptions(),
|
|
|
|
false /* skip_filters */, column_family_name, unknown_level),
|
|
|
|
TablePropertiesCollectorFactory::Context::kUnknownColumnFamily,
|
|
|
|
f.GetFileWriter()));
|
2020-10-08 20:11:52 +02:00
|
|
|
ASSERT_OK(f.ResetTableBuilder(std::move(builder)));
|
2020-02-11 00:42:46 +01:00
|
|
|
f.AddKVtoKVMap(1000);
|
2020-12-24 01:54:05 +01:00
|
|
|
ASSERT_OK(f.WriteKVAndFlushTable());
|
2020-02-11 00:42:46 +01:00
|
|
|
ASSERT_STREQ(f.GetFileChecksumFuncName(), "FileChecksumCrc32c");
|
2020-03-30 00:57:02 +02:00
|
|
|
|
2020-05-21 17:10:43 +02:00
|
|
|
std::unique_ptr<FileChecksumGenerator> checksum_crc32c_gen2 =
|
2020-03-30 00:57:02 +02:00
|
|
|
options.file_checksum_gen_factory->CreateFileChecksumGenerator(
|
|
|
|
gen_context);
|
2020-02-11 00:42:46 +01:00
|
|
|
std::string checksum;
|
2020-05-21 17:10:43 +02:00
|
|
|
ASSERT_OK(f.CalculateFileChecksum(checksum_crc32c_gen2.get(), &checksum));
|
2020-02-11 00:42:46 +01:00
|
|
|
EXPECT_STREQ(f.GetFileChecksum().c_str(), checksum.c_str());
|
|
|
|
}
|
|
|
|
|
2015-07-20 20:09:14 +02:00
|
|
|
#endif // !ROCKSDB_LITE
|
2014-01-24 21:14:08 +01:00
|
|
|
|
2015-03-17 22:08:00 +01:00
|
|
|
TEST_F(GeneralTableTest, ApproximateOffsetOfPlain) {
|
2016-08-20 00:10:31 +02:00
|
|
|
TableConstructor c(BytewiseComparator(), true /* convert_to_internal_key_ */);
|
2011-03-18 23:37:00 +01:00
|
|
|
c.Add("k01", "hello");
|
|
|
|
c.Add("k02", "hello2");
|
|
|
|
c.Add("k03", std::string(10000, 'x'));
|
|
|
|
c.Add("k04", std::string(200000, 'x'));
|
|
|
|
c.Add("k05", std::string(300000, 'x'));
|
|
|
|
c.Add("k06", "hello3");
|
|
|
|
c.Add("k07", std::string(100000, 'x'));
|
|
|
|
std::vector<std::string> keys;
|
2015-09-02 22:58:22 +02:00
|
|
|
stl_wrappers::KVMap kvmap;
|
2013-11-20 07:00:48 +01:00
|
|
|
Options options;
|
2020-10-19 20:37:05 +02:00
|
|
|
options.db_host_id = "";
|
2014-01-27 22:53:22 +01:00
|
|
|
test::PlainInternalKeyComparator internal_comparator(options.comparator);
|
2011-03-18 23:37:00 +01:00
|
|
|
options.compression = kNoCompression;
|
2014-08-25 23:22:05 +02:00
|
|
|
BlockBasedTableOptions table_options;
|
|
|
|
table_options.block_size = 1024;
|
2014-09-05 01:18:36 +02:00
|
|
|
const ImmutableCFOptions ioptions(options);
|
2018-05-21 23:33:55 +02:00
|
|
|
const MutableCFOptions moptions(options);
|
|
|
|
c.Finish(options, ioptions, moptions, table_options, internal_comparator,
|
2014-09-05 01:18:36 +02:00
|
|
|
&keys, &kvmap);
|
2011-03-18 23:37:00 +01:00
|
|
|
|
|
|
|
ASSERT_TRUE(Between(c.ApproximateOffsetOf("abc"), 0, 0));
|
|
|
|
ASSERT_TRUE(Between(c.ApproximateOffsetOf("k01"), 0, 0));
|
|
|
|
ASSERT_TRUE(Between(c.ApproximateOffsetOf("k01a"), 0, 0));
|
|
|
|
ASSERT_TRUE(Between(c.ApproximateOffsetOf("k02"), 0, 0));
|
|
|
|
ASSERT_TRUE(Between(c.ApproximateOffsetOf("k03"), 0, 0));
|
|
|
|
ASSERT_TRUE(Between(c.ApproximateOffsetOf("k04"), 10000, 11000));
|
2016-08-20 00:10:31 +02:00
|
|
|
// k04 and k05 will be in two consecutive blocks, the index is
|
|
|
|
// an arbitrary slice between k04 and k05, either before or after k04a
|
|
|
|
ASSERT_TRUE(Between(c.ApproximateOffsetOf("k04a"), 10000, 211000));
|
2011-03-18 23:37:00 +01:00
|
|
|
ASSERT_TRUE(Between(c.ApproximateOffsetOf("k05"), 210000, 211000));
|
|
|
|
ASSERT_TRUE(Between(c.ApproximateOffsetOf("k06"), 510000, 511000));
|
|
|
|
ASSERT_TRUE(Between(c.ApproximateOffsetOf("k07"), 510000, 511000));
|
2012-04-17 17:36:46 +02:00
|
|
|
ASSERT_TRUE(Between(c.ApproximateOffsetOf("xyz"), 610000, 612000));
|
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 19:42:39 +02:00
|
|
|
c.ResetTableReader();
|
2011-03-18 23:37:00 +01:00
|
|
|
}
|
|
|
|
|
2014-01-24 20:09:04 +01:00
|
|
|
static void DoCompressionTest(CompressionType comp) {
|
2011-03-18 23:37:00 +01:00
|
|
|
Random rnd(301);
|
2016-08-20 00:10:31 +02:00
|
|
|
TableConstructor c(BytewiseComparator(), true /* convert_to_internal_key_ */);
|
2011-03-18 23:37:00 +01:00
|
|
|
std::string tmp;
|
|
|
|
c.Add("k01", "hello");
|
|
|
|
c.Add("k02", test::CompressibleString(&rnd, 0.25, 10000, &tmp));
|
|
|
|
c.Add("k03", "hello3");
|
|
|
|
c.Add("k04", test::CompressibleString(&rnd, 0.25, 10000, &tmp));
|
|
|
|
std::vector<std::string> keys;
|
2015-09-02 22:58:22 +02:00
|
|
|
stl_wrappers::KVMap kvmap;
|
2013-11-20 07:00:48 +01:00
|
|
|
Options options;
|
2014-01-27 22:53:22 +01:00
|
|
|
test::PlainInternalKeyComparator ikc(options.comparator);
|
2012-06-28 08:41:33 +02:00
|
|
|
options.compression = comp;
|
2014-08-25 23:22:05 +02:00
|
|
|
BlockBasedTableOptions table_options;
|
|
|
|
table_options.block_size = 1024;
|
2014-09-05 01:18:36 +02:00
|
|
|
const ImmutableCFOptions ioptions(options);
|
2018-05-21 23:33:55 +02:00
|
|
|
const MutableCFOptions moptions(options);
|
|
|
|
c.Finish(options, ioptions, moptions, table_options, ikc, &keys, &kvmap);
|
2011-03-18 23:37:00 +01:00
|
|
|
|
For ApproximateSizes, pro-rate table metadata size over data blocks (#6784)
Summary:
The implementation of GetApproximateSizes was inconsistent in
its treatment of the size of non-data blocks of SST files, sometimes
including and sometimes now. This was at its worst with large portion
of table file used by filters and querying a small range that crossed
a table boundary: the size estimate would include large filter size.
It's conceivable that someone might want only to know the size in terms
of data blocks, but I believe that's unlikely enough to ignore for now.
Similarly, there's no evidence the internal function AppoximateOffsetOf
is used for anything other than a one-sided ApproximateSize, so I intend
to refactor to remove redundancy in a follow-up commit.
So to fix this, GetApproximateSizes (and implementation details
ApproximateSize and ApproximateOffsetOf) now consistently include in
their returned sizes a portion of table file metadata (incl filters
and indexes) based on the size portion of the data blocks in range. In
other words, if a key range covers data blocks that are X% by size of all
the table's data blocks, returned approximate size is X% of the total
file size. It would technically be more accurate to attribute metadata
based on number of keys, but that's not computationally efficient with
data available and rarely a meaningful difference.
Also includes miscellaneous comment improvements / clarifications.
Also included is a new approximatesizerandom benchmark for db_bench.
No significant performance difference seen with this change, whether ~700 ops/sec with cache_index_and_filter_blocks and small cache or ~150k ops/sec without cache_index_and_filter_blocks.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/6784
Test Plan:
Test added to DBTest.ApproximateSizesFilesWithErrorMargin.
Old code running new test...
[ RUN ] DBTest.ApproximateSizesFilesWithErrorMargin
db/db_test.cc:1562: Failure
Expected: (size) <= (11 * 100), actual: 9478 vs 1100
Other tests updated to reflect consistent accounting of metadata.
Reviewed By: siying
Differential Revision: D21334706
Pulled By: pdillinger
fbshipit-source-id: 6f86870e45213334fedbe9c73b4ebb1d8d611185
2020-06-02 21:27:59 +02:00
|
|
|
ASSERT_TRUE(Between(c.ApproximateOffsetOf("abc"), 0, 0));
|
|
|
|
ASSERT_TRUE(Between(c.ApproximateOffsetOf("k01"), 0, 0));
|
|
|
|
ASSERT_TRUE(Between(c.ApproximateOffsetOf("k02"), 0, 0));
|
|
|
|
ASSERT_TRUE(Between(c.ApproximateOffsetOf("k03"), 2000, 3500));
|
|
|
|
ASSERT_TRUE(Between(c.ApproximateOffsetOf("k04"), 2000, 3500));
|
|
|
|
ASSERT_TRUE(Between(c.ApproximateOffsetOf("xyz"), 4000, 7000));
|
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 19:42:39 +02:00
|
|
|
c.ResetTableReader();
|
2011-03-18 23:37:00 +01:00
|
|
|
}
|
|
|
|
|
2015-03-17 22:08:00 +01:00
|
|
|
TEST_F(GeneralTableTest, ApproximateOffsetOfCompressed) {
|
2014-02-11 02:02:02 +01:00
|
|
|
std::vector<CompressionType> compression_state;
|
2015-04-06 21:50:44 +02:00
|
|
|
if (!Snappy_Supported()) {
|
2012-06-28 08:41:33 +02:00
|
|
|
fprintf(stderr, "skipping snappy compression tests\n");
|
|
|
|
} else {
|
2014-02-11 02:02:02 +01:00
|
|
|
compression_state.push_back(kSnappyCompression);
|
2012-06-28 08:41:33 +02:00
|
|
|
}
|
|
|
|
|
2015-04-06 21:50:44 +02:00
|
|
|
if (!Zlib_Supported()) {
|
2012-06-28 08:41:33 +02:00
|
|
|
fprintf(stderr, "skipping zlib compression tests\n");
|
|
|
|
} else {
|
2014-02-11 02:02:02 +01:00
|
|
|
compression_state.push_back(kZlibCompression);
|
2012-06-28 08:41:33 +02:00
|
|
|
}
|
|
|
|
|
2014-02-11 02:02:02 +01:00
|
|
|
// TODO(kailiu) DoCompressionTest() doesn't work with BZip2.
|
|
|
|
/*
|
2015-04-06 21:50:44 +02:00
|
|
|
if (!BZip2_Supported()) {
|
2014-02-08 03:12:30 +01:00
|
|
|
fprintf(stderr, "skipping bzip2 compression tests\n");
|
|
|
|
} else {
|
2014-02-11 02:02:02 +01:00
|
|
|
compression_state.push_back(kBZip2Compression);
|
2014-02-08 03:12:30 +01:00
|
|
|
}
|
2014-02-11 02:02:02 +01:00
|
|
|
*/
|
2014-02-08 03:12:30 +01:00
|
|
|
|
2015-04-06 21:50:44 +02:00
|
|
|
if (!LZ4_Supported()) {
|
|
|
|
fprintf(stderr, "skipping lz4 and lz4hc compression tests\n");
|
2014-02-08 03:12:30 +01:00
|
|
|
} else {
|
2014-02-11 02:02:02 +01:00
|
|
|
compression_state.push_back(kLZ4Compression);
|
|
|
|
compression_state.push_back(kLZ4HCCompression);
|
2012-06-28 08:41:33 +02:00
|
|
|
}
|
|
|
|
|
2016-04-20 07:54:24 +02:00
|
|
|
if (!XPRESS_Supported()) {
|
|
|
|
fprintf(stderr, "skipping xpress and xpress compression tests\n");
|
|
|
|
}
|
|
|
|
else {
|
|
|
|
compression_state.push_back(kXpressCompression);
|
|
|
|
}
|
|
|
|
|
2014-02-11 02:02:02 +01:00
|
|
|
for (auto state : compression_state) {
|
|
|
|
DoCompressionTest(state);
|
2012-06-28 08:41:33 +02:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2018-11-06 23:19:51 +01:00
|
|
|
#ifndef ROCKSDB_VALGRIND_RUN
|
2020-06-13 02:29:22 +02:00
|
|
|
TEST_P(ParameterizedHarnessTest, RandomizedHarnessTest) {
|
|
|
|
Random rnd(test::RandomSeed() + 5);
|
|
|
|
for (int num_entries = 0; num_entries < 2000;
|
|
|
|
num_entries += (num_entries < 50 ? 1 : 200)) {
|
|
|
|
for (int e = 0; e < num_entries; e++) {
|
|
|
|
Add(test::RandomKey(&rnd, rnd.Skewed(4)),
|
2020-07-09 23:33:42 +02:00
|
|
|
rnd.RandomString(rnd.Skewed(5)));
|
2020-06-13 02:29:22 +02:00
|
|
|
}
|
|
|
|
Test(&rnd);
|
|
|
|
}
|
2013-11-10 10:17:32 +01:00
|
|
|
}
|
|
|
|
|
2015-10-13 19:32:05 +02:00
|
|
|
#ifndef ROCKSDB_LITE
|
2020-06-13 02:29:22 +02:00
|
|
|
TEST_F(DBHarnessTest, RandomizedLongDB) {
|
2013-11-10 10:17:32 +01:00
|
|
|
Random rnd(test::RandomSeed());
|
|
|
|
int num_entries = 100000;
|
|
|
|
for (int e = 0; e < num_entries; e++) {
|
|
|
|
std::string v;
|
2020-07-09 23:33:42 +02:00
|
|
|
Add(test::RandomKey(&rnd, rnd.Skewed(4)), rnd.RandomString(rnd.Skewed(5)));
|
2013-11-10 10:17:32 +01:00
|
|
|
}
|
|
|
|
Test(&rnd);
|
|
|
|
|
|
|
|
// We must have created enough data to force merging
|
|
|
|
int files = 0;
|
|
|
|
for (int level = 0; level < db()->NumberLevels(); level++) {
|
|
|
|
std::string value;
|
|
|
|
char name[100];
|
|
|
|
snprintf(name, sizeof(name), "rocksdb.num-files-at-level%d", level);
|
|
|
|
ASSERT_TRUE(db()->GetProperty(name, &value));
|
|
|
|
files += atoi(value.c_str());
|
|
|
|
}
|
|
|
|
ASSERT_GT(files, 0);
|
|
|
|
}
|
2015-10-13 19:32:05 +02:00
|
|
|
#endif // ROCKSDB_LITE
|
2018-11-06 23:19:51 +01:00
|
|
|
#endif // ROCKSDB_VALGRIND_RUN
|
2013-11-10 10:17:32 +01:00
|
|
|
|
2020-10-14 21:43:14 +02:00
|
|
|
class MemTableTest : public testing::Test {
|
|
|
|
public:
|
|
|
|
MemTableTest() {
|
|
|
|
InternalKeyComparator cmp(BytewiseComparator());
|
|
|
|
auto table_factory = std::make_shared<SkipListFactory>();
|
|
|
|
options_.memtable_factory = table_factory;
|
|
|
|
ImmutableCFOptions ioptions(options_);
|
|
|
|
wb_ = new WriteBufferManager(options_.db_write_buffer_size);
|
|
|
|
memtable_ = new MemTable(cmp, ioptions, MutableCFOptions(options_), wb_,
|
|
|
|
kMaxSequenceNumber, 0 /* column_family_id */);
|
|
|
|
memtable_->Ref();
|
|
|
|
}
|
|
|
|
|
|
|
|
~MemTableTest() {
|
|
|
|
delete memtable_->Unref();
|
|
|
|
delete wb_;
|
|
|
|
}
|
|
|
|
|
|
|
|
MemTable* GetMemTable() { return memtable_; }
|
|
|
|
|
|
|
|
private:
|
|
|
|
MemTable* memtable_;
|
|
|
|
Options options_;
|
|
|
|
WriteBufferManager* wb_;
|
|
|
|
};
|
2013-11-10 10:17:32 +01:00
|
|
|
|
2015-03-17 22:08:00 +01:00
|
|
|
TEST_F(MemTableTest, Simple) {
|
2013-11-10 10:17:32 +01:00
|
|
|
WriteBatch batch;
|
|
|
|
WriteBatchInternal::SetSequence(&batch, 100);
|
2020-10-08 20:11:52 +02:00
|
|
|
ASSERT_OK(batch.Put(std::string("k1"), std::string("v1")));
|
|
|
|
ASSERT_OK(batch.Put(std::string("k2"), std::string("v2")));
|
|
|
|
ASSERT_OK(batch.Put(std::string("k3"), std::string("v3")));
|
|
|
|
ASSERT_OK(batch.Put(std::string("largekey"), std::string("vlarge")));
|
|
|
|
ASSERT_OK(batch.DeleteRange(std::string("chi"), std::string("xigua")));
|
|
|
|
ASSERT_OK(batch.DeleteRange(std::string("begin"), std::string("end")));
|
2020-10-14 21:43:14 +02:00
|
|
|
ColumnFamilyMemTablesDefault cf_mems_default(GetMemTable());
|
support for concurrent adds to memtable
Summary:
This diff adds support for concurrent adds to the skiplist memtable
implementations. Memory allocation is made thread-safe by the addition of
a spinlock, with small per-core buffers to avoid contention. Concurrent
memtable writes are made via an additional method and don't impose a
performance overhead on the non-concurrent case, so parallelism can be
selected on a per-batch basis.
Write thread synchronization is an increasing bottleneck for higher levels
of concurrency, so this diff adds --enable_write_thread_adaptive_yield
(default off). This feature causes threads joining a write batch
group to spin for a short time (default 100 usec) using sched_yield,
rather than going to sleep on a mutex. If the timing of the yield calls
indicates that another thread has actually run during the yield then
spinning is avoided. This option improves performance for concurrent
situations even without parallel adds, although it has the potential to
increase CPU usage (and the heuristic adaptation is not yet mature).
Parallel writes are not currently compatible with
inplace updates, update callbacks, or delete filtering.
Enable it with --allow_concurrent_memtable_write (and
--enable_write_thread_adaptive_yield). Parallel memtable writes
are performance neutral when there is no actual parallelism, and in
my experiments (SSD server-class Linux and varying contention and key
sizes for fillrandom) they are always a performance win when there is
more than one thread.
Statistics are updated earlier in the write path, dropping the number
of DB mutex acquisitions from 2 to 1 for almost all cases.
This diff was motivated and inspired by Yahoo's cLSM work. It is more
conservative than cLSM: RocksDB's write batch group leader role is
preserved (along with all of the existing flush and write throttling
logic) and concurrent writers are blocked until all memtable insertions
have completed and the sequence number has been advanced, to preserve
linearizability.
My test config is "db_bench -benchmarks=fillrandom -threads=$T
-batch_size=1 -memtablerep=skip_list -value_size=100 --num=1000000/$T
-level0_slowdown_writes_trigger=9999 -level0_stop_writes_trigger=9999
-disable_auto_compactions --max_write_buffer_number=8
-max_background_flushes=8 --disable_wal --write_buffer_size=160000000
--block_size=16384 --allow_concurrent_memtable_write" on a two-socket
Xeon E5-2660 @ 2.2Ghz with lots of memory and an SSD hard drive. With 1
thread I get ~440Kops/sec. Peak performance for 1 socket (numactl
-N1) is slightly more than 1Mops/sec, at 16 threads. Peak performance
across both sockets happens at 30 threads, and is ~900Kops/sec, although
with fewer threads there is less performance loss when the system has
background work.
Test Plan:
1. concurrent stress tests for InlineSkipList and DynamicBloom
2. make clean; make check
3. make clean; DISABLE_JEMALLOC=1 make valgrind_check; valgrind db_bench
4. make clean; COMPILE_WITH_TSAN=1 make all check; db_bench
5. make clean; COMPILE_WITH_ASAN=1 make all check; db_bench
6. make clean; OPT=-DROCKSDB_LITE make check
7. verify no perf regressions when disabled
Reviewers: igor, sdong
Reviewed By: sdong
Subscribers: MarkCallaghan, IslamAbdelRahman, anthony, yhchiang, rven, sdong, guyg8, kradhakrishnan, dhruba
Differential Revision: https://reviews.facebook.net/D50589
2015-08-15 01:59:07 +02:00
|
|
|
ASSERT_TRUE(
|
Refactor trimming logic for immutable memtables (#5022)
Summary:
MyRocks currently sets `max_write_buffer_number_to_maintain` in order to maintain enough history for transaction conflict checking. The effectiveness of this approach depends on the size of memtables. When memtables are small, it may not keep enough history; when memtables are large, this may consume too much memory.
We are proposing a new way to configure memtable list history: by limiting the memory usage of immutable memtables. The new option is `max_write_buffer_size_to_maintain` and it will take precedence over the old `max_write_buffer_number_to_maintain` if they are both set to non-zero values. The new option accounts for the total memory usage of flushed immutable memtables and mutable memtable. When the total usage exceeds the limit, RocksDB may start dropping immutable memtables (which is also called trimming history), starting from the oldest one.
The semantics of the old option actually works both as an upper bound and lower bound. History trimming will start if number of immutable memtables exceeds the limit, but it will never go below (limit-1) due to history trimming.
In order the mimic the behavior with the new option, history trimming will stop if dropping the next immutable memtable causes the total memory usage go below the size limit. For example, assuming the size limit is set to 64MB, and there are 3 immutable memtables with sizes of 20, 30, 30. Although the total memory usage is 80MB > 64MB, dropping the oldest memtable will reduce the memory usage to 60MB < 64MB, so in this case no memtable will be dropped.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/5022
Differential Revision: D14394062
Pulled By: miasantreble
fbshipit-source-id: 60457a509c6af89d0993f988c9b5c2aa9e45f5c5
2019-08-23 22:54:09 +02:00
|
|
|
WriteBatchInternal::InsertInto(&batch, &cf_mems_default, nullptr, nullptr)
|
|
|
|
.ok());
|
2013-11-10 10:17:32 +01:00
|
|
|
|
2016-09-12 23:14:40 +02:00
|
|
|
for (int i = 0; i < 2; ++i) {
|
|
|
|
Arena arena;
|
2016-11-19 23:14:35 +01:00
|
|
|
ScopedArenaIterator arena_iter_guard;
|
|
|
|
std::unique_ptr<InternalIterator> iter_guard;
|
|
|
|
InternalIterator* iter;
|
|
|
|
if (i == 0) {
|
2020-10-14 21:43:14 +02:00
|
|
|
iter = GetMemTable()->NewIterator(ReadOptions(), &arena);
|
2016-11-19 23:14:35 +01:00
|
|
|
arena_iter_guard.set(iter);
|
|
|
|
} else {
|
2020-10-14 21:43:14 +02:00
|
|
|
iter = GetMemTable()->NewRangeTombstoneIterator(
|
2018-11-29 00:26:56 +01:00
|
|
|
ReadOptions(), kMaxSequenceNumber /* read_seq */);
|
2016-11-19 23:14:35 +01:00
|
|
|
iter_guard.reset(iter);
|
|
|
|
}
|
2016-11-21 21:07:09 +01:00
|
|
|
if (iter == nullptr) {
|
|
|
|
continue;
|
|
|
|
}
|
2016-09-12 23:14:40 +02:00
|
|
|
iter->SeekToFirst();
|
|
|
|
while (iter->Valid()) {
|
|
|
|
fprintf(stderr, "key: '%s' -> '%s'\n", iter->key().ToString().c_str(),
|
|
|
|
iter->value().ToString().c_str());
|
|
|
|
iter->Next();
|
|
|
|
}
|
2013-11-10 10:17:32 +01:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2013-12-06 01:51:26 +01:00
|
|
|
// Test the empty key
|
2020-06-13 02:29:22 +02:00
|
|
|
TEST_P(ParameterizedHarnessTest, SimpleEmptyKey) {
|
|
|
|
Random rnd(test::RandomSeed() + 1);
|
|
|
|
Add("", "v");
|
|
|
|
Test(&rnd);
|
2013-12-06 01:51:26 +01:00
|
|
|
}
|
|
|
|
|
2020-06-13 02:29:22 +02:00
|
|
|
TEST_P(ParameterizedHarnessTest, SimpleSingle) {
|
|
|
|
Random rnd(test::RandomSeed() + 2);
|
|
|
|
Add("abc", "v");
|
|
|
|
Test(&rnd);
|
2013-12-06 01:51:26 +01:00
|
|
|
}
|
|
|
|
|
2020-06-13 02:29:22 +02:00
|
|
|
TEST_P(ParameterizedHarnessTest, SimpleMulti) {
|
|
|
|
Random rnd(test::RandomSeed() + 3);
|
|
|
|
Add("abc", "v");
|
|
|
|
Add("abcd", "v");
|
|
|
|
Add("ac", "v2");
|
|
|
|
Test(&rnd);
|
2013-12-06 01:51:26 +01:00
|
|
|
}
|
|
|
|
|
2020-06-13 02:29:22 +02:00
|
|
|
TEST_P(ParameterizedHarnessTest, SimpleSpecialKey) {
|
|
|
|
Random rnd(test::RandomSeed() + 4);
|
|
|
|
Add("\xff\xff", "v3");
|
|
|
|
Test(&rnd);
|
2013-12-06 01:51:26 +01:00
|
|
|
}
|
2013-11-10 10:17:32 +01:00
|
|
|
|
2020-06-13 02:29:22 +02:00
|
|
|
TEST(TableTest, FooterTests) {
|
2014-05-01 20:09:32 +02:00
|
|
|
{
|
|
|
|
// upconvert legacy block based
|
|
|
|
std::string encoded;
|
2015-01-13 23:33:04 +01:00
|
|
|
Footer footer(kLegacyBlockBasedTableMagicNumber, 0);
|
2014-05-01 20:09:32 +02:00
|
|
|
BlockHandle meta_index(10, 5), index(20, 15);
|
|
|
|
footer.set_metaindex_handle(meta_index);
|
|
|
|
footer.set_index_handle(index);
|
|
|
|
footer.EncodeTo(&encoded);
|
|
|
|
Footer decoded_footer;
|
|
|
|
Slice encoded_slice(encoded);
|
2020-10-08 20:11:52 +02:00
|
|
|
ASSERT_OK(decoded_footer.DecodeFrom(&encoded_slice));
|
2014-05-01 20:09:32 +02:00
|
|
|
ASSERT_EQ(decoded_footer.table_magic_number(), kBlockBasedTableMagicNumber);
|
|
|
|
ASSERT_EQ(decoded_footer.checksum(), kCRC32c);
|
|
|
|
ASSERT_EQ(decoded_footer.metaindex_handle().offset(), meta_index.offset());
|
|
|
|
ASSERT_EQ(decoded_footer.metaindex_handle().size(), meta_index.size());
|
|
|
|
ASSERT_EQ(decoded_footer.index_handle().offset(), index.offset());
|
|
|
|
ASSERT_EQ(decoded_footer.index_handle().size(), index.size());
|
2015-01-13 23:33:04 +01:00
|
|
|
ASSERT_EQ(decoded_footer.version(), 0U);
|
2014-05-01 20:09:32 +02:00
|
|
|
}
|
|
|
|
{
|
|
|
|
// xxhash block based
|
|
|
|
std::string encoded;
|
2015-01-13 23:33:04 +01:00
|
|
|
Footer footer(kBlockBasedTableMagicNumber, 1);
|
2014-05-01 20:09:32 +02:00
|
|
|
BlockHandle meta_index(10, 5), index(20, 15);
|
|
|
|
footer.set_metaindex_handle(meta_index);
|
|
|
|
footer.set_index_handle(index);
|
|
|
|
footer.set_checksum(kxxHash);
|
|
|
|
footer.EncodeTo(&encoded);
|
|
|
|
Footer decoded_footer;
|
|
|
|
Slice encoded_slice(encoded);
|
2020-10-08 20:11:52 +02:00
|
|
|
ASSERT_OK(decoded_footer.DecodeFrom(&encoded_slice));
|
2014-05-01 20:09:32 +02:00
|
|
|
ASSERT_EQ(decoded_footer.table_magic_number(), kBlockBasedTableMagicNumber);
|
|
|
|
ASSERT_EQ(decoded_footer.checksum(), kxxHash);
|
|
|
|
ASSERT_EQ(decoded_footer.metaindex_handle().offset(), meta_index.offset());
|
|
|
|
ASSERT_EQ(decoded_footer.metaindex_handle().size(), meta_index.size());
|
|
|
|
ASSERT_EQ(decoded_footer.index_handle().offset(), index.offset());
|
|
|
|
ASSERT_EQ(decoded_footer.index_handle().size(), index.size());
|
2015-01-13 23:33:04 +01:00
|
|
|
ASSERT_EQ(decoded_footer.version(), 1U);
|
2014-05-01 20:09:32 +02:00
|
|
|
}
|
2018-11-01 23:39:40 +01:00
|
|
|
{
|
|
|
|
// xxhash64 block based
|
|
|
|
std::string encoded;
|
|
|
|
Footer footer(kBlockBasedTableMagicNumber, 1);
|
|
|
|
BlockHandle meta_index(10, 5), index(20, 15);
|
|
|
|
footer.set_metaindex_handle(meta_index);
|
|
|
|
footer.set_index_handle(index);
|
|
|
|
footer.set_checksum(kxxHash64);
|
|
|
|
footer.EncodeTo(&encoded);
|
|
|
|
Footer decoded_footer;
|
|
|
|
Slice encoded_slice(encoded);
|
2020-10-08 20:11:52 +02:00
|
|
|
ASSERT_OK(decoded_footer.DecodeFrom(&encoded_slice));
|
2018-11-01 23:39:40 +01:00
|
|
|
ASSERT_EQ(decoded_footer.table_magic_number(), kBlockBasedTableMagicNumber);
|
|
|
|
ASSERT_EQ(decoded_footer.checksum(), kxxHash64);
|
|
|
|
ASSERT_EQ(decoded_footer.metaindex_handle().offset(), meta_index.offset());
|
|
|
|
ASSERT_EQ(decoded_footer.metaindex_handle().size(), meta_index.size());
|
|
|
|
ASSERT_EQ(decoded_footer.index_handle().offset(), index.offset());
|
|
|
|
ASSERT_EQ(decoded_footer.index_handle().size(), index.size());
|
|
|
|
ASSERT_EQ(decoded_footer.version(), 1U);
|
|
|
|
}
|
2015-07-20 20:09:14 +02:00
|
|
|
// Plain table is not supported in ROCKSDB_LITE
|
|
|
|
#ifndef ROCKSDB_LITE
|
2014-05-01 20:09:32 +02:00
|
|
|
{
|
|
|
|
// upconvert legacy plain table
|
|
|
|
std::string encoded;
|
2015-01-13 23:33:04 +01:00
|
|
|
Footer footer(kLegacyPlainTableMagicNumber, 0);
|
2014-05-01 20:09:32 +02:00
|
|
|
BlockHandle meta_index(10, 5), index(20, 15);
|
|
|
|
footer.set_metaindex_handle(meta_index);
|
|
|
|
footer.set_index_handle(index);
|
|
|
|
footer.EncodeTo(&encoded);
|
|
|
|
Footer decoded_footer;
|
|
|
|
Slice encoded_slice(encoded);
|
2020-10-08 20:11:52 +02:00
|
|
|
ASSERT_OK(decoded_footer.DecodeFrom(&encoded_slice));
|
2014-05-01 20:09:32 +02:00
|
|
|
ASSERT_EQ(decoded_footer.table_magic_number(), kPlainTableMagicNumber);
|
|
|
|
ASSERT_EQ(decoded_footer.checksum(), kCRC32c);
|
|
|
|
ASSERT_EQ(decoded_footer.metaindex_handle().offset(), meta_index.offset());
|
|
|
|
ASSERT_EQ(decoded_footer.metaindex_handle().size(), meta_index.size());
|
|
|
|
ASSERT_EQ(decoded_footer.index_handle().offset(), index.offset());
|
|
|
|
ASSERT_EQ(decoded_footer.index_handle().size(), index.size());
|
2015-01-13 23:33:04 +01:00
|
|
|
ASSERT_EQ(decoded_footer.version(), 0U);
|
2014-05-01 20:09:32 +02:00
|
|
|
}
|
|
|
|
{
|
|
|
|
// xxhash block based
|
|
|
|
std::string encoded;
|
2015-01-13 23:33:04 +01:00
|
|
|
Footer footer(kPlainTableMagicNumber, 1);
|
2014-05-01 20:09:32 +02:00
|
|
|
BlockHandle meta_index(10, 5), index(20, 15);
|
|
|
|
footer.set_metaindex_handle(meta_index);
|
|
|
|
footer.set_index_handle(index);
|
|
|
|
footer.set_checksum(kxxHash);
|
|
|
|
footer.EncodeTo(&encoded);
|
|
|
|
Footer decoded_footer;
|
|
|
|
Slice encoded_slice(encoded);
|
2020-10-08 20:11:52 +02:00
|
|
|
ASSERT_OK(decoded_footer.DecodeFrom(&encoded_slice));
|
2014-05-01 20:09:32 +02:00
|
|
|
ASSERT_EQ(decoded_footer.table_magic_number(), kPlainTableMagicNumber);
|
|
|
|
ASSERT_EQ(decoded_footer.checksum(), kxxHash);
|
|
|
|
ASSERT_EQ(decoded_footer.metaindex_handle().offset(), meta_index.offset());
|
|
|
|
ASSERT_EQ(decoded_footer.metaindex_handle().size(), meta_index.size());
|
|
|
|
ASSERT_EQ(decoded_footer.index_handle().offset(), index.offset());
|
|
|
|
ASSERT_EQ(decoded_footer.index_handle().size(), index.size());
|
2015-01-13 23:33:04 +01:00
|
|
|
ASSERT_EQ(decoded_footer.version(), 1U);
|
|
|
|
}
|
2015-07-20 20:09:14 +02:00
|
|
|
#endif // !ROCKSDB_LITE
|
2015-01-13 23:33:04 +01:00
|
|
|
{
|
|
|
|
// version == 2
|
|
|
|
std::string encoded;
|
|
|
|
Footer footer(kBlockBasedTableMagicNumber, 2);
|
|
|
|
BlockHandle meta_index(10, 5), index(20, 15);
|
|
|
|
footer.set_metaindex_handle(meta_index);
|
|
|
|
footer.set_index_handle(index);
|
|
|
|
footer.EncodeTo(&encoded);
|
|
|
|
Footer decoded_footer;
|
|
|
|
Slice encoded_slice(encoded);
|
2020-10-08 20:11:52 +02:00
|
|
|
ASSERT_OK(decoded_footer.DecodeFrom(&encoded_slice));
|
2015-01-13 23:33:04 +01:00
|
|
|
ASSERT_EQ(decoded_footer.table_magic_number(), kBlockBasedTableMagicNumber);
|
|
|
|
ASSERT_EQ(decoded_footer.checksum(), kCRC32c);
|
|
|
|
ASSERT_EQ(decoded_footer.metaindex_handle().offset(), meta_index.offset());
|
|
|
|
ASSERT_EQ(decoded_footer.metaindex_handle().size(), meta_index.size());
|
|
|
|
ASSERT_EQ(decoded_footer.index_handle().offset(), index.offset());
|
|
|
|
ASSERT_EQ(decoded_footer.index_handle().size(), index.size());
|
|
|
|
ASSERT_EQ(decoded_footer.version(), 2U);
|
2014-05-01 20:09:32 +02:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2016-02-05 19:22:37 +01:00
|
|
|
class IndexBlockRestartIntervalTest
|
2018-06-05 04:59:44 +02:00
|
|
|
: public TableTest,
|
2018-08-10 01:49:45 +02:00
|
|
|
public ::testing::WithParamInterface<std::pair<int, bool>> {
|
2016-02-05 19:22:37 +01:00
|
|
|
public:
|
2018-08-10 01:49:45 +02:00
|
|
|
static std::vector<std::pair<int, bool>> GetRestartValues() {
|
|
|
|
return {{-1, false}, {0, false}, {1, false}, {8, false},
|
|
|
|
{16, false}, {32, false}, {-1, true}, {0, true},
|
|
|
|
{1, true}, {8, true}, {16, true}, {32, true}};
|
|
|
|
}
|
2016-02-05 19:22:37 +01:00
|
|
|
};
|
|
|
|
|
2020-06-04 00:53:09 +02:00
|
|
|
INSTANTIATE_TEST_CASE_P(
|
2016-02-05 19:22:37 +01:00
|
|
|
IndexBlockRestartIntervalTest, IndexBlockRestartIntervalTest,
|
|
|
|
::testing::ValuesIn(IndexBlockRestartIntervalTest::GetRestartValues()));
|
|
|
|
|
|
|
|
TEST_P(IndexBlockRestartIntervalTest, IndexBlockRestartInterval) {
|
|
|
|
const int kKeysInTable = 10000;
|
|
|
|
const int kKeySize = 100;
|
|
|
|
const int kValSize = 500;
|
|
|
|
|
2018-08-10 01:49:45 +02:00
|
|
|
const int index_block_restart_interval = std::get<0>(GetParam());
|
|
|
|
const bool value_delta_encoding = std::get<1>(GetParam());
|
2016-02-05 19:22:37 +01:00
|
|
|
|
|
|
|
Options options;
|
|
|
|
BlockBasedTableOptions table_options;
|
|
|
|
table_options.block_size = 64; // small block size to get big index block
|
|
|
|
table_options.index_block_restart_interval = index_block_restart_interval;
|
2018-08-10 01:49:45 +02:00
|
|
|
if (value_delta_encoding) {
|
|
|
|
table_options.format_version = 4;
|
|
|
|
}
|
2016-02-05 19:22:37 +01:00
|
|
|
options.table_factory.reset(new BlockBasedTableFactory(table_options));
|
|
|
|
|
|
|
|
TableConstructor c(BytewiseComparator());
|
|
|
|
static Random rnd(301);
|
|
|
|
for (int i = 0; i < kKeysInTable; i++) {
|
2020-07-09 23:33:42 +02:00
|
|
|
InternalKey k(rnd.RandomString(kKeySize), 0, kTypeValue);
|
|
|
|
c.Add(k.Encode().ToString(), rnd.RandomString(kValSize));
|
2016-02-05 19:22:37 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
std::vector<std::string> keys;
|
|
|
|
stl_wrappers::KVMap kvmap;
|
|
|
|
std::unique_ptr<InternalKeyComparator> comparator(
|
|
|
|
new InternalKeyComparator(BytewiseComparator()));
|
|
|
|
const ImmutableCFOptions ioptions(options);
|
2018-05-21 23:33:55 +02:00
|
|
|
const MutableCFOptions moptions(options);
|
|
|
|
c.Finish(options, ioptions, moptions, table_options, *comparator, &keys,
|
|
|
|
&kvmap);
|
2016-02-05 19:22:37 +01:00
|
|
|
auto reader = c.GetTableReader();
|
|
|
|
|
2020-08-04 00:21:56 +02:00
|
|
|
ReadOptions read_options;
|
2019-06-20 23:28:22 +02:00
|
|
|
std::unique_ptr<InternalIterator> db_iter(reader->NewIterator(
|
2020-08-04 00:21:56 +02:00
|
|
|
read_options, moptions.prefix_extractor.get(), /*arena=*/nullptr,
|
2019-06-20 23:28:22 +02:00
|
|
|
/*skip_filters=*/false, TableReaderCaller::kUncategorized));
|
2016-02-05 19:22:37 +01:00
|
|
|
|
|
|
|
// Test point lookup
|
|
|
|
for (auto& kv : kvmap) {
|
|
|
|
db_iter->Seek(kv.first);
|
|
|
|
|
|
|
|
ASSERT_TRUE(db_iter->Valid());
|
|
|
|
ASSERT_OK(db_iter->status());
|
|
|
|
ASSERT_EQ(db_iter->key(), kv.first);
|
|
|
|
ASSERT_EQ(db_iter->value(), kv.second);
|
|
|
|
}
|
|
|
|
|
|
|
|
// Test iterating
|
|
|
|
auto kv_iter = kvmap.begin();
|
|
|
|
for (db_iter->SeekToFirst(); db_iter->Valid(); db_iter->Next()) {
|
|
|
|
ASSERT_EQ(db_iter->key(), kv_iter->first);
|
|
|
|
ASSERT_EQ(db_iter->value(), kv_iter->second);
|
|
|
|
kv_iter++;
|
|
|
|
}
|
|
|
|
ASSERT_EQ(kv_iter, kvmap.end());
|
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 19:42:39 +02:00
|
|
|
c.ResetTableReader();
|
2016-02-05 19:22:37 +01:00
|
|
|
}
|
|
|
|
|
2016-02-23 01:33:26 +01:00
|
|
|
class PrefixTest : public testing::Test {
|
|
|
|
public:
|
|
|
|
PrefixTest() : testing::Test() {}
|
2019-02-14 22:52:47 +01:00
|
|
|
~PrefixTest() override {}
|
2016-02-23 01:33:26 +01:00
|
|
|
};
|
|
|
|
|
|
|
|
namespace {
|
|
|
|
// A simple PrefixExtractor that only works for test PrefixAndWholeKeyTest
|
2020-02-20 21:07:53 +01:00
|
|
|
class TestPrefixExtractor : public ROCKSDB_NAMESPACE::SliceTransform {
|
2016-02-23 01:33:26 +01:00
|
|
|
public:
|
|
|
|
~TestPrefixExtractor() override{};
|
|
|
|
const char* Name() const override { return "TestPrefixExtractor"; }
|
|
|
|
|
2020-02-20 21:07:53 +01:00
|
|
|
ROCKSDB_NAMESPACE::Slice Transform(
|
|
|
|
const ROCKSDB_NAMESPACE::Slice& src) const override {
|
2016-02-23 01:33:26 +01:00
|
|
|
assert(IsValid(src));
|
2020-02-20 21:07:53 +01:00
|
|
|
return ROCKSDB_NAMESPACE::Slice(src.data(), 3);
|
2016-02-23 01:33:26 +01:00
|
|
|
}
|
|
|
|
|
2020-02-20 21:07:53 +01:00
|
|
|
bool InDomain(const ROCKSDB_NAMESPACE::Slice& src) const override {
|
2016-02-23 01:33:26 +01:00
|
|
|
assert(IsValid(src));
|
|
|
|
return true;
|
|
|
|
}
|
|
|
|
|
2020-02-20 21:07:53 +01:00
|
|
|
bool InRange(const ROCKSDB_NAMESPACE::Slice& /*dst*/) const override {
|
|
|
|
return true;
|
|
|
|
}
|
2016-02-23 01:33:26 +01:00
|
|
|
|
2020-02-20 21:07:53 +01:00
|
|
|
bool IsValid(const ROCKSDB_NAMESPACE::Slice& src) const {
|
2016-02-23 01:33:26 +01:00
|
|
|
if (src.size() != 4) {
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
if (src[0] != '[') {
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
if (src[1] < '0' || src[1] > '9') {
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
if (src[2] != ']') {
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
if (src[3] < '0' || src[3] > '9') {
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
return true;
|
|
|
|
}
|
|
|
|
};
|
|
|
|
} // namespace
|
|
|
|
|
|
|
|
TEST_F(PrefixTest, PrefixAndWholeKeyTest) {
|
2020-02-20 21:07:53 +01:00
|
|
|
ROCKSDB_NAMESPACE::Options options;
|
|
|
|
options.compaction_style = ROCKSDB_NAMESPACE::kCompactionStyleUniversal;
|
2016-02-23 01:33:26 +01:00
|
|
|
options.num_levels = 20;
|
|
|
|
options.create_if_missing = true;
|
|
|
|
options.optimize_filters_for_hits = false;
|
|
|
|
options.target_file_size_base = 268435456;
|
|
|
|
options.prefix_extractor = std::make_shared<TestPrefixExtractor>();
|
2020-02-20 21:07:53 +01:00
|
|
|
ROCKSDB_NAMESPACE::BlockBasedTableOptions bbto;
|
|
|
|
bbto.filter_policy.reset(ROCKSDB_NAMESPACE::NewBloomFilterPolicy(10));
|
2016-02-23 01:33:26 +01:00
|
|
|
bbto.block_size = 262144;
|
|
|
|
bbto.whole_key_filtering = true;
|
|
|
|
|
2018-07-14 02:18:39 +02:00
|
|
|
const std::string kDBPath = test::PerThreadDBPath("table_prefix_test");
|
2016-02-23 01:33:26 +01:00
|
|
|
options.table_factory.reset(NewBlockBasedTableFactory(bbto));
|
2020-12-24 01:54:05 +01:00
|
|
|
ASSERT_OK(DestroyDB(kDBPath, options));
|
2020-02-20 21:07:53 +01:00
|
|
|
ROCKSDB_NAMESPACE::DB* db;
|
|
|
|
ASSERT_OK(ROCKSDB_NAMESPACE::DB::Open(options, kDBPath, &db));
|
2016-02-23 01:33:26 +01:00
|
|
|
|
|
|
|
// Create a bunch of keys with 10 filters.
|
|
|
|
for (int i = 0; i < 10; i++) {
|
|
|
|
std::string prefix = "[" + std::to_string(i) + "]";
|
|
|
|
for (int j = 0; j < 10; j++) {
|
|
|
|
std::string key = prefix + std::to_string(j);
|
2020-10-08 20:11:52 +02:00
|
|
|
ASSERT_OK(db->Put(ROCKSDB_NAMESPACE::WriteOptions(), key, "1"));
|
2016-02-23 01:33:26 +01:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
// Trigger compaction.
|
2020-10-08 20:11:52 +02:00
|
|
|
ASSERT_OK(db->CompactRange(CompactRangeOptions(), nullptr, nullptr));
|
2016-02-23 01:33:26 +01:00
|
|
|
delete db;
|
|
|
|
// In the second round, turn whole_key_filtering off and expect
|
|
|
|
// rocksdb still works.
|
|
|
|
}
|
|
|
|
|
2018-07-28 01:00:26 +02:00
|
|
|
/*
|
|
|
|
* Disable TableWithGlobalSeqno since RocksDB does not store global_seqno in
|
|
|
|
* the SST file any more. Instead, RocksDB deduces global_seqno from the
|
|
|
|
* MANIFEST while reading from an SST. Therefore, it's not possible to test the
|
|
|
|
* functionality of global_seqno in a single, isolated unit test without the
|
|
|
|
* involvement of Version, VersionSet, etc.
|
|
|
|
*/
|
|
|
|
TEST_P(BlockBasedTableTest, DISABLED_TableWithGlobalSeqno) {
|
2018-06-05 04:59:44 +02:00
|
|
|
BlockBasedTableOptions bbto = GetBlockBasedTableOptions();
|
2016-10-19 01:59:37 +02:00
|
|
|
test::StringSink* sink = new test::StringSink();
|
2021-01-05 00:59:52 +01:00
|
|
|
std::unique_ptr<FSWritableFile> holder(sink);
|
|
|
|
std::unique_ptr<WritableFileWriter> file_writer(new WritableFileWriter(
|
|
|
|
std::move(holder), "" /* don't care */, FileOptions()));
|
2016-10-19 01:59:37 +02:00
|
|
|
Options options;
|
|
|
|
options.table_factory.reset(NewBlockBasedTableFactory(bbto));
|
|
|
|
const ImmutableCFOptions ioptions(options);
|
2018-05-21 23:33:55 +02:00
|
|
|
const MutableCFOptions moptions(options);
|
2016-10-19 01:59:37 +02:00
|
|
|
InternalKeyComparator ikc(options.comparator);
|
|
|
|
std::vector<std::unique_ptr<IntTblPropCollectorFactory>>
|
|
|
|
int_tbl_prop_collector_factories;
|
|
|
|
int_tbl_prop_collector_factories.emplace_back(
|
|
|
|
new SstFileWriterPropertiesCollectorFactory(2 /* version */,
|
|
|
|
0 /* global_seqno*/));
|
|
|
|
std::string column_family_name;
|
|
|
|
std::unique_ptr<TableBuilder> builder(options.table_factory->NewTableBuilder(
|
2018-05-21 23:33:55 +02:00
|
|
|
TableBuilderOptions(ioptions, moptions, ikc,
|
|
|
|
&int_tbl_prop_collector_factories, kNoCompression,
|
2019-03-18 20:07:35 +01:00
|
|
|
0 /* sample_for_compression */, CompressionOptions(),
|
|
|
|
false /* skip_filters */, column_family_name, -1),
|
2016-10-19 01:59:37 +02:00
|
|
|
TablePropertiesCollectorFactory::Context::kUnknownColumnFamily,
|
|
|
|
file_writer.get()));
|
|
|
|
|
|
|
|
for (char c = 'a'; c <= 'z'; ++c) {
|
|
|
|
std::string key(8, c);
|
|
|
|
std::string value = key;
|
|
|
|
InternalKey ik(key, 0, kTypeValue);
|
|
|
|
|
|
|
|
builder->Add(ik.Encode(), value);
|
|
|
|
}
|
|
|
|
ASSERT_OK(builder->Finish());
|
2020-12-24 01:54:05 +01:00
|
|
|
ASSERT_OK(file_writer->Flush());
|
2016-10-19 01:59:37 +02:00
|
|
|
|
|
|
|
test::RandomRWStringSink ss_rw(sink);
|
|
|
|
uint32_t version;
|
|
|
|
uint64_t global_seqno;
|
|
|
|
uint64_t global_seqno_offset;
|
|
|
|
|
|
|
|
// Helper function to get version, global_seqno, global_seqno_offset
|
|
|
|
std::function<void()> GetVersionAndGlobalSeqno = [&]() {
|
2021-01-05 00:59:52 +01:00
|
|
|
std::unique_ptr<FSRandomAccessFile> source(
|
|
|
|
new test::StringSource(ss_rw.contents(), 73342, true));
|
2018-11-09 20:17:34 +01:00
|
|
|
std::unique_ptr<RandomAccessFileReader> file_reader(
|
2021-01-05 00:59:52 +01:00
|
|
|
new RandomAccessFileReader(std::move(source), ""));
|
2016-10-19 01:59:37 +02:00
|
|
|
|
|
|
|
TableProperties* props = nullptr;
|
|
|
|
ASSERT_OK(ReadTableProperties(file_reader.get(), ss_rw.contents().size(),
|
|
|
|
kBlockBasedTableMagicNumber, ioptions,
|
2018-06-16 04:24:21 +02:00
|
|
|
&props, true /* compression_type_missing */));
|
2016-10-19 01:59:37 +02:00
|
|
|
|
|
|
|
UserCollectedProperties user_props = props->user_collected_properties;
|
|
|
|
version = DecodeFixed32(
|
|
|
|
user_props[ExternalSstFilePropertyNames::kVersion].c_str());
|
|
|
|
global_seqno = DecodeFixed64(
|
|
|
|
user_props[ExternalSstFilePropertyNames::kGlobalSeqno].c_str());
|
|
|
|
global_seqno_offset =
|
|
|
|
props->properties_offsets[ExternalSstFilePropertyNames::kGlobalSeqno];
|
|
|
|
|
|
|
|
delete props;
|
|
|
|
};
|
|
|
|
|
|
|
|
// Helper function to update the value of the global seqno in the file
|
|
|
|
std::function<void(uint64_t)> SetGlobalSeqno = [&](uint64_t val) {
|
|
|
|
std::string new_global_seqno;
|
|
|
|
PutFixed64(&new_global_seqno, val);
|
|
|
|
|
2021-01-05 00:59:52 +01:00
|
|
|
ASSERT_OK(ss_rw.Write(global_seqno_offset, new_global_seqno, IOOptions(),
|
|
|
|
nullptr));
|
2016-10-19 01:59:37 +02:00
|
|
|
};
|
|
|
|
|
|
|
|
// Helper function to get the contents of the table InternalIterator
|
2018-11-09 20:17:34 +01:00
|
|
|
std::unique_ptr<TableReader> table_reader;
|
2020-08-04 00:21:56 +02:00
|
|
|
const ReadOptions read_options;
|
2016-10-19 01:59:37 +02:00
|
|
|
std::function<InternalIterator*()> GetTableInternalIter = [&]() {
|
2021-01-05 00:59:52 +01:00
|
|
|
std::unique_ptr<FSRandomAccessFile> source(
|
|
|
|
new test::StringSource(ss_rw.contents(), 73342, true));
|
2018-11-09 20:17:34 +01:00
|
|
|
std::unique_ptr<RandomAccessFileReader> file_reader(
|
2021-01-05 00:59:52 +01:00
|
|
|
new RandomAccessFileReader(std::move(source), ""));
|
2016-10-19 01:59:37 +02:00
|
|
|
|
|
|
|
options.table_factory->NewTableReader(
|
2018-05-21 23:33:55 +02:00
|
|
|
TableReaderOptions(ioptions, moptions.prefix_extractor.get(),
|
|
|
|
EnvOptions(), ikc),
|
|
|
|
std::move(file_reader), ss_rw.contents().size(), &table_reader);
|
2016-10-19 01:59:37 +02:00
|
|
|
|
2019-06-20 23:28:22 +02:00
|
|
|
return table_reader->NewIterator(
|
2020-08-04 00:21:56 +02:00
|
|
|
read_options, moptions.prefix_extractor.get(), /*arena=*/nullptr,
|
2019-06-20 23:28:22 +02:00
|
|
|
/*skip_filters=*/false, TableReaderCaller::kUncategorized);
|
2016-10-19 01:59:37 +02:00
|
|
|
};
|
|
|
|
|
|
|
|
GetVersionAndGlobalSeqno();
|
2019-09-09 20:22:28 +02:00
|
|
|
ASSERT_EQ(2u, version);
|
|
|
|
ASSERT_EQ(0u, global_seqno);
|
2016-10-19 01:59:37 +02:00
|
|
|
|
|
|
|
InternalIterator* iter = GetTableInternalIter();
|
|
|
|
char current_c = 'a';
|
|
|
|
for (iter->SeekToFirst(); iter->Valid(); iter->Next()) {
|
|
|
|
ParsedInternalKey pik;
|
2020-10-28 18:11:13 +01:00
|
|
|
ASSERT_OK(ParseInternalKey(iter->key(), &pik, true /* log_err_key */));
|
2016-10-19 01:59:37 +02:00
|
|
|
|
|
|
|
ASSERT_EQ(pik.type, ValueType::kTypeValue);
|
|
|
|
ASSERT_EQ(pik.sequence, 0);
|
|
|
|
ASSERT_EQ(pik.user_key, iter->value());
|
|
|
|
ASSERT_EQ(pik.user_key.ToString(), std::string(8, current_c));
|
|
|
|
current_c++;
|
|
|
|
}
|
|
|
|
ASSERT_EQ(current_c, 'z' + 1);
|
|
|
|
delete iter;
|
|
|
|
|
|
|
|
// Update global sequence number to 10
|
|
|
|
SetGlobalSeqno(10);
|
|
|
|
GetVersionAndGlobalSeqno();
|
2019-09-09 20:22:28 +02:00
|
|
|
ASSERT_EQ(2u, version);
|
|
|
|
ASSERT_EQ(10u, global_seqno);
|
2016-10-19 01:59:37 +02:00
|
|
|
|
|
|
|
iter = GetTableInternalIter();
|
|
|
|
current_c = 'a';
|
|
|
|
for (iter->SeekToFirst(); iter->Valid(); iter->Next()) {
|
|
|
|
ParsedInternalKey pik;
|
2020-10-28 18:11:13 +01:00
|
|
|
ASSERT_OK(ParseInternalKey(iter->key(), &pik, true /* log_err_key */));
|
2016-10-19 01:59:37 +02:00
|
|
|
|
|
|
|
ASSERT_EQ(pik.type, ValueType::kTypeValue);
|
|
|
|
ASSERT_EQ(pik.sequence, 10);
|
|
|
|
ASSERT_EQ(pik.user_key, iter->value());
|
|
|
|
ASSERT_EQ(pik.user_key.ToString(), std::string(8, current_c));
|
|
|
|
current_c++;
|
|
|
|
}
|
|
|
|
ASSERT_EQ(current_c, 'z' + 1);
|
|
|
|
|
|
|
|
// Verify Seek
|
|
|
|
for (char c = 'a'; c <= 'z'; c++) {
|
|
|
|
std::string k = std::string(8, c);
|
|
|
|
InternalKey ik(k, 10, kValueTypeForSeek);
|
|
|
|
iter->Seek(ik.Encode());
|
|
|
|
ASSERT_TRUE(iter->Valid());
|
|
|
|
|
|
|
|
ParsedInternalKey pik;
|
2020-10-28 18:11:13 +01:00
|
|
|
ASSERT_OK(ParseInternalKey(iter->key(), &pik, true /* log_err_key */));
|
2016-10-19 01:59:37 +02:00
|
|
|
|
|
|
|
ASSERT_EQ(pik.type, ValueType::kTypeValue);
|
|
|
|
ASSERT_EQ(pik.sequence, 10);
|
|
|
|
ASSERT_EQ(pik.user_key.ToString(), k);
|
|
|
|
ASSERT_EQ(iter->value().ToString(), k);
|
|
|
|
}
|
|
|
|
delete iter;
|
|
|
|
|
|
|
|
// Update global sequence number to 3
|
|
|
|
SetGlobalSeqno(3);
|
|
|
|
GetVersionAndGlobalSeqno();
|
2019-09-09 20:22:28 +02:00
|
|
|
ASSERT_EQ(2u, version);
|
|
|
|
ASSERT_EQ(3u, global_seqno);
|
2016-10-19 01:59:37 +02:00
|
|
|
|
|
|
|
iter = GetTableInternalIter();
|
|
|
|
current_c = 'a';
|
|
|
|
for (iter->SeekToFirst(); iter->Valid(); iter->Next()) {
|
|
|
|
ParsedInternalKey pik;
|
2020-10-28 18:11:13 +01:00
|
|
|
ASSERT_OK(ParseInternalKey(iter->key(), &pik, true /* log_err_key */));
|
2016-10-19 01:59:37 +02:00
|
|
|
|
|
|
|
ASSERT_EQ(pik.type, ValueType::kTypeValue);
|
|
|
|
ASSERT_EQ(pik.sequence, 3);
|
|
|
|
ASSERT_EQ(pik.user_key, iter->value());
|
|
|
|
ASSERT_EQ(pik.user_key.ToString(), std::string(8, current_c));
|
|
|
|
current_c++;
|
|
|
|
}
|
|
|
|
ASSERT_EQ(current_c, 'z' + 1);
|
|
|
|
|
|
|
|
// Verify Seek
|
|
|
|
for (char c = 'a'; c <= 'z'; c++) {
|
|
|
|
std::string k = std::string(8, c);
|
|
|
|
// seqno=4 is less than 3 so we still should get our key
|
|
|
|
InternalKey ik(k, 4, kValueTypeForSeek);
|
|
|
|
iter->Seek(ik.Encode());
|
|
|
|
ASSERT_TRUE(iter->Valid());
|
|
|
|
|
|
|
|
ParsedInternalKey pik;
|
2020-10-28 18:11:13 +01:00
|
|
|
ASSERT_OK(ParseInternalKey(iter->key(), &pik, true /* log_err_key */));
|
2016-10-19 01:59:37 +02:00
|
|
|
|
|
|
|
ASSERT_EQ(pik.type, ValueType::kTypeValue);
|
|
|
|
ASSERT_EQ(pik.sequence, 3);
|
|
|
|
ASSERT_EQ(pik.user_key.ToString(), k);
|
|
|
|
ASSERT_EQ(iter->value().ToString(), k);
|
|
|
|
}
|
|
|
|
|
|
|
|
delete iter;
|
|
|
|
}
|
|
|
|
|
2018-06-05 04:59:44 +02:00
|
|
|
TEST_P(BlockBasedTableTest, BlockAlignTest) {
|
|
|
|
BlockBasedTableOptions bbto = GetBlockBasedTableOptions();
|
2018-03-27 05:14:24 +02:00
|
|
|
bbto.block_align = true;
|
|
|
|
test::StringSink* sink = new test::StringSink();
|
2021-01-05 00:59:52 +01:00
|
|
|
std::unique_ptr<FSWritableFile> holder(sink);
|
|
|
|
std::unique_ptr<WritableFileWriter> file_writer(new WritableFileWriter(
|
|
|
|
std::move(holder), "" /* don't care */, FileOptions()));
|
2018-03-27 05:14:24 +02:00
|
|
|
Options options;
|
|
|
|
options.compression = kNoCompression;
|
|
|
|
options.table_factory.reset(NewBlockBasedTableFactory(bbto));
|
|
|
|
const ImmutableCFOptions ioptions(options);
|
2018-05-21 23:33:55 +02:00
|
|
|
const MutableCFOptions moptions(options);
|
2018-03-27 05:14:24 +02:00
|
|
|
InternalKeyComparator ikc(options.comparator);
|
|
|
|
std::vector<std::unique_ptr<IntTblPropCollectorFactory>>
|
|
|
|
int_tbl_prop_collector_factories;
|
|
|
|
std::string column_family_name;
|
|
|
|
std::unique_ptr<TableBuilder> builder(options.table_factory->NewTableBuilder(
|
2018-05-21 23:33:55 +02:00
|
|
|
TableBuilderOptions(ioptions, moptions, ikc,
|
|
|
|
&int_tbl_prop_collector_factories, kNoCompression,
|
2019-03-18 20:07:35 +01:00
|
|
|
0 /* sample_for_compression */, CompressionOptions(),
|
|
|
|
false /* skip_filters */, column_family_name, -1),
|
2018-03-27 05:14:24 +02:00
|
|
|
TablePropertiesCollectorFactory::Context::kUnknownColumnFamily,
|
|
|
|
file_writer.get()));
|
|
|
|
|
|
|
|
for (int i = 1; i <= 10000; ++i) {
|
|
|
|
std::ostringstream ostr;
|
|
|
|
ostr << std::setfill('0') << std::setw(5) << i;
|
|
|
|
std::string key = ostr.str();
|
|
|
|
std::string value = "val";
|
|
|
|
InternalKey ik(key, 0, kTypeValue);
|
|
|
|
|
|
|
|
builder->Add(ik.Encode(), value);
|
|
|
|
}
|
|
|
|
ASSERT_OK(builder->Finish());
|
2020-12-24 01:54:05 +01:00
|
|
|
ASSERT_OK(file_writer->Flush());
|
2018-03-27 05:14:24 +02:00
|
|
|
|
2021-01-05 00:59:52 +01:00
|
|
|
std::unique_ptr<FSRandomAccessFile> source(
|
|
|
|
new test::StringSource(sink->contents(), 73342, false));
|
2018-11-09 20:17:34 +01:00
|
|
|
std::unique_ptr<RandomAccessFileReader> file_reader(
|
2021-01-05 00:59:52 +01:00
|
|
|
new RandomAccessFileReader(std::move(source), "test"));
|
2018-03-27 05:14:24 +02:00
|
|
|
// Helper function to get version, global_seqno, global_seqno_offset
|
|
|
|
std::function<void()> VerifyBlockAlignment = [&]() {
|
|
|
|
TableProperties* props = nullptr;
|
2021-01-05 00:59:52 +01:00
|
|
|
ASSERT_OK(ReadTableProperties(file_reader.get(), sink->contents().size(),
|
|
|
|
kBlockBasedTableMagicNumber, ioptions, &props,
|
|
|
|
true /* compression_type_missing */));
|
2018-03-27 05:14:24 +02:00
|
|
|
|
|
|
|
uint64_t data_block_size = props->data_size / props->num_data_blocks;
|
|
|
|
ASSERT_EQ(data_block_size, 4096);
|
|
|
|
ASSERT_EQ(props->data_size, data_block_size * props->num_data_blocks);
|
|
|
|
delete props;
|
|
|
|
};
|
|
|
|
|
|
|
|
VerifyBlockAlignment();
|
|
|
|
|
|
|
|
// The below block of code verifies that we can read back the keys. Set
|
|
|
|
// block_align to false when creating the reader to ensure we can flip between
|
|
|
|
// the two modes without any issues
|
|
|
|
std::unique_ptr<TableReader> table_reader;
|
|
|
|
bbto.block_align = false;
|
|
|
|
Options options2;
|
|
|
|
options2.table_factory.reset(NewBlockBasedTableFactory(bbto));
|
|
|
|
ImmutableCFOptions ioptions2(options2);
|
2018-05-21 23:33:55 +02:00
|
|
|
const MutableCFOptions moptions2(options2);
|
|
|
|
|
2018-03-27 05:14:24 +02:00
|
|
|
ASSERT_OK(ioptions.table_factory->NewTableReader(
|
2018-05-21 23:33:55 +02:00
|
|
|
TableReaderOptions(ioptions2, moptions2.prefix_extractor.get(),
|
|
|
|
EnvOptions(),
|
2018-03-27 05:14:24 +02:00
|
|
|
GetPlainInternalComparator(options2.comparator)),
|
2021-01-05 00:59:52 +01:00
|
|
|
std::move(file_reader), sink->contents().size(), &table_reader));
|
2018-03-27 05:14:24 +02:00
|
|
|
|
2020-08-04 00:21:56 +02:00
|
|
|
ReadOptions read_options;
|
2018-05-21 23:33:55 +02:00
|
|
|
std::unique_ptr<InternalIterator> db_iter(table_reader->NewIterator(
|
2020-08-04 00:21:56 +02:00
|
|
|
read_options, moptions2.prefix_extractor.get(), /*arena=*/nullptr,
|
2019-06-20 23:28:22 +02:00
|
|
|
/*skip_filters=*/false, TableReaderCaller::kUncategorized));
|
2018-03-27 05:14:24 +02:00
|
|
|
|
|
|
|
int expected_key = 1;
|
|
|
|
for (db_iter->SeekToFirst(); db_iter->Valid(); db_iter->Next()) {
|
|
|
|
std::ostringstream ostr;
|
|
|
|
ostr << std::setfill('0') << std::setw(5) << expected_key++;
|
|
|
|
std::string key = ostr.str();
|
|
|
|
std::string value = "val";
|
|
|
|
|
|
|
|
ASSERT_OK(db_iter->status());
|
|
|
|
ASSERT_EQ(ExtractUserKey(db_iter->key()).ToString(), key);
|
|
|
|
ASSERT_EQ(db_iter->value().ToString(), value);
|
|
|
|
}
|
|
|
|
expected_key--;
|
|
|
|
ASSERT_EQ(expected_key, 10000);
|
|
|
|
table_reader.reset();
|
|
|
|
}
|
|
|
|
|
2018-06-12 21:46:48 +02:00
|
|
|
TEST_P(BlockBasedTableTest, PropertiesBlockRestartPointTest) {
|
|
|
|
BlockBasedTableOptions bbto = GetBlockBasedTableOptions();
|
|
|
|
bbto.block_align = true;
|
|
|
|
test::StringSink* sink = new test::StringSink();
|
2021-01-05 00:59:52 +01:00
|
|
|
std::unique_ptr<FSWritableFile> holder(sink);
|
|
|
|
std::unique_ptr<WritableFileWriter> file_writer(new WritableFileWriter(
|
|
|
|
std::move(holder), "" /* don't care */, FileOptions()));
|
2018-06-12 21:46:48 +02:00
|
|
|
|
|
|
|
Options options;
|
|
|
|
options.compression = kNoCompression;
|
|
|
|
options.table_factory.reset(NewBlockBasedTableFactory(bbto));
|
|
|
|
|
|
|
|
const ImmutableCFOptions ioptions(options);
|
|
|
|
const MutableCFOptions moptions(options);
|
|
|
|
InternalKeyComparator ikc(options.comparator);
|
|
|
|
std::vector<std::unique_ptr<IntTblPropCollectorFactory>>
|
|
|
|
int_tbl_prop_collector_factories;
|
|
|
|
std::string column_family_name;
|
|
|
|
|
|
|
|
std::unique_ptr<TableBuilder> builder(options.table_factory->NewTableBuilder(
|
|
|
|
TableBuilderOptions(ioptions, moptions, ikc,
|
|
|
|
&int_tbl_prop_collector_factories, kNoCompression,
|
2019-03-18 20:07:35 +01:00
|
|
|
0 /* sample_for_compression */, CompressionOptions(),
|
|
|
|
false /* skip_filters */, column_family_name, -1),
|
2018-06-12 21:46:48 +02:00
|
|
|
TablePropertiesCollectorFactory::Context::kUnknownColumnFamily,
|
|
|
|
file_writer.get()));
|
|
|
|
|
|
|
|
for (int i = 1; i <= 10000; ++i) {
|
|
|
|
std::ostringstream ostr;
|
|
|
|
ostr << std::setfill('0') << std::setw(5) << i;
|
|
|
|
std::string key = ostr.str();
|
|
|
|
std::string value = "val";
|
|
|
|
InternalKey ik(key, 0, kTypeValue);
|
|
|
|
|
|
|
|
builder->Add(ik.Encode(), value);
|
|
|
|
}
|
|
|
|
ASSERT_OK(builder->Finish());
|
2020-12-24 01:54:05 +01:00
|
|
|
ASSERT_OK(file_writer->Flush());
|
2018-06-12 21:46:48 +02:00
|
|
|
|
2021-01-05 00:59:52 +01:00
|
|
|
std::unique_ptr<FSRandomAccessFile> source(
|
|
|
|
new test::StringSource(sink->contents(), 73342, true));
|
2018-11-09 20:17:34 +01:00
|
|
|
std::unique_ptr<RandomAccessFileReader> file_reader(
|
2021-01-05 00:59:52 +01:00
|
|
|
new RandomAccessFileReader(std::move(source), "test"));
|
2018-06-12 21:46:48 +02:00
|
|
|
|
|
|
|
{
|
|
|
|
RandomAccessFileReader* file = file_reader.get();
|
2021-01-05 00:59:52 +01:00
|
|
|
uint64_t file_size = sink->contents().size();
|
2018-06-12 21:46:48 +02:00
|
|
|
|
|
|
|
Footer footer;
|
2020-06-29 23:51:57 +02:00
|
|
|
IOOptions opts;
|
|
|
|
ASSERT_OK(ReadFooterFromFile(opts, file, nullptr /* prefetch_buffer */,
|
|
|
|
file_size, &footer,
|
|
|
|
kBlockBasedTableMagicNumber));
|
2018-06-12 21:46:48 +02:00
|
|
|
|
2019-06-19 04:00:03 +02:00
|
|
|
auto BlockFetchHelper = [&](const BlockHandle& handle, BlockType block_type,
|
2018-06-12 21:46:48 +02:00
|
|
|
BlockContents* contents) {
|
|
|
|
ReadOptions read_options;
|
|
|
|
read_options.verify_checksums = false;
|
|
|
|
PersistentCacheOptions cache_options;
|
|
|
|
|
2018-11-29 02:58:08 +01:00
|
|
|
BlockFetcher block_fetcher(
|
|
|
|
file, nullptr /* prefetch_buffer */, footer, read_options, handle,
|
|
|
|
contents, ioptions, false /* decompress */,
|
2019-06-19 04:00:03 +02:00
|
|
|
false /*maybe_compressed*/, block_type,
|
|
|
|
UncompressionDict::GetEmptyDict(), cache_options);
|
2018-06-12 21:46:48 +02:00
|
|
|
|
|
|
|
ASSERT_OK(block_fetcher.ReadBlockContents());
|
|
|
|
};
|
|
|
|
|
|
|
|
// -- Read metaindex block
|
|
|
|
auto metaindex_handle = footer.metaindex_handle();
|
|
|
|
BlockContents metaindex_contents;
|
|
|
|
|
2019-06-19 04:00:03 +02:00
|
|
|
BlockFetchHelper(metaindex_handle, BlockType::kMetaIndex,
|
2019-06-19 23:07:36 +02:00
|
|
|
&metaindex_contents);
|
2020-02-26 00:29:17 +01:00
|
|
|
Block metaindex_block(std::move(metaindex_contents));
|
2018-06-12 21:46:48 +02:00
|
|
|
|
Add an option to put first key of each sst block in the index (#5289)
Summary:
The first key is used to defer reading the data block until this file gets to the top of merging iterator's heap. For short range scans, most files never make it to the top of the heap, so this change can reduce read amplification by a lot sometimes.
Consider the following workload. There are a few data streams (we'll be calling them "logs"), each stream consisting of a sequence of blobs (we'll be calling them "records"). Each record is identified by log ID and a sequence number within the log. RocksDB key is concatenation of log ID and sequence number (big endian). Reads are mostly relatively short range scans, each within a single log. Writes are mostly sequential for each log, but writes to different logs are randomly interleaved. Compactions are disabled; instead, when we accumulate a few tens of sst files, we create a new column family and start writing to it.
So, a typical sst file consists of a few ranges of blocks, each range corresponding to one log ID (we use FlushBlockPolicy to cut blocks at log boundaries). A typical read would go like this. First, iterator Seek() reads one block from each sst file. Then a series of Next()s move through one sst file (since writes to each log are mostly sequential) until the subiterator reaches the end of this log in this sst file; then Next() switches to the next sst file and reads sequentially from that, and so on. Often a range scan will only return records from a small number of blocks in small number of sst files; in this case, the cost of initial Seek() reading one block from each file may be bigger than the cost of reading the actually useful blocks.
Neither iterate_upper_bound nor bloom filters can prevent reading one block from each file in Seek(). But this PR can: if the index contains first key from each block, we don't have to read the block until this block actually makes it to the top of merging iterator's heap, so for short range scans we won't read any blocks from most of the sst files.
This PR does the deferred block loading inside value() call. This is not ideal: there's no good way to report an IO error from inside value(). As discussed with siying offline, it would probably be better to change InternalIterator's interface to explicitly fetch deferred value and get status. I'll do it in a separate PR.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/5289
Differential Revision: D15256423
Pulled By: al13n321
fbshipit-source-id: 750e4c39ce88e8d41662f701cf6275d9388ba46a
2019-06-25 05:50:35 +02:00
|
|
|
std::unique_ptr<InternalIterator> meta_iter(metaindex_block.NewDataIterator(
|
2020-07-08 02:25:08 +02:00
|
|
|
BytewiseComparator(), kDisableGlobalSequenceNumber));
|
2018-06-12 21:46:48 +02:00
|
|
|
bool found_properties_block = true;
|
|
|
|
ASSERT_OK(SeekToPropertiesBlock(meta_iter.get(), &found_properties_block));
|
|
|
|
ASSERT_TRUE(found_properties_block);
|
|
|
|
|
|
|
|
// -- Read properties block
|
|
|
|
Slice v = meta_iter->value();
|
|
|
|
BlockHandle properties_handle;
|
|
|
|
ASSERT_OK(properties_handle.DecodeFrom(&v));
|
|
|
|
BlockContents properties_contents;
|
|
|
|
|
2019-06-19 04:00:03 +02:00
|
|
|
BlockFetchHelper(properties_handle, BlockType::kProperties,
|
2019-06-19 23:07:36 +02:00
|
|
|
&properties_contents);
|
2020-02-26 00:29:17 +01:00
|
|
|
Block properties_block(std::move(properties_contents));
|
2018-06-12 21:46:48 +02:00
|
|
|
|
2019-09-09 20:22:28 +02:00
|
|
|
ASSERT_EQ(properties_block.NumRestarts(), 1u);
|
2018-06-12 21:46:48 +02:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2018-07-20 18:00:33 +02:00
|
|
|
TEST_P(BlockBasedTableTest, PropertiesMetaBlockLast) {
|
|
|
|
// The properties meta-block should come at the end since we always need to
|
|
|
|
// read it when opening a file, unlike index/filter/other meta-blocks, which
|
|
|
|
// are sometimes read depending on the user's configuration. This ordering
|
|
|
|
// allows us to do a small readahead on the end of the file to read properties
|
|
|
|
// and meta-index blocks with one I/O.
|
|
|
|
TableConstructor c(BytewiseComparator(), true /* convert_to_internal_key_ */);
|
|
|
|
c.Add("a1", "val1");
|
|
|
|
c.Add("b2", "val2");
|
|
|
|
c.Add("c3", "val3");
|
|
|
|
c.Add("d4", "val4");
|
|
|
|
c.Add("e5", "val5");
|
|
|
|
c.Add("f6", "val6");
|
|
|
|
c.Add("g7", "val7");
|
|
|
|
c.Add("h8", "val8");
|
|
|
|
c.Add("j9", "val9");
|
|
|
|
|
|
|
|
// write an SST file
|
|
|
|
Options options;
|
|
|
|
BlockBasedTableOptions table_options = GetBlockBasedTableOptions();
|
|
|
|
table_options.filter_policy.reset(NewBloomFilterPolicy(
|
|
|
|
8 /* bits_per_key */, false /* use_block_based_filter */));
|
|
|
|
options.table_factory.reset(NewBlockBasedTableFactory(table_options));
|
|
|
|
ImmutableCFOptions ioptions(options);
|
|
|
|
MutableCFOptions moptions(options);
|
|
|
|
std::vector<std::string> keys;
|
|
|
|
stl_wrappers::KVMap kvmap;
|
|
|
|
c.Finish(options, ioptions, moptions, table_options,
|
|
|
|
GetPlainInternalComparator(options.comparator), &keys, &kvmap);
|
|
|
|
|
|
|
|
// get file reader
|
|
|
|
test::StringSink* table_sink = c.TEST_GetSink();
|
2021-01-05 00:59:52 +01:00
|
|
|
std::unique_ptr<FSRandomAccessFile> source(new test::StringSource(
|
|
|
|
table_sink->contents(), 0 /* unique_id */, false /* allow_mmap_reads */));
|
|
|
|
|
|
|
|
std::unique_ptr<RandomAccessFileReader> table_reader(
|
|
|
|
new RandomAccessFileReader(std::move(source), "test"));
|
2018-07-20 18:00:33 +02:00
|
|
|
size_t table_size = table_sink->contents().size();
|
|
|
|
|
|
|
|
// read footer
|
|
|
|
Footer footer;
|
2020-06-29 23:51:57 +02:00
|
|
|
IOOptions opts;
|
|
|
|
ASSERT_OK(ReadFooterFromFile(opts, table_reader.get(),
|
2018-07-20 18:00:33 +02:00
|
|
|
nullptr /* prefetch_buffer */, table_size,
|
|
|
|
&footer, kBlockBasedTableMagicNumber));
|
|
|
|
|
|
|
|
// read metaindex
|
|
|
|
auto metaindex_handle = footer.metaindex_handle();
|
|
|
|
BlockContents metaindex_contents;
|
|
|
|
PersistentCacheOptions pcache_opts;
|
|
|
|
BlockFetcher block_fetcher(
|
|
|
|
table_reader.get(), nullptr /* prefetch_buffer */, footer, ReadOptions(),
|
|
|
|
metaindex_handle, &metaindex_contents, ioptions, false /* decompress */,
|
2019-06-19 04:00:03 +02:00
|
|
|
false /*maybe_compressed*/, BlockType::kMetaIndex,
|
|
|
|
UncompressionDict::GetEmptyDict(), pcache_opts,
|
|
|
|
nullptr /*memory_allocator*/);
|
2018-07-20 18:00:33 +02:00
|
|
|
ASSERT_OK(block_fetcher.ReadBlockContents());
|
2020-02-26 00:29:17 +01:00
|
|
|
Block metaindex_block(std::move(metaindex_contents));
|
2018-07-20 18:00:33 +02:00
|
|
|
|
|
|
|
// verify properties block comes last
|
|
|
|
std::unique_ptr<InternalIterator> metaindex_iter{
|
2020-07-08 02:25:08 +02:00
|
|
|
metaindex_block.NewDataIterator(options.comparator,
|
2020-02-26 00:29:17 +01:00
|
|
|
kDisableGlobalSequenceNumber)};
|
2018-07-20 18:00:33 +02:00
|
|
|
uint64_t max_offset = 0;
|
|
|
|
std::string key_at_max_offset;
|
|
|
|
for (metaindex_iter->SeekToFirst(); metaindex_iter->Valid();
|
|
|
|
metaindex_iter->Next()) {
|
|
|
|
BlockHandle handle;
|
|
|
|
Slice value = metaindex_iter->value();
|
|
|
|
ASSERT_OK(handle.DecodeFrom(&value));
|
|
|
|
if (handle.offset() > max_offset) {
|
|
|
|
max_offset = handle.offset();
|
|
|
|
key_at_max_offset = metaindex_iter->key().ToString();
|
|
|
|
}
|
|
|
|
}
|
|
|
|
ASSERT_EQ(kPropertiesBlock, key_at_max_offset);
|
|
|
|
// index handle is stored in footer rather than metaindex block, so need
|
|
|
|
// separate logic to verify it comes before properties block.
|
|
|
|
ASSERT_GT(max_offset, footer.index_handle().offset());
|
|
|
|
c.ResetTableReader();
|
|
|
|
}
|
|
|
|
|
2018-06-05 04:59:44 +02:00
|
|
|
TEST_P(BlockBasedTableTest, BadOptions) {
|
2020-02-20 21:07:53 +01:00
|
|
|
ROCKSDB_NAMESPACE::Options options;
|
2018-03-27 05:14:24 +02:00
|
|
|
options.compression = kNoCompression;
|
2018-06-05 04:59:44 +02:00
|
|
|
BlockBasedTableOptions bbto = GetBlockBasedTableOptions();
|
2018-03-27 05:14:24 +02:00
|
|
|
bbto.block_size = 4000;
|
|
|
|
bbto.block_align = true;
|
|
|
|
|
2018-06-08 21:53:23 +02:00
|
|
|
const std::string kDBPath =
|
2018-07-14 02:18:39 +02:00
|
|
|
test::PerThreadDBPath("block_based_table_bad_options_test");
|
2018-03-27 05:14:24 +02:00
|
|
|
options.table_factory.reset(NewBlockBasedTableFactory(bbto));
|
2020-12-24 01:54:05 +01:00
|
|
|
ASSERT_OK(DestroyDB(kDBPath, options));
|
2020-02-20 21:07:53 +01:00
|
|
|
ROCKSDB_NAMESPACE::DB* db;
|
|
|
|
ASSERT_NOK(ROCKSDB_NAMESPACE::DB::Open(options, kDBPath, &db));
|
2018-03-27 05:14:24 +02:00
|
|
|
|
|
|
|
bbto.block_size = 4096;
|
|
|
|
options.compression = kSnappyCompression;
|
|
|
|
options.table_factory.reset(NewBlockBasedTableFactory(bbto));
|
2020-02-20 21:07:53 +01:00
|
|
|
ASSERT_NOK(ROCKSDB_NAMESPACE::DB::Open(options, kDBPath, &db));
|
2018-03-27 05:14:24 +02:00
|
|
|
}
|
|
|
|
|
2018-07-20 23:31:27 +02:00
|
|
|
TEST_F(BBTTailPrefetchTest, TestTailPrefetchStats) {
|
|
|
|
TailPrefetchStats tpstats;
|
|
|
|
ASSERT_EQ(0, tpstats.GetSuggestedPrefetchSize());
|
|
|
|
tpstats.RecordEffectiveSize(size_t{1000});
|
|
|
|
tpstats.RecordEffectiveSize(size_t{1005});
|
|
|
|
tpstats.RecordEffectiveSize(size_t{1002});
|
|
|
|
ASSERT_EQ(1005, tpstats.GetSuggestedPrefetchSize());
|
|
|
|
|
|
|
|
// One single super large value shouldn't influence much
|
|
|
|
tpstats.RecordEffectiveSize(size_t{1002000});
|
|
|
|
tpstats.RecordEffectiveSize(size_t{999});
|
|
|
|
ASSERT_LE(1005, tpstats.GetSuggestedPrefetchSize());
|
|
|
|
ASSERT_GT(1200, tpstats.GetSuggestedPrefetchSize());
|
|
|
|
|
|
|
|
// Only history of 32 is kept
|
|
|
|
for (int i = 0; i < 32; i++) {
|
|
|
|
tpstats.RecordEffectiveSize(size_t{100});
|
|
|
|
}
|
|
|
|
ASSERT_EQ(100, tpstats.GetSuggestedPrefetchSize());
|
|
|
|
|
|
|
|
// 16 large values and 16 small values. The result should be closer
|
|
|
|
// to the small value as the algorithm.
|
|
|
|
for (int i = 0; i < 16; i++) {
|
|
|
|
tpstats.RecordEffectiveSize(size_t{1000});
|
|
|
|
}
|
|
|
|
tpstats.RecordEffectiveSize(size_t{10});
|
|
|
|
tpstats.RecordEffectiveSize(size_t{20});
|
|
|
|
for (int i = 0; i < 6; i++) {
|
|
|
|
tpstats.RecordEffectiveSize(size_t{100});
|
|
|
|
}
|
|
|
|
ASSERT_LE(80, tpstats.GetSuggestedPrefetchSize());
|
|
|
|
ASSERT_GT(200, tpstats.GetSuggestedPrefetchSize());
|
|
|
|
}
|
|
|
|
|
|
|
|
TEST_F(BBTTailPrefetchTest, FilePrefetchBufferMinOffset) {
|
|
|
|
TailPrefetchStats tpstats;
|
|
|
|
FilePrefetchBuffer buffer(nullptr, 0, 0, false, true);
|
2020-06-29 23:51:57 +02:00
|
|
|
IOOptions opts;
|
2020-12-30 18:24:04 +01:00
|
|
|
buffer.TryReadFromCache(opts, 500, 10, nullptr, nullptr);
|
|
|
|
buffer.TryReadFromCache(opts, 480, 10, nullptr, nullptr);
|
|
|
|
buffer.TryReadFromCache(opts, 490, 10, nullptr, nullptr);
|
2018-07-20 23:31:27 +02:00
|
|
|
ASSERT_EQ(480, buffer.min_offset_read());
|
|
|
|
}
|
|
|
|
|
2018-08-15 23:27:47 +02:00
|
|
|
TEST_P(BlockBasedTableTest, DataBlockHashIndex) {
|
|
|
|
const int kNumKeys = 500;
|
|
|
|
const int kKeySize = 8;
|
|
|
|
const int kValSize = 40;
|
|
|
|
|
|
|
|
BlockBasedTableOptions table_options = GetBlockBasedTableOptions();
|
|
|
|
table_options.data_block_index_type =
|
2018-08-17 03:29:13 +02:00
|
|
|
BlockBasedTableOptions::kDataBlockBinaryAndHash;
|
2018-08-15 23:27:47 +02:00
|
|
|
|
|
|
|
Options options;
|
|
|
|
options.comparator = BytewiseComparator();
|
|
|
|
|
|
|
|
options.table_factory.reset(new BlockBasedTableFactory(table_options));
|
|
|
|
|
|
|
|
TableConstructor c(options.comparator);
|
|
|
|
|
|
|
|
static Random rnd(1048);
|
|
|
|
for (int i = 0; i < kNumKeys; i++) {
|
|
|
|
// padding one "0" to mark existent keys.
|
2020-07-09 23:33:42 +02:00
|
|
|
std::string random_key(rnd.RandomString(kKeySize - 1) + "1");
|
2018-08-15 23:27:47 +02:00
|
|
|
InternalKey k(random_key, 0, kTypeValue);
|
2020-07-09 23:33:42 +02:00
|
|
|
c.Add(k.Encode().ToString(), rnd.RandomString(kValSize));
|
2018-08-15 23:27:47 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
std::vector<std::string> keys;
|
|
|
|
stl_wrappers::KVMap kvmap;
|
|
|
|
const ImmutableCFOptions ioptions(options);
|
|
|
|
const MutableCFOptions moptions(options);
|
|
|
|
const InternalKeyComparator internal_comparator(options.comparator);
|
|
|
|
c.Finish(options, ioptions, moptions, table_options, internal_comparator,
|
|
|
|
&keys, &kvmap);
|
|
|
|
|
|
|
|
auto reader = c.GetTableReader();
|
|
|
|
|
|
|
|
std::unique_ptr<InternalIterator> seek_iter;
|
2020-08-04 00:21:56 +02:00
|
|
|
ReadOptions read_options;
|
2019-06-20 23:28:22 +02:00
|
|
|
seek_iter.reset(reader->NewIterator(
|
2020-08-04 00:21:56 +02:00
|
|
|
read_options, moptions.prefix_extractor.get(), /*arena=*/nullptr,
|
2019-06-20 23:28:22 +02:00
|
|
|
/*skip_filters=*/false, TableReaderCaller::kUncategorized));
|
2018-08-15 23:27:47 +02:00
|
|
|
for (int i = 0; i < 2; ++i) {
|
|
|
|
ReadOptions ro;
|
|
|
|
// for every kv, we seek using two method: Get() and Seek()
|
|
|
|
// Get() will use the SuffixIndexHash in Block. For non-existent key it
|
|
|
|
// will invalidate the iterator
|
|
|
|
// Seek() will use the default BinarySeek() in Block. So for non-existent
|
|
|
|
// key it will land at the closest key that is large than target.
|
|
|
|
|
|
|
|
// Search for existent keys
|
|
|
|
for (auto& kv : kvmap) {
|
|
|
|
if (i == 0) {
|
|
|
|
// Search using Seek()
|
|
|
|
seek_iter->Seek(kv.first);
|
|
|
|
ASSERT_OK(seek_iter->status());
|
|
|
|
ASSERT_TRUE(seek_iter->Valid());
|
|
|
|
ASSERT_EQ(seek_iter->key(), kv.first);
|
|
|
|
ASSERT_EQ(seek_iter->value(), kv.second);
|
|
|
|
} else {
|
|
|
|
// Search using Get()
|
|
|
|
PinnableSlice value;
|
|
|
|
std::string user_key = ExtractUserKey(kv.first).ToString();
|
|
|
|
GetContext get_context(options.comparator, nullptr, nullptr, nullptr,
|
|
|
|
GetContext::kNotFound, user_key, &value, nullptr,
|
New API to get all merge operands for a Key (#5604)
Summary:
This is a new API added to db.h to allow for fetching all merge operands associated with a Key. The main motivation for this API is to support use cases where doing a full online merge is not necessary as it is performance sensitive. Example use-cases:
1. Update subset of columns and read subset of columns -
Imagine a SQL Table, a row is encoded as a K/V pair (as it is done in MyRocks). If there are many columns and users only updated one of them, we can use merge operator to reduce write amplification. While users only read one or two columns in the read query, this feature can avoid a full merging of the whole row, and save some CPU.
2. Updating very few attributes in a value which is a JSON-like document -
Updating one attribute can be done efficiently using merge operator, while reading back one attribute can be done more efficiently if we don't need to do a full merge.
----------------------------------------------------------------------------------------------------
API :
Status GetMergeOperands(
const ReadOptions& options, ColumnFamilyHandle* column_family,
const Slice& key, PinnableSlice* merge_operands,
GetMergeOperandsOptions* get_merge_operands_options,
int* number_of_operands)
Example usage :
int size = 100;
int number_of_operands = 0;
std::vector<PinnableSlice> values(size);
GetMergeOperandsOptions merge_operands_info;
db_->GetMergeOperands(ReadOptions(), db_->DefaultColumnFamily(), "k1", values.data(), merge_operands_info, &number_of_operands);
Description :
Returns all the merge operands corresponding to the key. If the number of merge operands in DB is greater than merge_operands_options.expected_max_number_of_operands no merge operands are returned and status is Incomplete. Merge operands returned are in the order of insertion.
merge_operands-> Points to an array of at-least merge_operands_options.expected_max_number_of_operands and the caller is responsible for allocating it. If the status returned is Incomplete then number_of_operands will contain the total number of merge operands found in DB for key.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/5604
Test Plan:
Added unit test and perf test in db_bench that can be run using the command:
./db_bench -benchmarks=getmergeoperands --merge_operator=sortlist
Differential Revision: D16657366
Pulled By: vjnadimpalli
fbshipit-source-id: 0faadd752351745224ee12d4ae9ef3cb529951bf
2019-08-06 23:22:34 +02:00
|
|
|
nullptr, true, nullptr, nullptr);
|
2018-08-15 23:27:47 +02:00
|
|
|
ASSERT_OK(reader->Get(ro, kv.first, &get_context,
|
|
|
|
moptions.prefix_extractor.get()));
|
|
|
|
ASSERT_EQ(get_context.State(), GetContext::kFound);
|
|
|
|
ASSERT_EQ(value, Slice(kv.second));
|
|
|
|
value.Reset();
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
// Search for non-existent keys
|
|
|
|
for (auto& kv : kvmap) {
|
|
|
|
std::string user_key = ExtractUserKey(kv.first).ToString();
|
2018-08-17 03:29:13 +02:00
|
|
|
user_key.back() = '0'; // make it non-existent key
|
2018-08-15 23:27:47 +02:00
|
|
|
InternalKey internal_key(user_key, 0, kTypeValue);
|
|
|
|
std::string encoded_key = internal_key.Encode().ToString();
|
2018-08-17 03:29:13 +02:00
|
|
|
if (i == 0) { // Search using Seek()
|
2018-08-15 23:27:47 +02:00
|
|
|
seek_iter->Seek(encoded_key);
|
|
|
|
ASSERT_OK(seek_iter->status());
|
2018-08-17 03:29:13 +02:00
|
|
|
if (seek_iter->Valid()) {
|
2018-08-15 23:27:47 +02:00
|
|
|
ASSERT_TRUE(BytewiseComparator()->Compare(
|
|
|
|
user_key, ExtractUserKey(seek_iter->key())) < 0);
|
|
|
|
}
|
2018-08-17 03:29:13 +02:00
|
|
|
} else { // Search using Get()
|
2018-08-15 23:27:47 +02:00
|
|
|
PinnableSlice value;
|
|
|
|
GetContext get_context(options.comparator, nullptr, nullptr, nullptr,
|
|
|
|
GetContext::kNotFound, user_key, &value, nullptr,
|
New API to get all merge operands for a Key (#5604)
Summary:
This is a new API added to db.h to allow for fetching all merge operands associated with a Key. The main motivation for this API is to support use cases where doing a full online merge is not necessary as it is performance sensitive. Example use-cases:
1. Update subset of columns and read subset of columns -
Imagine a SQL Table, a row is encoded as a K/V pair (as it is done in MyRocks). If there are many columns and users only updated one of them, we can use merge operator to reduce write amplification. While users only read one or two columns in the read query, this feature can avoid a full merging of the whole row, and save some CPU.
2. Updating very few attributes in a value which is a JSON-like document -
Updating one attribute can be done efficiently using merge operator, while reading back one attribute can be done more efficiently if we don't need to do a full merge.
----------------------------------------------------------------------------------------------------
API :
Status GetMergeOperands(
const ReadOptions& options, ColumnFamilyHandle* column_family,
const Slice& key, PinnableSlice* merge_operands,
GetMergeOperandsOptions* get_merge_operands_options,
int* number_of_operands)
Example usage :
int size = 100;
int number_of_operands = 0;
std::vector<PinnableSlice> values(size);
GetMergeOperandsOptions merge_operands_info;
db_->GetMergeOperands(ReadOptions(), db_->DefaultColumnFamily(), "k1", values.data(), merge_operands_info, &number_of_operands);
Description :
Returns all the merge operands corresponding to the key. If the number of merge operands in DB is greater than merge_operands_options.expected_max_number_of_operands no merge operands are returned and status is Incomplete. Merge operands returned are in the order of insertion.
merge_operands-> Points to an array of at-least merge_operands_options.expected_max_number_of_operands and the caller is responsible for allocating it. If the status returned is Incomplete then number_of_operands will contain the total number of merge operands found in DB for key.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/5604
Test Plan:
Added unit test and perf test in db_bench that can be run using the command:
./db_bench -benchmarks=getmergeoperands --merge_operator=sortlist
Differential Revision: D16657366
Pulled By: vjnadimpalli
fbshipit-source-id: 0faadd752351745224ee12d4ae9ef3cb529951bf
2019-08-06 23:22:34 +02:00
|
|
|
nullptr, true, nullptr, nullptr);
|
2018-08-15 23:27:47 +02:00
|
|
|
ASSERT_OK(reader->Get(ro, encoded_key, &get_context,
|
|
|
|
moptions.prefix_extractor.get()));
|
|
|
|
ASSERT_EQ(get_context.State(), GetContext::kNotFound);
|
|
|
|
value.Reset();
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2019-04-16 20:32:03 +02:00
|
|
|
// BlockBasedTableIterator should invalidate itself and return
|
|
|
|
// OutOfBound()=true immediately after Seek(), to allow LevelIterator
|
|
|
|
// filter out corresponding level.
|
|
|
|
TEST_P(BlockBasedTableTest, OutOfBoundOnSeek) {
|
|
|
|
TableConstructor c(BytewiseComparator(), true /*convert_to_internal_key*/);
|
|
|
|
c.Add("foo", "v1");
|
|
|
|
std::vector<std::string> keys;
|
|
|
|
stl_wrappers::KVMap kvmap;
|
|
|
|
Options options;
|
|
|
|
BlockBasedTableOptions table_opt(GetBlockBasedTableOptions());
|
|
|
|
options.table_factory.reset(NewBlockBasedTableFactory(table_opt));
|
|
|
|
const ImmutableCFOptions ioptions(options);
|
|
|
|
const MutableCFOptions moptions(options);
|
|
|
|
c.Finish(options, ioptions, moptions, table_opt,
|
|
|
|
GetPlainInternalComparator(BytewiseComparator()), &keys, &kvmap);
|
|
|
|
auto* reader = c.GetTableReader();
|
|
|
|
ReadOptions read_opt;
|
|
|
|
std::string upper_bound = "bar";
|
|
|
|
Slice upper_bound_slice(upper_bound);
|
|
|
|
read_opt.iterate_upper_bound = &upper_bound_slice;
|
|
|
|
std::unique_ptr<InternalIterator> iter;
|
2019-06-20 23:28:22 +02:00
|
|
|
iter.reset(new KeyConvertingIterator(reader->NewIterator(
|
|
|
|
read_opt, /*prefix_extractor=*/nullptr, /*arena=*/nullptr,
|
|
|
|
/*skip_filters=*/false, TableReaderCaller::kUncategorized)));
|
2019-04-16 20:32:03 +02:00
|
|
|
iter->SeekToFirst();
|
|
|
|
ASSERT_FALSE(iter->Valid());
|
2020-10-08 20:11:52 +02:00
|
|
|
ASSERT_OK(iter->status());
|
2020-08-05 19:42:56 +02:00
|
|
|
ASSERT_TRUE(iter->UpperBoundCheckResult() == IterBoundCheck::kOutOfBound);
|
2019-06-20 23:28:22 +02:00
|
|
|
iter.reset(new KeyConvertingIterator(reader->NewIterator(
|
|
|
|
read_opt, /*prefix_extractor=*/nullptr, /*arena=*/nullptr,
|
|
|
|
/*skip_filters=*/false, TableReaderCaller::kUncategorized)));
|
2019-04-16 20:32:03 +02:00
|
|
|
iter->Seek("foo");
|
|
|
|
ASSERT_FALSE(iter->Valid());
|
2020-10-08 20:11:52 +02:00
|
|
|
ASSERT_OK(iter->status());
|
2020-08-05 19:42:56 +02:00
|
|
|
ASSERT_TRUE(iter->UpperBoundCheckResult() == IterBoundCheck::kOutOfBound);
|
2019-04-16 20:32:03 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
// BlockBasedTableIterator should invalidate itself and return
|
|
|
|
// OutOfBound()=true after Next(), if it finds current index key is no smaller
|
|
|
|
// than upper bound, unless it is pointing to the last data block.
|
|
|
|
TEST_P(BlockBasedTableTest, OutOfBoundOnNext) {
|
|
|
|
TableConstructor c(BytewiseComparator(), true /*convert_to_internal_key*/);
|
|
|
|
c.Add("bar", "v");
|
|
|
|
c.Add("foo", "v");
|
|
|
|
std::vector<std::string> keys;
|
|
|
|
stl_wrappers::KVMap kvmap;
|
|
|
|
Options options;
|
|
|
|
BlockBasedTableOptions table_opt(GetBlockBasedTableOptions());
|
|
|
|
table_opt.flush_block_policy_factory =
|
|
|
|
std::make_shared<FlushBlockEveryKeyPolicyFactory>();
|
|
|
|
options.table_factory.reset(NewBlockBasedTableFactory(table_opt));
|
|
|
|
const ImmutableCFOptions ioptions(options);
|
|
|
|
const MutableCFOptions moptions(options);
|
|
|
|
c.Finish(options, ioptions, moptions, table_opt,
|
|
|
|
GetPlainInternalComparator(BytewiseComparator()), &keys, &kvmap);
|
|
|
|
auto* reader = c.GetTableReader();
|
|
|
|
ReadOptions read_opt;
|
|
|
|
std::string ub1 = "bar_after";
|
|
|
|
Slice ub_slice1(ub1);
|
|
|
|
read_opt.iterate_upper_bound = &ub_slice1;
|
|
|
|
std::unique_ptr<InternalIterator> iter;
|
2019-06-20 23:28:22 +02:00
|
|
|
iter.reset(new KeyConvertingIterator(reader->NewIterator(
|
|
|
|
read_opt, /*prefix_extractor=*/nullptr, /*arena=*/nullptr,
|
|
|
|
/*skip_filters=*/false, TableReaderCaller::kUncategorized)));
|
2019-04-16 20:32:03 +02:00
|
|
|
iter->Seek("bar");
|
|
|
|
ASSERT_TRUE(iter->Valid());
|
|
|
|
ASSERT_EQ("bar", iter->key());
|
|
|
|
iter->Next();
|
|
|
|
ASSERT_FALSE(iter->Valid());
|
2020-08-05 19:42:56 +02:00
|
|
|
ASSERT_TRUE(iter->UpperBoundCheckResult() == IterBoundCheck::kOutOfBound);
|
2019-04-16 20:32:03 +02:00
|
|
|
std::string ub2 = "foo_after";
|
|
|
|
Slice ub_slice2(ub2);
|
|
|
|
read_opt.iterate_upper_bound = &ub_slice2;
|
2019-06-20 23:28:22 +02:00
|
|
|
iter.reset(new KeyConvertingIterator(reader->NewIterator(
|
|
|
|
read_opt, /*prefix_extractor=*/nullptr, /*arena=*/nullptr,
|
|
|
|
/*skip_filters=*/false, TableReaderCaller::kUncategorized)));
|
2019-04-16 20:32:03 +02:00
|
|
|
iter->Seek("foo");
|
|
|
|
ASSERT_TRUE(iter->Valid());
|
|
|
|
ASSERT_EQ("foo", iter->key());
|
|
|
|
iter->Next();
|
|
|
|
ASSERT_FALSE(iter->Valid());
|
2020-08-05 19:42:56 +02:00
|
|
|
ASSERT_FALSE(iter->UpperBoundCheckResult() == IterBoundCheck::kOutOfBound);
|
2019-04-16 20:32:03 +02:00
|
|
|
}
|
|
|
|
|
2020-02-20 21:07:53 +01:00
|
|
|
} // namespace ROCKSDB_NAMESPACE
|
2011-03-18 23:37:00 +01:00
|
|
|
|
|
|
|
int main(int argc, char** argv) {
|
2015-03-17 22:08:00 +01:00
|
|
|
::testing::InitGoogleTest(&argc, argv);
|
|
|
|
return RUN_ALL_TESTS();
|
2011-03-18 23:37:00 +01:00
|
|
|
}
|