rocksdb/table/get_context.cc
Sagar Vemuri 93c2b91740 Introduce conditional merge-operator invocation in point lookups
Summary:
For every merge operand encountered for a key in the read path we now have the ability to decide whether to look further (to retrieve more merge operands for the key) or stop and invoke the merge operator to return the value. The user needs to override `ShouldMerge()` method with a condition to terminate search when true to avail this facility.

This has a couple of advantages:
1. It helps in limiting the number of merge operands that are looked at to compute a value as part of a user Get operation.
2. It allows to peek at a merge key-value to see if further merge operands need to look at.

Example: Limiting the number of merge operands that are looked at: Lets say you have 10 merge operands for a key spread over various levels. If you only want RocksDB to look at the latest two merge operands instead of all 10 to compute the value, it is now possible with this PR. You can set the condition in `ShouldMerge()` to return true when the size of the operand list is 2. Look at the example implementation in the unit test. Without this PR, a Get might look at all the 10 merge operands in different levels before invoking the merge-operator.

Added a new unit test.
Made sure that there is no perf regression by running benchmarks.

Command line to Load data:
```
TEST_TMPDIR=/dev/shm ./db_bench --benchmarks="mergerandom" --merge_operator="uint64add" --num=10000000
...
mergerandom  :      12.861 micros/op 77757 ops/sec;    8.6 MB/s ( updates:10000000)
```

**ReadRandomMergeRandom bechmark results:**
Command line:
```
TEST_TMPDIR=/dev/shm ./db_bench --benchmarks="readrandommergerandom" --merge_operator="uint64add" --num=10000000
```

Base -- Without this code change (on commit fc7476b):
```
readrandommergerandom :      38.586 micros/op 25916 ops/sec; (reads:3001599 merges:6998401 total:10000000 hits:842235 maxlength:8)
```

With this code change:
```
readrandommergerandom :      38.653 micros/op 25870 ops/sec; (reads:3001599 merges:6998401 total:10000000 hits:842235 maxlength:8)
```
Closes https://github.com/facebook/rocksdb/pull/2923

Differential Revision: D5898239

Pulled By: sagar0

fbshipit-source-id: daefa325019f77968639a75c851d46352c2303ef
2017-09-28 15:58:49 -07:00

226 lines
7.4 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).
#include "table/get_context.h"
#include "db/merge_helper.h"
#include "db/pinned_iterators_manager.h"
#include "db/read_callback.h"
#include "monitoring/file_read_sample.h"
#include "monitoring/perf_context_imp.h"
#include "monitoring/statistics.h"
#include "rocksdb/env.h"
#include "rocksdb/merge_operator.h"
#include "rocksdb/statistics.h"
namespace rocksdb {
namespace {
void appendToReplayLog(std::string* replay_log, ValueType type, Slice value) {
#ifndef ROCKSDB_LITE
if (replay_log) {
if (replay_log->empty()) {
// Optimization: in the common case of only one operation in the
// log, we allocate the exact amount of space needed.
replay_log->reserve(1 + VarintLength(value.size()) + value.size());
}
replay_log->push_back(type);
PutLengthPrefixedSlice(replay_log, value);
}
#endif // ROCKSDB_LITE
}
} // namespace
GetContext::GetContext(
const Comparator* ucmp, const MergeOperator* merge_operator, Logger* logger,
Statistics* statistics, GetState init_state, const Slice& user_key,
PinnableSlice* pinnable_val, bool* value_found, MergeContext* merge_context,
RangeDelAggregator* _range_del_agg, Env* env, SequenceNumber* seq,
PinnedIteratorsManager* _pinned_iters_mgr, ReadCallback* callback)
: ucmp_(ucmp),
merge_operator_(merge_operator),
logger_(logger),
statistics_(statistics),
state_(init_state),
user_key_(user_key),
pinnable_val_(pinnable_val),
value_found_(value_found),
merge_context_(merge_context),
range_del_agg_(_range_del_agg),
env_(env),
seq_(seq),
replay_log_(nullptr),
pinned_iters_mgr_(_pinned_iters_mgr),
callback_(callback) {
if (seq_) {
*seq_ = kMaxSequenceNumber;
}
sample_ = should_sample_file_read();
}
// Called from TableCache::Get and Table::Get when file/block in which
// key may exist are not there in TableCache/BlockCache respectively. In this
// case we can't guarantee that key does not exist and are not permitted to do
// IO to be certain.Set the status=kFound and value_found=false to let the
// caller know that key may exist but is not there in memory
void GetContext::MarkKeyMayExist() {
state_ = kFound;
if (value_found_ != nullptr) {
*value_found_ = false;
}
}
void GetContext::SaveValue(const Slice& value, SequenceNumber seq) {
assert(state_ == kNotFound);
appendToReplayLog(replay_log_, kTypeValue, value);
state_ = kFound;
if (LIKELY(pinnable_val_ != nullptr)) {
pinnable_val_->PinSelf(value);
}
}
bool GetContext::SaveValue(const ParsedInternalKey& parsed_key,
const Slice& value, Cleanable* value_pinner) {
assert((state_ != kMerge && parsed_key.type != kTypeMerge) ||
merge_context_ != nullptr);
if (ucmp_->Equal(parsed_key.user_key, user_key_)) {
// If the value is not in the snapshot, skip it
if (!CheckCallback(parsed_key.sequence)) {
return true; // to continue to the next seq
}
appendToReplayLog(replay_log_, parsed_key.type, value);
if (seq_ != nullptr) {
// Set the sequence number if it is uninitialized
if (*seq_ == kMaxSequenceNumber) {
*seq_ = parsed_key.sequence;
}
}
auto type = parsed_key.type;
// Key matches. Process it
if ((type == kTypeValue || type == kTypeMerge) &&
range_del_agg_ != nullptr && range_del_agg_->ShouldDelete(parsed_key)) {
type = kTypeRangeDeletion;
}
switch (type) {
case kTypeValue:
assert(state_ == kNotFound || state_ == kMerge);
if (kNotFound == state_) {
state_ = kFound;
if (LIKELY(pinnable_val_ != nullptr)) {
if (LIKELY(value_pinner != nullptr)) {
// If the backing resources for the value are provided, pin them
pinnable_val_->PinSlice(value, value_pinner);
} else {
// Otherwise copy the value
pinnable_val_->PinSelf(value);
}
}
} else if (kMerge == state_) {
assert(merge_operator_ != nullptr);
state_ = kFound;
if (LIKELY(pinnable_val_ != nullptr)) {
Status merge_status = MergeHelper::TimedFullMerge(
merge_operator_, user_key_, &value,
merge_context_->GetOperands(), pinnable_val_->GetSelf(),
logger_, statistics_, env_);
pinnable_val_->PinSelf();
if (!merge_status.ok()) {
state_ = kCorrupt;
}
}
}
return false;
case kTypeDeletion:
case kTypeSingleDeletion:
case kTypeRangeDeletion:
// TODO(noetzli): Verify correctness once merge of single-deletes
// is supported
assert(state_ == kNotFound || state_ == kMerge);
if (kNotFound == state_) {
state_ = kDeleted;
} else if (kMerge == state_) {
state_ = kFound;
if (LIKELY(pinnable_val_ != nullptr)) {
Status merge_status = MergeHelper::TimedFullMerge(
merge_operator_, user_key_, nullptr,
merge_context_->GetOperands(), pinnable_val_->GetSelf(),
logger_, statistics_, env_);
pinnable_val_->PinSelf();
if (!merge_status.ok()) {
state_ = kCorrupt;
}
}
}
return false;
case kTypeMerge:
assert(state_ == kNotFound || state_ == kMerge);
state_ = kMerge;
// value_pinner is not set from plain_table_reader.cc for example.
if (pinned_iters_mgr() && pinned_iters_mgr()->PinningEnabled() &&
value_pinner != nullptr) {
value_pinner->DelegateCleanupsTo(pinned_iters_mgr());
merge_context_->PushOperand(value, true /*value_pinned*/);
} else {
merge_context_->PushOperand(value, false);
}
if (merge_operator_ != nullptr &&
merge_operator_->ShouldMerge(merge_context_->GetOperands())) {
state_ = kFound;
if (LIKELY(pinnable_val_ != nullptr)) {
Status merge_status = MergeHelper::TimedFullMerge(
merge_operator_, user_key_, nullptr,
merge_context_->GetOperands(), pinnable_val_->GetSelf(),
logger_, statistics_, env_);
pinnable_val_->PinSelf();
if (!merge_status.ok()) {
state_ = kCorrupt;
}
}
return false;
}
return true;
default:
assert(false);
break;
}
}
// state_ could be Corrupt, merge or notfound
return false;
}
void replayGetContextLog(const Slice& replay_log, const Slice& user_key,
GetContext* get_context, Cleanable* value_pinner) {
#ifndef ROCKSDB_LITE
Slice s = replay_log;
while (s.size()) {
auto type = static_cast<ValueType>(*s.data());
s.remove_prefix(1);
Slice value;
bool ret = GetLengthPrefixedSlice(&s, &value);
assert(ret);
(void)ret;
// Since SequenceNumber is not stored and unknown, we will use
// kMaxSequenceNumber.
get_context->SaveValue(
ParsedInternalKey(user_key, kMaxSequenceNumber, type), value,
value_pinner);
}
#else // ROCKSDB_LITE
assert(false);
#endif // ROCKSDB_LITE
}
} // namespace rocksdb