rocksdb/db/pre_release_callback.h
Maysam Yabandeh c292dc8540 WritePrepared: reduce prepared_mutex_ overhead (#5420)
Summary:
The patch reduces the contention over prepared_mutex_ using these techniques:
1) Move ::RemovePrepared() to be called from the commit callback when we have two write queues.
2) Use two separate mutex for PreparedHeap, one prepared_mutex_ needed for ::RemovePrepared, and one ::push_pop_mutex() needed for ::AddPrepared(). Given that we call ::AddPrepared only from the first write queue and ::RemovePrepared mostly from the 2nd, this will result into each the two write queues not competing with each other over a single mutex. ::RemovePrepared might occasionally need to acquire ::push_pop_mutex() if ::erase() ends up with calling ::pop()
3) Acquire ::push_pop_mutex() on the first callback of the write queue and release it on the last.
Pull Request resolved: https://github.com/facebook/rocksdb/pull/5420

Differential Revision: D15741985

Pulled By: maysamyabandeh

fbshipit-source-id: 84ce8016007e88bb6e10da5760ba1f0d26347735
2019-06-10 11:53:31 -07:00

39 lines
1.6 KiB
C++

// Copyright (c) 2011-present, Facebook, Inc. All rights reserved.
// This source code is licensed under both the GPLv2 (found in the
// COPYING file in the root directory) and Apache 2.0 License
// (found in the LICENSE.Apache file in the root directory).
#pragma once
#include "rocksdb/status.h"
namespace rocksdb {
class DB;
class PreReleaseCallback {
public:
virtual ~PreReleaseCallback() {}
// Will be called while on the write thread after the write to the WAL and
// before the write to memtable. This is useful if any operation needs to be
// done before the write gets visible to the readers, or if we want to reduce
// the overhead of locking by updating something sequentially while we are on
// the write thread. If the callback fails, this function returns a non-OK
// status, the sequence number will not be released, and same status will be
// propagated to all the writers in the write group.
// seq is the sequence number that is used for this write and will be
// released.
// is_mem_disabled is currently used for debugging purposes to assert that
// the callback is done from the right write queue.
// If non-zero, log_number indicates the WAL log to which we wrote.
// index >= 0 specifies the order of callback in the same write thread.
// total > index specifies the total number of callbacks in the same write
// thread. Together with index, could be used to reduce the redundant
// operations among the callbacks.
virtual Status Callback(SequenceNumber seq, bool is_mem_disabled,
uint64_t log_number, size_t index, size_t total) = 0;
};
} // namespace rocksdb