dolysis/rocksdb
3
0
Fork 0
mirror of https://github.com/facebook/rocksdb.git synced 2024-11-27 11:43:49 +00:00
Code Issues Packages Projects Releases Wiki Activity
rocksdb/db/db_merge_operand_test.cc
Andrew Kryczka 504fe4de80 Avoid allocations/copies for large GetMergeOperands() results (#10458) 
Summary:
This PR avoids allocations and copies for the result of `GetMergeOperands()` when the average operand size is at least 256 bytes and the total operands size is at least 32KB. The `GetMergeOperands()` already included `PinnableSlice` but was calling `PinSelf()` (i.e., allocating and copying) for each operand. When this optimization takes effect, we instead call `PinSlice()` to skip that allocation and copy. Resources are pinned in order for the `PinnableSlice` to point to valid memory even after `GetMergeOperands()` returns.

The pinned resources include a referenced `SuperVersion`, a `MergingContext`, and a `PinnedIteratorsManager`. They are bundled into a `GetMergeOperandsState`. We use `SharedCleanablePtr` to share that bundle among all `PinnableSlice`s populated by `GetMergeOperands()`. That way, the last `PinnableSlice` to be `Reset()` will cleanup the bundle, including unreferencing the `SuperVersion`.

Pull Request resolved: https://github.com/facebook/rocksdb/pull/10458

Test Plan:
- new DB level test
- measured benefit/regression in a number of memtable scenarios

Setup command:
```
$ ./db_bench -benchmarks=mergerandom -merge_operator=StringAppendOperator -num=$num -writes=16384 -key_size=16 -value_size=$value_sz -compression_type=none -write_buffer_size=1048576000
```

Benchmark command:
```
./db_bench -threads=$threads -use_existing_db=true -avoid_flush_during_recovery=true -write_buffer_size=1048576000 -benchmarks=readrandomoperands -merge_operator=StringAppendOperator -num=$num -duration=10
```

Worst regression is when a key has many tiny operands:

- Parameters: num=1 (implying 16384 operands per key), value_sz=8, threads=1
- `GetMergeOperands()` latency increases 682 micros -> 800 micros (+17%)

The regression disappears into the noise (<1% difference) if we remove the `Reset()` loop and the size counting loop. The former is arguably needed regardless of this PR as the convention in `Get()` and `MultiGet()` is to `Reset()` the input `PinnableSlice`s at the start. The latter could be optimized to count the size as we accumulate operands rather than after the fact.

Best improvement is when a key has large operands and high concurrency:

- Parameters: num=4 (implying 4096 operands per key), value_sz=2KB, threads=32
- `GetMergeOperands()` latency decreases 11492 micros -> 437 micros (-96%).

Reviewed By: cbi42

Differential Revision: D38336578

Pulled By: ajkr

fbshipit-source-id: 48146d127e04cb7f2d4d2939a2b9dff3aba18258
2022-08-04 00:42:13 -07:00

449 lines
16 KiB
C++
Raw Blame History

// Copyright (c) 2018-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 "db/db_test_util.h"
#include "port/stack_trace.h"
#include "rocksdb/perf_context.h"
#include "rocksdb/utilities/debug.h"
#include "table/block_based/block_builder.h"
#if !defined(ROCKSDB_LITE)
#include "test_util/sync_point.h"
#endif
#include "rocksdb/merge_operator.h"
#include "utilities/fault_injection_env.h"
#include "utilities/merge_operators.h"
#include "utilities/merge_operators/sortlist.h"
#include "utilities/merge_operators/string_append/stringappend2.h"
namespace ROCKSDB_NAMESPACE {
namespace {
class LimitedStringAppendMergeOp : public StringAppendTESTOperator {
public:
LimitedStringAppendMergeOp(int limit, char delim)
: StringAppendTESTOperator(delim), limit_(limit) {}
const char* Name() const override {
return "DBMergeOperatorTest::LimitedStringAppendMergeOp";
}
bool ShouldMerge(const std::vector<Slice>& operands) const override {
if (operands.size() > 0 && limit_ > 0 && operands.size() >= limit_) {
return true;
}
return false;
}
private:
size_t limit_ = 0;
};
} // namespace
class DBMergeOperandTest : public DBTestBase {
public:
DBMergeOperandTest()
: DBTestBase("db_merge_operand_test", /*env_do_fsync=*/true) {}
};
TEST_F(DBMergeOperandTest, CacheEvictedMergeOperandReadAfterFreeBug) {
// There was a bug of reading merge operands after they are mistakely freed
// in DB::GetMergeOperands, which is surfaced by cache full.
// See PR#9507 for more.
Options options;
options.create_if_missing = true;
options.merge_operator = MergeOperators::CreateStringAppendOperator();
options.env = env_;
BlockBasedTableOptions table_options;
// Small cache to simulate cache full
table_options.block_cache = NewLRUCache(1);
options.table_factory.reset(NewBlockBasedTableFactory(table_options));
Reopen(options);
int num_records = 4;
int number_of_operands = 0;
std::vector<PinnableSlice> values(num_records);
GetMergeOperandsOptions merge_operands_info;
merge_operands_info.expected_max_number_of_operands = num_records;
ASSERT_OK(Merge("k1", "v1"));
ASSERT_OK(Flush());
ASSERT_OK(Merge("k1", "v2"));
ASSERT_OK(Flush());
ASSERT_OK(Merge("k1", "v3"));
ASSERT_OK(Flush());
ASSERT_OK(Merge("k1", "v4"));
ASSERT_OK(db_->GetMergeOperands(ReadOptions(), db_->DefaultColumnFamily(),
"k1", values.data(), &merge_operands_info,
&number_of_operands));
ASSERT_EQ(number_of_operands, 4);
ASSERT_EQ(values[0].ToString(), "v1");
ASSERT_EQ(values[1].ToString(), "v2");
ASSERT_EQ(values[2].ToString(), "v3");
ASSERT_EQ(values[3].ToString(), "v4");
}
TEST_F(DBMergeOperandTest, FlushedMergeOperandReadAfterFreeBug) {
// Repro for a bug where a memtable containing a merge operand could be
// deleted before the merge operand was saved to the result.
auto options = CurrentOptions();
options.merge_operator = MergeOperators::CreateStringAppendOperator();
Reopen(options);
ASSERT_OK(Merge("key", "value"));
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->LoadDependency(
{{"DBImpl::GetImpl:PostMemTableGet:0",
"DBMergeOperandTest::FlushedMergeOperandReadAfterFreeBug:PreFlush"},
{"DBMergeOperandTest::FlushedMergeOperandReadAfterFreeBug:PostFlush",
"DBImpl::GetImpl:PostMemTableGet:1"}});
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->EnableProcessing();
auto flush_thread = port::Thread([&]() {
TEST_SYNC_POINT(
"DBMergeOperandTest::FlushedMergeOperandReadAfterFreeBug:PreFlush");
ASSERT_OK(Flush());
TEST_SYNC_POINT(
"DBMergeOperandTest::FlushedMergeOperandReadAfterFreeBug:PostFlush");
});
PinnableSlice value;
GetMergeOperandsOptions merge_operands_info;
merge_operands_info.expected_max_number_of_operands = 1;
int number_of_operands;
ASSERT_OK(db_->GetMergeOperands(ReadOptions(), db_->DefaultColumnFamily(),
"key", &value, &merge_operands_info,
&number_of_operands));
ASSERT_EQ(1, number_of_operands);
flush_thread.join();
}
TEST_F(DBMergeOperandTest, GetMergeOperandsBasic) {
Options options;
options.create_if_missing = true;
// Use only the latest two merge operands.
options.merge_operator = std::make_shared<LimitedStringAppendMergeOp>(2, ',');
options.env = env_;
Reopen(options);
int num_records = 4;
int number_of_operands = 0;
std::vector<PinnableSlice> values(num_records);
GetMergeOperandsOptions merge_operands_info;
merge_operands_info.expected_max_number_of_operands = num_records;
// k0 value in memtable
ASSERT_OK(Put("k0", "PutARock"));
ASSERT_OK(db_->GetMergeOperands(ReadOptions(), db_->DefaultColumnFamily(),
"k0", values.data(), &merge_operands_info,
&number_of_operands));
ASSERT_EQ(values[0], "PutARock");
// k0.1 value in SST
ASSERT_OK(Put("k0.1", "RockInSST"));
ASSERT_OK(Flush());
ASSERT_OK(db_->GetMergeOperands(ReadOptions(), db_->DefaultColumnFamily(),
"k0.1", values.data(), &merge_operands_info,
&number_of_operands));
ASSERT_EQ(values[0], "RockInSST");
// All k1 values are in memtable.
ASSERT_OK(Merge("k1", "a"));
ASSERT_OK(Put("k1", "x"));
ASSERT_OK(Merge("k1", "b"));
ASSERT_OK(Merge("k1", "c"));
ASSERT_OK(Merge("k1", "d"));
ASSERT_OK(db_->GetMergeOperands(ReadOptions(), db_->DefaultColumnFamily(),
"k1", values.data(), &merge_operands_info,
&number_of_operands));
ASSERT_EQ(values[0], "x");
ASSERT_EQ(values[1], "b");
ASSERT_EQ(values[2], "c");
ASSERT_EQ(values[3], "d");
// expected_max_number_of_operands is less than number of merge operands so
// status should be Incomplete.
merge_operands_info.expected_max_number_of_operands = num_records - 1;
Status status = db_->GetMergeOperands(
ReadOptions(), db_->DefaultColumnFamily(), "k1", values.data(),
&merge_operands_info, &number_of_operands);
ASSERT_EQ(status.IsIncomplete(), true);
merge_operands_info.expected_max_number_of_operands = num_records;
// All k1.1 values are in memtable.
ASSERT_OK(Merge("k1.1", "r"));
ASSERT_OK(Delete("k1.1"));
ASSERT_OK(Merge("k1.1", "c"));
ASSERT_OK(Merge("k1.1", "k"));
ASSERT_OK(Merge("k1.1", "s"));
ASSERT_OK(db_->GetMergeOperands(ReadOptions(), db_->DefaultColumnFamily(),
"k1.1", values.data(), &merge_operands_info,
&number_of_operands));
ASSERT_EQ(values[0], "c");
ASSERT_EQ(values[1], "k");
ASSERT_EQ(values[2], "s");
// All k2 values are flushed to L0 into a single file.
ASSERT_OK(Merge("k2", "q"));
ASSERT_OK(Merge("k2", "w"));
ASSERT_OK(Merge("k2", "e"));
ASSERT_OK(Merge("k2", "r"));
ASSERT_OK(Flush());
ASSERT_OK(db_->GetMergeOperands(ReadOptions(), db_->DefaultColumnFamily(),
"k2", values.data(), &merge_operands_info,
&number_of_operands));
ASSERT_EQ(values[0], "q");
ASSERT_EQ(values[1], "w");
ASSERT_EQ(values[2], "e");
ASSERT_EQ(values[3], "r");
// All k2.1 values are flushed to L0 into a single file.
ASSERT_OK(Merge("k2.1", "m"));
ASSERT_OK(Put("k2.1", "l"));
ASSERT_OK(Merge("k2.1", "n"));
ASSERT_OK(Merge("k2.1", "o"));
ASSERT_OK(Flush());
ASSERT_OK(db_->GetMergeOperands(ReadOptions(), db_->DefaultColumnFamily(),
"k2.1", values.data(), &merge_operands_info,
&number_of_operands));
ASSERT_EQ(values[0], "l,n,o");
// All k2.2 values are flushed to L0 into a single file.
ASSERT_OK(Merge("k2.2", "g"));
ASSERT_OK(Delete("k2.2"));
ASSERT_OK(Merge("k2.2", "o"));
ASSERT_OK(Merge("k2.2", "t"));
ASSERT_OK(Flush());
ASSERT_OK(db_->GetMergeOperands(ReadOptions(), db_->DefaultColumnFamily(),
"k2.2", values.data(), &merge_operands_info,
&number_of_operands));
ASSERT_EQ(values[0], "o,t");
// Do some compaction that will make the following tests more predictable
// Slice start("PutARock");
// Slice end("t");
ASSERT_OK(db_->CompactRange(CompactRangeOptions(), nullptr, nullptr));
// All k3 values are flushed and are in different files.
ASSERT_OK(Merge("k3", "ab"));
ASSERT_OK(Flush());
ASSERT_OK(Merge("k3", "bc"));
ASSERT_OK(Flush());
ASSERT_OK(Merge("k3", "cd"));
ASSERT_OK(Flush());
ASSERT_OK(Merge("k3", "de"));
ASSERT_OK(db_->GetMergeOperands(ReadOptions(), db_->DefaultColumnFamily(),
"k3", values.data(), &merge_operands_info,
&number_of_operands));
ASSERT_EQ(values[0], "ab");
ASSERT_EQ(values[1], "bc");
ASSERT_EQ(values[2], "cd");
ASSERT_EQ(values[3], "de");
// All k3.1 values are flushed and are in different files.
ASSERT_OK(Merge("k3.1", "ab"));
ASSERT_OK(Flush());
ASSERT_OK(Put("k3.1", "bc"));
ASSERT_OK(Flush());
ASSERT_OK(Merge("k3.1", "cd"));
ASSERT_OK(Flush());
ASSERT_OK(Merge("k3.1", "de"));
ASSERT_OK(db_->GetMergeOperands(ReadOptions(), db_->DefaultColumnFamily(),
"k3.1", values.data(), &merge_operands_info,
&number_of_operands));
ASSERT_EQ(values[0], "bc");
ASSERT_EQ(values[1], "cd");
ASSERT_EQ(values[2], "de");
// All k3.2 values are flushed and are in different files.
ASSERT_OK(Merge("k3.2", "ab"));
ASSERT_OK(Flush());
ASSERT_OK(Delete("k3.2"));
ASSERT_OK(Flush());
ASSERT_OK(Merge("k3.2", "cd"));
ASSERT_OK(Flush());
ASSERT_OK(Merge("k3.2", "de"));
ASSERT_OK(db_->GetMergeOperands(ReadOptions(), db_->DefaultColumnFamily(),
"k3.2", values.data(), &merge_operands_info,
&number_of_operands));
ASSERT_EQ(values[0], "cd");
ASSERT_EQ(values[1], "de");
// All K4 values are in different levels
ASSERT_OK(Merge("k4", "ba"));
ASSERT_OK(Flush());
MoveFilesToLevel(4);
ASSERT_OK(Merge("k4", "cb"));
ASSERT_OK(Flush());
MoveFilesToLevel(3);
ASSERT_OK(Merge("k4", "dc"));
ASSERT_OK(Flush());
MoveFilesToLevel(1);
ASSERT_OK(Merge("k4", "ed"));
ASSERT_OK(db_->GetMergeOperands(ReadOptions(), db_->DefaultColumnFamily(),
"k4", values.data(), &merge_operands_info,
&number_of_operands));
ASSERT_EQ(values[0], "ba");
ASSERT_EQ(values[1], "cb");
ASSERT_EQ(values[2], "dc");
ASSERT_EQ(values[3], "ed");
// First 3 k5 values are in SST and next 4 k5 values are in Immutable
// Memtable
ASSERT_OK(Merge("k5", "who"));
ASSERT_OK(Merge("k5", "am"));
ASSERT_OK(Merge("k5", "i"));
ASSERT_OK(Flush());
ASSERT_OK(Put("k5", "remember"));
ASSERT_OK(Merge("k5", "i"));
ASSERT_OK(Merge("k5", "am"));
ASSERT_OK(Merge("k5", "rocks"));
ASSERT_OK(dbfull()->TEST_SwitchMemtable());
ASSERT_OK(db_->GetMergeOperands(ReadOptions(), db_->DefaultColumnFamily(),
"k5", values.data(), &merge_operands_info,
&number_of_operands));
ASSERT_EQ(values[0], "remember");
ASSERT_EQ(values[1], "i");
ASSERT_EQ(values[2], "am");
}
TEST_F(DBMergeOperandTest, BlobDBGetMergeOperandsBasic) {
Options options;
options.create_if_missing = true;
options.enable_blob_files = true;
options.min_blob_size = 0;
// Use only the latest two merge operands.
options.merge_operator = std::make_shared<LimitedStringAppendMergeOp>(2, ',');
options.env = env_;
Reopen(options);
int num_records = 4;
int number_of_operands = 0;
std::vector<PinnableSlice> values(num_records);
GetMergeOperandsOptions merge_operands_info;
merge_operands_info.expected_max_number_of_operands = num_records;
// All k1 values are in memtable.
ASSERT_OK(Put("k1", "x"));
ASSERT_OK(Merge("k1", "b"));
ASSERT_OK(Merge("k1", "c"));
ASSERT_OK(Merge("k1", "d"));
ASSERT_OK(db_->GetMergeOperands(ReadOptions(), db_->DefaultColumnFamily(),
"k1", values.data(), &merge_operands_info,
&number_of_operands));
ASSERT_EQ(values[0], "x");
ASSERT_EQ(values[1], "b");
ASSERT_EQ(values[2], "c");
ASSERT_EQ(values[3], "d");
// expected_max_number_of_operands is less than number of merge operands so
// status should be Incomplete.
merge_operands_info.expected_max_number_of_operands = num_records - 1;
Status status = db_->GetMergeOperands(
ReadOptions(), db_->DefaultColumnFamily(), "k1", values.data(),
&merge_operands_info, &number_of_operands);
ASSERT_EQ(status.IsIncomplete(), true);
merge_operands_info.expected_max_number_of_operands = num_records;
// All k2 values are flushed to L0 into a single file.
ASSERT_OK(Put("k2", "q"));
ASSERT_OK(Merge("k2", "w"));
ASSERT_OK(Merge("k2", "e"));
ASSERT_OK(Merge("k2", "r"));
ASSERT_OK(Flush());
ASSERT_OK(db_->GetMergeOperands(ReadOptions(), db_->DefaultColumnFamily(),
"k2", values.data(), &merge_operands_info,
&number_of_operands));
ASSERT_EQ(values[0], "q,w,e,r");
// Do some compaction that will make the following tests more predictable
ASSERT_OK(db_->CompactRange(CompactRangeOptions(), nullptr, nullptr));
// All k3 values are flushed and are in different files.
ASSERT_OK(Put("k3", "ab"));
ASSERT_OK(Flush());
ASSERT_OK(Merge("k3", "bc"));
ASSERT_OK(Flush());
ASSERT_OK(Merge("k3", "cd"));
ASSERT_OK(Flush());
ASSERT_OK(Merge("k3", "de"));
ASSERT_OK(db_->GetMergeOperands(ReadOptions(), db_->DefaultColumnFamily(),
"k3", values.data(), &merge_operands_info,
&number_of_operands));
ASSERT_EQ(values[0], "ab");
ASSERT_EQ(values[1], "bc");
ASSERT_EQ(values[2], "cd");
ASSERT_EQ(values[3], "de");
// All K4 values are in different levels
ASSERT_OK(Put("k4", "ba"));
ASSERT_OK(Flush());
MoveFilesToLevel(4);
ASSERT_OK(Merge("k4", "cb"));
ASSERT_OK(Flush());
MoveFilesToLevel(3);
ASSERT_OK(Merge("k4", "dc"));
ASSERT_OK(Flush());
MoveFilesToLevel(1);
ASSERT_OK(Merge("k4", "ed"));
ASSERT_OK(db_->GetMergeOperands(ReadOptions(), db_->DefaultColumnFamily(),
"k4", values.data(), &merge_operands_info,
&number_of_operands));
ASSERT_EQ(values[0], "ba");
ASSERT_EQ(values[1], "cb");
ASSERT_EQ(values[2], "dc");
ASSERT_EQ(values[3], "ed");
}
TEST_F(DBMergeOperandTest, GetMergeOperandsLargeResultOptimization) {
// These constants are chosen to trigger the large result optimization
// (pinning a bundle of `DBImpl` resources).
const int kNumOperands = 1024;
const int kOperandLen = 1024;
Options options;
options.create_if_missing = true;
options.merge_operator = MergeOperators::CreateStringAppendOperator();
DestroyAndReopen(options);
Random rnd(301);
std::vector<std::string> expected_merge_operands;
expected_merge_operands.reserve(kNumOperands);
for (int i = 0; i < kNumOperands; ++i) {
expected_merge_operands.emplace_back(rnd.RandomString(kOperandLen));
ASSERT_OK(Merge("key", expected_merge_operands.back()));
}
std::vector<PinnableSlice> merge_operands(kNumOperands);
GetMergeOperandsOptions merge_operands_info;
merge_operands_info.expected_max_number_of_operands = kNumOperands;
int num_merge_operands = 0;
ASSERT_OK(db_->GetMergeOperands(ReadOptions(), db_->DefaultColumnFamily(),
"key", merge_operands.data(),
&merge_operands_info, &num_merge_operands));
ASSERT_EQ(num_merge_operands, kNumOperands);
// Ensures the large result optimization was used.
for (int i = 0; i < kNumOperands; ++i) {
ASSERT_TRUE(merge_operands[i].IsPinned());
}
// Add a Flush() to change the `SuperVersion` to challenge the resource
// pinning.
ASSERT_OK(Flush());
for (int i = 0; i < kNumOperands; ++i) {
ASSERT_EQ(expected_merge_operands[i], merge_operands[i]);
}
}
} // namespace ROCKSDB_NAMESPACE
int main(int argc, char** argv) {
ROCKSDB_NAMESPACE::port::InstallStackTraceHandler();
::testing::InitGoogleTest(&argc, argv);
return RUN_ALL_TESTS();
}
Reference in a new issue View git blame Copy permalink