rocksdb/db/merge_test.cc

889 lines
28 KiB
C++
Raw Normal View History

// 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 <assert.h>
Perform post-flush updates of memtable list in a callback (#6069) Summary: Currently, the following interleaving of events can lead to SuperVersion containing both immutable memtables as well as the resulting L0. This can cause Get to return incorrect result if there are merge operands. This may also affect other operations such as single deletes. ``` time main_thr bg_flush_thr bg_compact_thr compact_thr set_opts_thr 0 | WriteManifest:0 1 | issue compact 2 | wait 3 | Merge(counter) 4 | issue flush 5 | wait 6 | WriteManifest:1 7 | wake up 8 | write manifest 9 | wake up 10 | Get(counter) 11 | remove imm V ``` The reason behind is that: one bg flush thread's installing new `Version` can be batched and performed by another thread that is the "leader" MANIFEST writer. This bg thread removes the memtables from current super version only after `LogAndApply` returns. After the leader MANIFEST writer signals (releasing mutex) this bg flush thread, it is possible that another thread sees this cf with both memtables (whose data have been flushed to the newest L0) and the L0 before this bg flush thread removes the memtables. To address this issue, each bg flush thread can pass a callback function to `LogAndApply`. The callback is responsible for removing the memtables. Therefore, the leader MANIFEST writer can call this callback and remove the memtables before releasing the mutex. Test plan (devserver) ``` $make merge_test $./merge_test --gtest_filter=MergeTest.MergeWithCompactionAndFlush $make check ``` Pull Request resolved: https://github.com/facebook/rocksdb/pull/6069 Reviewed By: cheng-chang Differential Revision: D18790894 Pulled By: riversand963 fbshipit-source-id: e41bd600c0448b4f4b2deb3f7677f95e3076b4ed
2020-10-27 01:20:43 +00:00
#include <iostream>
Perform post-flush updates of memtable list in a callback (#6069) Summary: Currently, the following interleaving of events can lead to SuperVersion containing both immutable memtables as well as the resulting L0. This can cause Get to return incorrect result if there are merge operands. This may also affect other operations such as single deletes. ``` time main_thr bg_flush_thr bg_compact_thr compact_thr set_opts_thr 0 | WriteManifest:0 1 | issue compact 2 | wait 3 | Merge(counter) 4 | issue flush 5 | wait 6 | WriteManifest:1 7 | wake up 8 | write manifest 9 | wake up 10 | Get(counter) 11 | remove imm V ``` The reason behind is that: one bg flush thread's installing new `Version` can be batched and performed by another thread that is the "leader" MANIFEST writer. This bg thread removes the memtables from current super version only after `LogAndApply` returns. After the leader MANIFEST writer signals (releasing mutex) this bg flush thread, it is possible that another thread sees this cf with both memtables (whose data have been flushed to the newest L0) and the L0 before this bg flush thread removes the memtables. To address this issue, each bg flush thread can pass a callback function to `LogAndApply`. The callback is responsible for removing the memtables. Therefore, the leader MANIFEST writer can call this callback and remove the memtables before releasing the mutex. Test plan (devserver) ``` $make merge_test $./merge_test --gtest_filter=MergeTest.MergeWithCompactionAndFlush $make check ``` Pull Request resolved: https://github.com/facebook/rocksdb/pull/6069 Reviewed By: cheng-chang Differential Revision: D18790894 Pulled By: riversand963 fbshipit-source-id: e41bd600c0448b4f4b2deb3f7677f95e3076b4ed
2020-10-27 01:20:43 +00:00
#include <memory>
#include "db/db_impl/db_impl.h"
#include "db/dbformat.h"
#include "db/write_batch_internal.h"
#include "port/stack_trace.h"
#include "rocksdb/cache.h"
#include "rocksdb/comparator.h"
#include "rocksdb/db.h"
#include "rocksdb/env.h"
#include "rocksdb/merge_operator.h"
#include "rocksdb/utilities/db_ttl.h"
Introduce a wide column aware MergeOperator API (#11807) Summary: Pull Request resolved: https://github.com/facebook/rocksdb/pull/11807 For now, RocksDB has limited support for using merge with wide columns: when a bunch of merge operands have to be applied to a wide-column base value, RocksDB currently passes only the value of the default column to the application's `MergeOperator`, which means there is no way to update any other columns during a merge. As a first step in making this more general, the patch adds a new API `FullMergeV3` to `MergeOperator`. `FullMergeV3`'s interface enables applications to receive a plain, wide-column, or non-existent base value as merge input, and to produce a new plain value, a new wide-column value, or an existing operand as merge output. Note that there are no limitations on the column names and values if the merge result is a wide-column entity. Also, the interface is general in the sense that it makes it possible e.g. for a merge that takes a plain base value and some deltas to produce a wide-column entity as a result. For backward compatibility, the default implementation of `FullMergeV3` falls back to `FullMergeV2` and implements the current logic where merge operands are applied to the default column of the base entity and any other columns are unchanged. (Note that with `FullMergeV3` in the `MergeOperator` interface, this behavior will become customizable.) This patch just introduces the new API and the default backward compatible implementation. I plan to integrate `FullMergeV3` into the query and compaction logic in subsequent diffs. Reviewed By: jaykorean Differential Revision: D49117253 fbshipit-source-id: 109e016f25cd130fc504790818d927bae7fec6bd
2023-09-11 19:13:58 +00:00
#include "rocksdb/wide_columns.h"
#include "test_util/testharness.h"
Perform post-flush updates of memtable list in a callback (#6069) Summary: Currently, the following interleaving of events can lead to SuperVersion containing both immutable memtables as well as the resulting L0. This can cause Get to return incorrect result if there are merge operands. This may also affect other operations such as single deletes. ``` time main_thr bg_flush_thr bg_compact_thr compact_thr set_opts_thr 0 | WriteManifest:0 1 | issue compact 2 | wait 3 | Merge(counter) 4 | issue flush 5 | wait 6 | WriteManifest:1 7 | wake up 8 | write manifest 9 | wake up 10 | Get(counter) 11 | remove imm V ``` The reason behind is that: one bg flush thread's installing new `Version` can be batched and performed by another thread that is the "leader" MANIFEST writer. This bg thread removes the memtables from current super version only after `LogAndApply` returns. After the leader MANIFEST writer signals (releasing mutex) this bg flush thread, it is possible that another thread sees this cf with both memtables (whose data have been flushed to the newest L0) and the L0 before this bg flush thread removes the memtables. To address this issue, each bg flush thread can pass a callback function to `LogAndApply`. The callback is responsible for removing the memtables. Therefore, the leader MANIFEST writer can call this callback and remove the memtables before releasing the mutex. Test plan (devserver) ``` $make merge_test $./merge_test --gtest_filter=MergeTest.MergeWithCompactionAndFlush $make check ``` Pull Request resolved: https://github.com/facebook/rocksdb/pull/6069 Reviewed By: cheng-chang Differential Revision: D18790894 Pulled By: riversand963 fbshipit-source-id: e41bd600c0448b4f4b2deb3f7677f95e3076b4ed
2020-10-27 01:20:43 +00:00
#include "util/coding.h"
#include "utilities/merge_operators.h"
namespace ROCKSDB_NAMESPACE {
Converted db/merge_test.cc to use gtest (#4114) Summary: Picked up a task to convert this to use the gtest framework. It can't be this simple, can it? It works, but should all the std::cout be removed? ``` [$] ~/git/rocksdb [gft !]: ./merge_test [==========] Running 2 tests from 1 test case. [----------] Global test environment set-up. [----------] 2 tests from MergeTest [ RUN ] MergeTest.MergeDbTest Test read-modify-write counters... a: 3 1 2 a: 3 b: 1225 3 Compaction started ... Compaction ended a: 3 b: 1225 Test merge-based counters... a: 3 1 2 a: 3 b: 1225 3 Test merge in memtable... a: 3 1 2 a: 3 b: 1225 3 Test Partial-Merge Test merge-operator not set after reopen [ OK ] MergeTest.MergeDbTest (93 ms) [ RUN ] MergeTest.MergeDbTtlTest Opening database with TTL Test read-modify-write counters... a: 3 1 2 a: 3 b: 1225 3 Compaction started ... Compaction ended a: 3 b: 1225 Test merge-based counters... a: 3 1 2 a: 3 b: 1225 3 Test merge in memtable... Opening database with TTL a: 3 1 2 a: 3 b: 1225 3 Test Partial-Merge Opening database with TTL Opening database with TTL Opening database with TTL Opening database with TTL Test merge-operator not set after reopen [ OK ] MergeTest.MergeDbTtlTest (97 ms) [----------] 2 tests from MergeTest (190 ms total) [----------] Global test environment tear-down [==========] 2 tests from 1 test case ran. (190 ms total) [ PASSED ] 2 tests. ``` Pull Request resolved: https://github.com/facebook/rocksdb/pull/4114 Differential Revision: D8822886 Pulled By: gfosco fbshipit-source-id: c299d008e883c3bb911d2b357a2e9e4423f8e91a
2018-07-13 21:07:53 +00:00
bool use_compression;
class MergeTest : public testing::Test {};
size_t num_merge_operator_calls;
void resetNumMergeOperatorCalls() { num_merge_operator_calls = 0; }
size_t num_partial_merge_calls;
void resetNumPartialMergeCalls() { num_partial_merge_calls = 0; }
class CountMergeOperator : public AssociativeMergeOperator {
public:
CountMergeOperator() {
mergeOperator_ = MergeOperators::CreateUInt64AddOperator();
}
bool Merge(const Slice& key, const Slice* existing_value, const Slice& value,
std::string* new_value, Logger* logger) const override {
assert(new_value->empty());
++num_merge_operator_calls;
if (existing_value == nullptr) {
new_value->assign(value.data(), value.size());
return true;
}
return mergeOperator_->PartialMerge(key, *existing_value, value, new_value,
logger);
}
bool PartialMergeMulti(const Slice& key,
const std::deque<Slice>& operand_list,
std::string* new_value,
Logger* logger) const override {
assert(new_value->empty());
++num_partial_merge_calls;
return mergeOperator_->PartialMergeMulti(key, operand_list, new_value,
logger);
}
const char* Name() const override { return "UInt64AddOperator"; }
private:
std::shared_ptr<MergeOperator> mergeOperator_;
};
class EnvMergeTest : public EnvWrapper {
public:
EnvMergeTest() : EnvWrapper(Env::Default()) {}
static const char* kClassName() { return "MergeEnv"; }
const char* Name() const override { return kClassName(); }
// ~EnvMergeTest() override {}
uint64_t NowNanos() override {
++now_nanos_count_;
return target()->NowNanos();
}
static uint64_t now_nanos_count_;
static std::unique_ptr<EnvMergeTest> singleton_;
static EnvMergeTest* GetInstance() {
if (nullptr == singleton_) singleton_.reset(new EnvMergeTest);
return singleton_.get();
}
};
uint64_t EnvMergeTest::now_nanos_count_{0};
std::unique_ptr<EnvMergeTest> EnvMergeTest::singleton_;
std::shared_ptr<DB> OpenDb(const std::string& dbname, const bool ttl = false,
const size_t max_successive_merges = 0) {
DB* db;
Options options;
options.create_if_missing = true;
options.merge_operator = std::make_shared<CountMergeOperator>();
options.max_successive_merges = max_successive_merges;
options.env = EnvMergeTest::GetInstance();
EXPECT_OK(DestroyDB(dbname, Options()));
Status s;
if (ttl) {
2014-04-29 03:44:33 +00:00
DBWithTTL* db_with_ttl;
s = DBWithTTL::Open(options, dbname, &db_with_ttl);
db = db_with_ttl;
} else {
2013-08-06 18:42:21 +00:00
s = DB::Open(options, dbname, &db);
}
EXPECT_OK(s);
assert(s.ok());
Built-in support for generating unique IDs, bug fix (#8708) Summary: Env::GenerateUniqueId() works fine on Windows and on POSIX where /proc/sys/kernel/random/uuid exists. Our other implementation is flawed and easily produces collision in a new multi-threaded test. As we rely more heavily on DB session ID uniqueness, this becomes a serious issue. This change combines several individually suitable entropy sources for reliable generation of random unique IDs, with goal of uniqueness and portability, not cryptographic strength nor maximum speed. Specifically: * Moves code for getting UUIDs from the OS to port::GenerateRfcUuid rather than in Env implementation details. Callers are now told whether the operation fails or succeeds. * Adds an internal API GenerateRawUniqueId for generating high-quality 128-bit unique identifiers, by combining entropy from three "tracks": * Lots of info from default Env like time, process id, and hostname. * std::random_device * port::GenerateRfcUuid (when working) * Built-in implementations of Env::GenerateUniqueId() will now always produce an RFC 4122 UUID string, either from platform-specific API or by converting the output of GenerateRawUniqueId. DB session IDs now use GenerateRawUniqueId while DB IDs (not as critical) try to use port::GenerateRfcUuid but fall back on GenerateRawUniqueId with conversion to an RFC 4122 UUID. GenerateRawUniqueId is declared and defined under env/ rather than util/ or even port/ because of the Env dependency. Likely follow-up: enhance GenerateRawUniqueId to be faster after the first call and to guarantee uniqueness within the lifetime of a single process (imparting the same property onto DB session IDs). Pull Request resolved: https://github.com/facebook/rocksdb/pull/8708 Test Plan: A new mini-stress test in env_test checks the various public and internal APIs for uniqueness, including each track of GenerateRawUniqueId individually. We can't hope to verify anywhere close to 128 bits of entropy, but it can at least detect flaws as bad as the old code. Serial execution of the new tests takes about 350 ms on my machine. Reviewed By: zhichao-cao, mrambacher Differential Revision: D30563780 Pulled By: pdillinger fbshipit-source-id: de4c9ff4b2f581cf784fcedb5f39f16e5185c364
2021-08-30 22:19:39 +00:00
// Allowed to call NowNanos during DB creation (in GenerateRawUniqueId() for
// session ID)
EnvMergeTest::now_nanos_count_ = 0;
return std::shared_ptr<DB>(db);
}
// Imagine we are maintaining a set of uint64 counters.
// Each counter has a distinct name. And we would like
// to support four high level operations:
// set, add, get and remove
// This is a quick implementation without a Merge operation.
class Counters {
protected:
std::shared_ptr<DB> db_;
WriteOptions put_option_;
ReadOptions get_option_;
WriteOptions delete_option_;
uint64_t default_;
public:
explicit Counters(std::shared_ptr<DB> db, uint64_t defaultCount = 0)
: db_(db),
put_option_(),
get_option_(),
delete_option_(),
default_(defaultCount) {
assert(db_);
}
virtual ~Counters() {}
// public interface of Counters.
// All four functions return false
// if the underlying level db operation failed.
// mapped to a levedb Put
bool set(const std::string& key, uint64_t value) {
// just treat the internal rep of int64 as the string
char buf[sizeof(value)];
EncodeFixed64(buf, value);
Slice slice(buf, sizeof(value));
auto s = db_->Put(put_option_, key, slice);
if (s.ok()) {
return true;
} else {
std::cerr << s.ToString() << std::endl;
return false;
}
}
// mapped to a rocksdb Delete
bool remove(const std::string& key) {
auto s = db_->Delete(delete_option_, key);
if (s.ok()) {
return true;
} else {
std::cerr << s.ToString() << std::endl;
return false;
}
}
// mapped to a rocksdb Get
bool get(const std::string& key, uint64_t* value) {
std::string str;
auto s = db_->Get(get_option_, key, &str);
if (s.IsNotFound()) {
// return default value if not found;
*value = default_;
return true;
} else if (s.ok()) {
// deserialization
if (str.size() != sizeof(uint64_t)) {
std::cerr << "value corruption\n";
return false;
}
*value = DecodeFixed64(&str[0]);
return true;
} else {
std::cerr << s.ToString() << std::endl;
return false;
}
}
// 'add' is implemented as get -> modify -> set
// An alternative is a single merge operation, see MergeBasedCounters
virtual bool add(const std::string& key, uint64_t value) {
uint64_t base = default_;
return get(key, &base) && set(key, base + value);
}
// convenience functions for testing
void assert_set(const std::string& key, uint64_t value) {
assert(set(key, value));
}
void assert_remove(const std::string& key) { assert(remove(key)); }
uint64_t assert_get(const std::string& key) {
uint64_t value = default_;
int result = get(key, &value);
assert(result);
if (result == 0) exit(1); // Disable unused variable warning.
return value;
}
void assert_add(const std::string& key, uint64_t value) {
int result = add(key, value);
assert(result);
if (result == 0) exit(1); // Disable unused variable warning.
}
};
// Implement 'add' directly with the new Merge operation
class MergeBasedCounters : public Counters {
private:
WriteOptions merge_option_; // for merge
public:
explicit MergeBasedCounters(std::shared_ptr<DB> db, uint64_t defaultCount = 0)
: Counters(db, defaultCount), merge_option_() {}
// mapped to a rocksdb Merge operation
bool add(const std::string& key, uint64_t value) override {
char encoded[sizeof(uint64_t)];
EncodeFixed64(encoded, value);
Slice slice(encoded, sizeof(uint64_t));
auto s = db_->Merge(merge_option_, key, slice);
if (s.ok()) {
return true;
} else {
std::cerr << s.ToString() << std::endl;
return false;
}
}
};
void dumpDb(DB* db) {
auto it = std::unique_ptr<Iterator>(db->NewIterator(ReadOptions()));
for (it->SeekToFirst(); it->Valid(); it->Next()) {
// uint64_t value = DecodeFixed64(it->value().data());
// std::cout << it->key().ToString() << ": " << value << std::endl;
}
assert(it->status().ok()); // Check for any errors found during the scan
}
void testCounters(Counters& counters, DB* db, bool test_compaction) {
FlushOptions o;
o.wait = true;
counters.assert_set("a", 1);
if (test_compaction) {
ASSERT_OK(db->Flush(o));
}
ASSERT_EQ(counters.assert_get("a"), 1);
counters.assert_remove("b");
// defaut value is 0 if non-existent
ASSERT_EQ(counters.assert_get("b"), 0);
counters.assert_add("a", 2);
if (test_compaction) {
ASSERT_OK(db->Flush(o));
}
// 1+2 = 3
ASSERT_EQ(counters.assert_get("a"), 3);
dumpDb(db);
// 1+...+49 = ?
uint64_t sum = 0;
for (int i = 1; i < 50; i++) {
counters.assert_add("b", i);
sum += i;
}
ASSERT_EQ(counters.assert_get("b"), sum);
dumpDb(db);
if (test_compaction) {
ASSERT_OK(db->Flush(o));
ASSERT_OK(db->CompactRange(CompactRangeOptions(), nullptr, nullptr));
dumpDb(db);
ASSERT_EQ(counters.assert_get("a"), 3);
ASSERT_EQ(counters.assert_get("b"), sum);
}
}
Perform post-flush updates of memtable list in a callback (#6069) Summary: Currently, the following interleaving of events can lead to SuperVersion containing both immutable memtables as well as the resulting L0. This can cause Get to return incorrect result if there are merge operands. This may also affect other operations such as single deletes. ``` time main_thr bg_flush_thr bg_compact_thr compact_thr set_opts_thr 0 | WriteManifest:0 1 | issue compact 2 | wait 3 | Merge(counter) 4 | issue flush 5 | wait 6 | WriteManifest:1 7 | wake up 8 | write manifest 9 | wake up 10 | Get(counter) 11 | remove imm V ``` The reason behind is that: one bg flush thread's installing new `Version` can be batched and performed by another thread that is the "leader" MANIFEST writer. This bg thread removes the memtables from current super version only after `LogAndApply` returns. After the leader MANIFEST writer signals (releasing mutex) this bg flush thread, it is possible that another thread sees this cf with both memtables (whose data have been flushed to the newest L0) and the L0 before this bg flush thread removes the memtables. To address this issue, each bg flush thread can pass a callback function to `LogAndApply`. The callback is responsible for removing the memtables. Therefore, the leader MANIFEST writer can call this callback and remove the memtables before releasing the mutex. Test plan (devserver) ``` $make merge_test $./merge_test --gtest_filter=MergeTest.MergeWithCompactionAndFlush $make check ``` Pull Request resolved: https://github.com/facebook/rocksdb/pull/6069 Reviewed By: cheng-chang Differential Revision: D18790894 Pulled By: riversand963 fbshipit-source-id: e41bd600c0448b4f4b2deb3f7677f95e3076b4ed
2020-10-27 01:20:43 +00:00
void testCountersWithFlushAndCompaction(Counters& counters, DB* db) {
ASSERT_OK(db->Put({}, "1", "1"));
ASSERT_OK(db->Flush(FlushOptions()));
std::atomic<int> cnt{0};
const auto get_thread_id = [&cnt]() {
thread_local int thread_id{cnt++};
return thread_id;
};
SyncPoint::GetInstance()->DisableProcessing();
SyncPoint::GetInstance()->ClearAllCallBacks();
SyncPoint::GetInstance()->SetCallBack(
"VersionSet::LogAndApply:BeforeWriterWaiting", [&](void* /*arg*/) {
int thread_id = get_thread_id();
if (1 == thread_id) {
TEST_SYNC_POINT(
"testCountersWithFlushAndCompaction::bg_compact_thread:0");
} else if (2 == thread_id) {
TEST_SYNC_POINT(
"testCountersWithFlushAndCompaction::bg_flush_thread:0");
}
});
SyncPoint::GetInstance()->SetCallBack(
"VersionSet::LogAndApply:WriteManifest", [&](void* /*arg*/) {
int thread_id = get_thread_id();
if (0 == thread_id) {
TEST_SYNC_POINT(
"testCountersWithFlushAndCompaction::set_options_thread:0");
TEST_SYNC_POINT(
"testCountersWithFlushAndCompaction::set_options_thread:1");
}
});
SyncPoint::GetInstance()->SetCallBack(
"VersionSet::LogAndApply:WakeUpAndDone", [&](void* arg) {
auto* mutex = reinterpret_cast<InstrumentedMutex*>(arg);
mutex->AssertHeld();
int thread_id = get_thread_id();
ASSERT_EQ(2, thread_id);
mutex->Unlock();
TEST_SYNC_POINT(
"testCountersWithFlushAndCompaction::bg_flush_thread:1");
TEST_SYNC_POINT(
"testCountersWithFlushAndCompaction::bg_flush_thread:2");
mutex->Lock();
});
SyncPoint::GetInstance()->LoadDependency({
{"testCountersWithFlushAndCompaction::set_options_thread:0",
"testCountersWithCompactionAndFlush:BeforeCompact"},
{"testCountersWithFlushAndCompaction::bg_compact_thread:0",
"testCountersWithFlushAndCompaction:BeforeIncCounters"},
{"testCountersWithFlushAndCompaction::bg_flush_thread:0",
"testCountersWithFlushAndCompaction::set_options_thread:1"},
{"testCountersWithFlushAndCompaction::bg_flush_thread:1",
"testCountersWithFlushAndCompaction:BeforeVerification"},
{"testCountersWithFlushAndCompaction:AfterGet",
"testCountersWithFlushAndCompaction::bg_flush_thread:2"},
});
SyncPoint::GetInstance()->EnableProcessing();
port::Thread set_options_thread([&]() {
ASSERT_OK(reinterpret_cast<DBImpl*>(db)->SetOptions(
{{"disable_auto_compactions", "false"}}));
});
TEST_SYNC_POINT("testCountersWithCompactionAndFlush:BeforeCompact");
port::Thread compact_thread([&]() {
ASSERT_OK(reinterpret_cast<DBImpl*>(db)->CompactRange(
CompactRangeOptions(), db->DefaultColumnFamily(), nullptr, nullptr));
});
TEST_SYNC_POINT("testCountersWithFlushAndCompaction:BeforeIncCounters");
counters.add("test-key", 1);
FlushOptions flush_opts;
flush_opts.wait = false;
ASSERT_OK(db->Flush(flush_opts));
TEST_SYNC_POINT("testCountersWithFlushAndCompaction:BeforeVerification");
std::string expected;
PutFixed64(&expected, 1);
std::string actual;
Status s = db->Get(ReadOptions(), "test-key", &actual);
TEST_SYNC_POINT("testCountersWithFlushAndCompaction:AfterGet");
set_options_thread.join();
compact_thread.join();
ASSERT_OK(s);
ASSERT_EQ(expected, actual);
SyncPoint::GetInstance()->DisableProcessing();
SyncPoint::GetInstance()->ClearAllCallBacks();
}
void testSuccessiveMerge(Counters& counters, size_t max_num_merges,
size_t num_merges) {
counters.assert_remove("z");
uint64_t sum = 0;
for (size_t i = 1; i <= num_merges; ++i) {
resetNumMergeOperatorCalls();
counters.assert_add("z", i);
sum += i;
if (i % (max_num_merges + 1) == 0) {
ASSERT_EQ(num_merge_operator_calls, max_num_merges + 1);
} else {
ASSERT_EQ(num_merge_operator_calls, 0);
}
resetNumMergeOperatorCalls();
ASSERT_EQ(counters.assert_get("z"), sum);
ASSERT_EQ(num_merge_operator_calls, i % (max_num_merges + 1));
}
}
void testPartialMerge(Counters* counters, DB* db, size_t max_merge,
size_t min_merge, size_t count) {
FlushOptions o;
o.wait = true;
// Test case 1: partial merge should be called when the number of merge
// operands exceeds the threshold.
uint64_t tmp_sum = 0;
resetNumPartialMergeCalls();
for (size_t i = 1; i <= count; i++) {
counters->assert_add("b", i);
tmp_sum += i;
}
ASSERT_OK(db->Flush(o));
ASSERT_OK(db->CompactRange(CompactRangeOptions(), nullptr, nullptr));
ASSERT_EQ(tmp_sum, counters->assert_get("b"));
if (count > max_merge) {
// in this case, FullMerge should be called instead.
ASSERT_EQ(num_partial_merge_calls, 0U);
} else {
// if count >= min_merge, then partial merge should be called once.
ASSERT_EQ((count >= min_merge), (num_partial_merge_calls == 1));
}
// Test case 2: partial merge should not be called when a put is found.
resetNumPartialMergeCalls();
tmp_sum = 0;
ASSERT_OK(db->Put(ROCKSDB_NAMESPACE::WriteOptions(), "c", "10"));
for (size_t i = 1; i <= count; i++) {
counters->assert_add("c", i);
tmp_sum += i;
}
ASSERT_OK(db->Flush(o));
ASSERT_OK(db->CompactRange(CompactRangeOptions(), nullptr, nullptr));
ASSERT_EQ(tmp_sum, counters->assert_get("c"));
ASSERT_EQ(num_partial_merge_calls, 0U);
Built-in support for generating unique IDs, bug fix (#8708) Summary: Env::GenerateUniqueId() works fine on Windows and on POSIX where /proc/sys/kernel/random/uuid exists. Our other implementation is flawed and easily produces collision in a new multi-threaded test. As we rely more heavily on DB session ID uniqueness, this becomes a serious issue. This change combines several individually suitable entropy sources for reliable generation of random unique IDs, with goal of uniqueness and portability, not cryptographic strength nor maximum speed. Specifically: * Moves code for getting UUIDs from the OS to port::GenerateRfcUuid rather than in Env implementation details. Callers are now told whether the operation fails or succeeds. * Adds an internal API GenerateRawUniqueId for generating high-quality 128-bit unique identifiers, by combining entropy from three "tracks": * Lots of info from default Env like time, process id, and hostname. * std::random_device * port::GenerateRfcUuid (when working) * Built-in implementations of Env::GenerateUniqueId() will now always produce an RFC 4122 UUID string, either from platform-specific API or by converting the output of GenerateRawUniqueId. DB session IDs now use GenerateRawUniqueId while DB IDs (not as critical) try to use port::GenerateRfcUuid but fall back on GenerateRawUniqueId with conversion to an RFC 4122 UUID. GenerateRawUniqueId is declared and defined under env/ rather than util/ or even port/ because of the Env dependency. Likely follow-up: enhance GenerateRawUniqueId to be faster after the first call and to guarantee uniqueness within the lifetime of a single process (imparting the same property onto DB session IDs). Pull Request resolved: https://github.com/facebook/rocksdb/pull/8708 Test Plan: A new mini-stress test in env_test checks the various public and internal APIs for uniqueness, including each track of GenerateRawUniqueId individually. We can't hope to verify anywhere close to 128 bits of entropy, but it can at least detect flaws as bad as the old code. Serial execution of the new tests takes about 350 ms on my machine. Reviewed By: zhichao-cao, mrambacher Differential Revision: D30563780 Pulled By: pdillinger fbshipit-source-id: de4c9ff4b2f581cf784fcedb5f39f16e5185c364
2021-08-30 22:19:39 +00:00
// NowNanos was previously called in MergeHelper::FilterMerge(), which
// harmed performance.
ASSERT_EQ(EnvMergeTest::now_nanos_count_, 0U);
}
void testSingleBatchSuccessiveMerge(DB* db, size_t max_num_merges,
size_t num_merges) {
ASSERT_GT(num_merges, max_num_merges);
Slice key("BatchSuccessiveMerge");
uint64_t merge_value = 1;
char buf[sizeof(merge_value)];
EncodeFixed64(buf, merge_value);
Slice merge_value_slice(buf, sizeof(merge_value));
// Create the batch
WriteBatch batch;
for (size_t i = 0; i < num_merges; ++i) {
ASSERT_OK(batch.Merge(key, merge_value_slice));
}
// Apply to memtable and count the number of merges
resetNumMergeOperatorCalls();
ASSERT_OK(db->Write(WriteOptions(), &batch));
ASSERT_EQ(
num_merge_operator_calls,
static_cast<size_t>(num_merges - (num_merges % (max_num_merges + 1))));
// Get the value
resetNumMergeOperatorCalls();
std::string get_value_str;
ASSERT_OK(db->Get(ReadOptions(), key, &get_value_str));
assert(get_value_str.size() == sizeof(uint64_t));
uint64_t get_value = DecodeFixed64(&get_value_str[0]);
ASSERT_EQ(get_value, num_merges * merge_value);
ASSERT_EQ(num_merge_operator_calls,
static_cast<size_t>((num_merges % (max_num_merges + 1))));
}
Converted db/merge_test.cc to use gtest (#4114) Summary: Picked up a task to convert this to use the gtest framework. It can't be this simple, can it? It works, but should all the std::cout be removed? ``` [$] ~/git/rocksdb [gft !]: ./merge_test [==========] Running 2 tests from 1 test case. [----------] Global test environment set-up. [----------] 2 tests from MergeTest [ RUN ] MergeTest.MergeDbTest Test read-modify-write counters... a: 3 1 2 a: 3 b: 1225 3 Compaction started ... Compaction ended a: 3 b: 1225 Test merge-based counters... a: 3 1 2 a: 3 b: 1225 3 Test merge in memtable... a: 3 1 2 a: 3 b: 1225 3 Test Partial-Merge Test merge-operator not set after reopen [ OK ] MergeTest.MergeDbTest (93 ms) [ RUN ] MergeTest.MergeDbTtlTest Opening database with TTL Test read-modify-write counters... a: 3 1 2 a: 3 b: 1225 3 Compaction started ... Compaction ended a: 3 b: 1225 Test merge-based counters... a: 3 1 2 a: 3 b: 1225 3 Test merge in memtable... Opening database with TTL a: 3 1 2 a: 3 b: 1225 3 Test Partial-Merge Opening database with TTL Opening database with TTL Opening database with TTL Opening database with TTL Test merge-operator not set after reopen [ OK ] MergeTest.MergeDbTtlTest (97 ms) [----------] 2 tests from MergeTest (190 ms total) [----------] Global test environment tear-down [==========] 2 tests from 1 test case ran. (190 ms total) [ PASSED ] 2 tests. ``` Pull Request resolved: https://github.com/facebook/rocksdb/pull/4114 Differential Revision: D8822886 Pulled By: gfosco fbshipit-source-id: c299d008e883c3bb911d2b357a2e9e4423f8e91a
2018-07-13 21:07:53 +00:00
void runTest(const std::string& dbname, const bool use_ttl = false) {
{
auto db = OpenDb(dbname, use_ttl);
{
Counters counters(db, 0);
testCounters(counters, db.get(), true);
}
{
MergeBasedCounters counters(db, 0);
Converted db/merge_test.cc to use gtest (#4114) Summary: Picked up a task to convert this to use the gtest framework. It can't be this simple, can it? It works, but should all the std::cout be removed? ``` [$] ~/git/rocksdb [gft !]: ./merge_test [==========] Running 2 tests from 1 test case. [----------] Global test environment set-up. [----------] 2 tests from MergeTest [ RUN ] MergeTest.MergeDbTest Test read-modify-write counters... a: 3 1 2 a: 3 b: 1225 3 Compaction started ... Compaction ended a: 3 b: 1225 Test merge-based counters... a: 3 1 2 a: 3 b: 1225 3 Test merge in memtable... a: 3 1 2 a: 3 b: 1225 3 Test Partial-Merge Test merge-operator not set after reopen [ OK ] MergeTest.MergeDbTest (93 ms) [ RUN ] MergeTest.MergeDbTtlTest Opening database with TTL Test read-modify-write counters... a: 3 1 2 a: 3 b: 1225 3 Compaction started ... Compaction ended a: 3 b: 1225 Test merge-based counters... a: 3 1 2 a: 3 b: 1225 3 Test merge in memtable... Opening database with TTL a: 3 1 2 a: 3 b: 1225 3 Test Partial-Merge Opening database with TTL Opening database with TTL Opening database with TTL Opening database with TTL Test merge-operator not set after reopen [ OK ] MergeTest.MergeDbTtlTest (97 ms) [----------] 2 tests from MergeTest (190 ms total) [----------] Global test environment tear-down [==========] 2 tests from 1 test case ran. (190 ms total) [ PASSED ] 2 tests. ``` Pull Request resolved: https://github.com/facebook/rocksdb/pull/4114 Differential Revision: D8822886 Pulled By: gfosco fbshipit-source-id: c299d008e883c3bb911d2b357a2e9e4423f8e91a
2018-07-13 21:07:53 +00:00
testCounters(counters, db.get(), use_compression);
}
}
ASSERT_OK(DestroyDB(dbname, Options()));
{
size_t max_merge = 5;
auto db = OpenDb(dbname, use_ttl, max_merge);
MergeBasedCounters counters(db, 0);
Converted db/merge_test.cc to use gtest (#4114) Summary: Picked up a task to convert this to use the gtest framework. It can't be this simple, can it? It works, but should all the std::cout be removed? ``` [$] ~/git/rocksdb [gft !]: ./merge_test [==========] Running 2 tests from 1 test case. [----------] Global test environment set-up. [----------] 2 tests from MergeTest [ RUN ] MergeTest.MergeDbTest Test read-modify-write counters... a: 3 1 2 a: 3 b: 1225 3 Compaction started ... Compaction ended a: 3 b: 1225 Test merge-based counters... a: 3 1 2 a: 3 b: 1225 3 Test merge in memtable... a: 3 1 2 a: 3 b: 1225 3 Test Partial-Merge Test merge-operator not set after reopen [ OK ] MergeTest.MergeDbTest (93 ms) [ RUN ] MergeTest.MergeDbTtlTest Opening database with TTL Test read-modify-write counters... a: 3 1 2 a: 3 b: 1225 3 Compaction started ... Compaction ended a: 3 b: 1225 Test merge-based counters... a: 3 1 2 a: 3 b: 1225 3 Test merge in memtable... Opening database with TTL a: 3 1 2 a: 3 b: 1225 3 Test Partial-Merge Opening database with TTL Opening database with TTL Opening database with TTL Opening database with TTL Test merge-operator not set after reopen [ OK ] MergeTest.MergeDbTtlTest (97 ms) [----------] 2 tests from MergeTest (190 ms total) [----------] Global test environment tear-down [==========] 2 tests from 1 test case ran. (190 ms total) [ PASSED ] 2 tests. ``` Pull Request resolved: https://github.com/facebook/rocksdb/pull/4114 Differential Revision: D8822886 Pulled By: gfosco fbshipit-source-id: c299d008e883c3bb911d2b357a2e9e4423f8e91a
2018-07-13 21:07:53 +00:00
testCounters(counters, db.get(), use_compression);
testSuccessiveMerge(counters, max_merge, max_merge * 2);
testSingleBatchSuccessiveMerge(db.get(), 5, 7);
ASSERT_OK(db->Close());
ASSERT_OK(DestroyDB(dbname, Options()));
}
{
size_t max_merge = 100;
// Min merge is hard-coded to 2.
uint32_t min_merge = 2;
for (uint32_t count = min_merge - 1; count <= min_merge + 1; count++) {
auto db = OpenDb(dbname, use_ttl, max_merge);
MergeBasedCounters counters(db, 0);
testPartialMerge(&counters, db.get(), max_merge, min_merge, count);
ASSERT_OK(db->Close());
ASSERT_OK(DestroyDB(dbname, Options()));
}
{
auto db = OpenDb(dbname, use_ttl, max_merge);
MergeBasedCounters counters(db, 0);
testPartialMerge(&counters, db.get(), max_merge, min_merge,
min_merge * 10);
ASSERT_OK(db->Close());
ASSERT_OK(DestroyDB(dbname, Options()));
}
}
{
{
auto db = OpenDb(dbname);
MergeBasedCounters counters(db, 0);
counters.add("test-key", 1);
counters.add("test-key", 1);
counters.add("test-key", 1);
ASSERT_OK(db->CompactRange(CompactRangeOptions(), nullptr, nullptr));
}
DB* reopen_db;
ASSERT_OK(DB::Open(Options(), dbname, &reopen_db));
std::string value;
ASSERT_NOK(reopen_db->Get(ReadOptions(), "test-key", &value));
delete reopen_db;
ASSERT_OK(DestroyDB(dbname, Options()));
}
/* Temporary remove this test
{
std::cout << "Test merge-operator not set after reopen (recovery case)\n";
{
auto db = OpenDb(dbname);
MergeBasedCounters counters(db, 0);
counters.add("test-key", 1);
counters.add("test-key", 1);
counters.add("test-key", 1);
}
DB* reopen_db;
ASSERT_TRUE(DB::Open(Options(), dbname, &reopen_db).IsInvalidArgument());
}
*/
}
Converted db/merge_test.cc to use gtest (#4114) Summary: Picked up a task to convert this to use the gtest framework. It can't be this simple, can it? It works, but should all the std::cout be removed? ``` [$] ~/git/rocksdb [gft !]: ./merge_test [==========] Running 2 tests from 1 test case. [----------] Global test environment set-up. [----------] 2 tests from MergeTest [ RUN ] MergeTest.MergeDbTest Test read-modify-write counters... a: 3 1 2 a: 3 b: 1225 3 Compaction started ... Compaction ended a: 3 b: 1225 Test merge-based counters... a: 3 1 2 a: 3 b: 1225 3 Test merge in memtable... a: 3 1 2 a: 3 b: 1225 3 Test Partial-Merge Test merge-operator not set after reopen [ OK ] MergeTest.MergeDbTest (93 ms) [ RUN ] MergeTest.MergeDbTtlTest Opening database with TTL Test read-modify-write counters... a: 3 1 2 a: 3 b: 1225 3 Compaction started ... Compaction ended a: 3 b: 1225 Test merge-based counters... a: 3 1 2 a: 3 b: 1225 3 Test merge in memtable... Opening database with TTL a: 3 1 2 a: 3 b: 1225 3 Test Partial-Merge Opening database with TTL Opening database with TTL Opening database with TTL Opening database with TTL Test merge-operator not set after reopen [ OK ] MergeTest.MergeDbTtlTest (97 ms) [----------] 2 tests from MergeTest (190 ms total) [----------] Global test environment tear-down [==========] 2 tests from 1 test case ran. (190 ms total) [ PASSED ] 2 tests. ``` Pull Request resolved: https://github.com/facebook/rocksdb/pull/4114 Differential Revision: D8822886 Pulled By: gfosco fbshipit-source-id: c299d008e883c3bb911d2b357a2e9e4423f8e91a
2018-07-13 21:07:53 +00:00
TEST_F(MergeTest, MergeDbTest) {
runTest(test::PerThreadDBPath("merge_testdb"));
Converted db/merge_test.cc to use gtest (#4114) Summary: Picked up a task to convert this to use the gtest framework. It can't be this simple, can it? It works, but should all the std::cout be removed? ``` [$] ~/git/rocksdb [gft !]: ./merge_test [==========] Running 2 tests from 1 test case. [----------] Global test environment set-up. [----------] 2 tests from MergeTest [ RUN ] MergeTest.MergeDbTest Test read-modify-write counters... a: 3 1 2 a: 3 b: 1225 3 Compaction started ... Compaction ended a: 3 b: 1225 Test merge-based counters... a: 3 1 2 a: 3 b: 1225 3 Test merge in memtable... a: 3 1 2 a: 3 b: 1225 3 Test Partial-Merge Test merge-operator not set after reopen [ OK ] MergeTest.MergeDbTest (93 ms) [ RUN ] MergeTest.MergeDbTtlTest Opening database with TTL Test read-modify-write counters... a: 3 1 2 a: 3 b: 1225 3 Compaction started ... Compaction ended a: 3 b: 1225 Test merge-based counters... a: 3 1 2 a: 3 b: 1225 3 Test merge in memtable... Opening database with TTL a: 3 1 2 a: 3 b: 1225 3 Test Partial-Merge Opening database with TTL Opening database with TTL Opening database with TTL Opening database with TTL Test merge-operator not set after reopen [ OK ] MergeTest.MergeDbTtlTest (97 ms) [----------] 2 tests from MergeTest (190 ms total) [----------] Global test environment tear-down [==========] 2 tests from 1 test case ran. (190 ms total) [ PASSED ] 2 tests. ``` Pull Request resolved: https://github.com/facebook/rocksdb/pull/4114 Differential Revision: D8822886 Pulled By: gfosco fbshipit-source-id: c299d008e883c3bb911d2b357a2e9e4423f8e91a
2018-07-13 21:07:53 +00:00
}
TEST_F(MergeTest, MergeDbTtlTest) {
runTest(test::PerThreadDBPath("merge_testdbttl"),
true); // Run test on TTL database
Converted db/merge_test.cc to use gtest (#4114) Summary: Picked up a task to convert this to use the gtest framework. It can't be this simple, can it? It works, but should all the std::cout be removed? ``` [$] ~/git/rocksdb [gft !]: ./merge_test [==========] Running 2 tests from 1 test case. [----------] Global test environment set-up. [----------] 2 tests from MergeTest [ RUN ] MergeTest.MergeDbTest Test read-modify-write counters... a: 3 1 2 a: 3 b: 1225 3 Compaction started ... Compaction ended a: 3 b: 1225 Test merge-based counters... a: 3 1 2 a: 3 b: 1225 3 Test merge in memtable... a: 3 1 2 a: 3 b: 1225 3 Test Partial-Merge Test merge-operator not set after reopen [ OK ] MergeTest.MergeDbTest (93 ms) [ RUN ] MergeTest.MergeDbTtlTest Opening database with TTL Test read-modify-write counters... a: 3 1 2 a: 3 b: 1225 3 Compaction started ... Compaction ended a: 3 b: 1225 Test merge-based counters... a: 3 1 2 a: 3 b: 1225 3 Test merge in memtable... Opening database with TTL a: 3 1 2 a: 3 b: 1225 3 Test Partial-Merge Opening database with TTL Opening database with TTL Opening database with TTL Opening database with TTL Test merge-operator not set after reopen [ OK ] MergeTest.MergeDbTtlTest (97 ms) [----------] 2 tests from MergeTest (190 ms total) [----------] Global test environment tear-down [==========] 2 tests from 1 test case ran. (190 ms total) [ PASSED ] 2 tests. ``` Pull Request resolved: https://github.com/facebook/rocksdb/pull/4114 Differential Revision: D8822886 Pulled By: gfosco fbshipit-source-id: c299d008e883c3bb911d2b357a2e9e4423f8e91a
2018-07-13 21:07:53 +00:00
}
Perform post-flush updates of memtable list in a callback (#6069) Summary: Currently, the following interleaving of events can lead to SuperVersion containing both immutable memtables as well as the resulting L0. This can cause Get to return incorrect result if there are merge operands. This may also affect other operations such as single deletes. ``` time main_thr bg_flush_thr bg_compact_thr compact_thr set_opts_thr 0 | WriteManifest:0 1 | issue compact 2 | wait 3 | Merge(counter) 4 | issue flush 5 | wait 6 | WriteManifest:1 7 | wake up 8 | write manifest 9 | wake up 10 | Get(counter) 11 | remove imm V ``` The reason behind is that: one bg flush thread's installing new `Version` can be batched and performed by another thread that is the "leader" MANIFEST writer. This bg thread removes the memtables from current super version only after `LogAndApply` returns. After the leader MANIFEST writer signals (releasing mutex) this bg flush thread, it is possible that another thread sees this cf with both memtables (whose data have been flushed to the newest L0) and the L0 before this bg flush thread removes the memtables. To address this issue, each bg flush thread can pass a callback function to `LogAndApply`. The callback is responsible for removing the memtables. Therefore, the leader MANIFEST writer can call this callback and remove the memtables before releasing the mutex. Test plan (devserver) ``` $make merge_test $./merge_test --gtest_filter=MergeTest.MergeWithCompactionAndFlush $make check ``` Pull Request resolved: https://github.com/facebook/rocksdb/pull/6069 Reviewed By: cheng-chang Differential Revision: D18790894 Pulled By: riversand963 fbshipit-source-id: e41bd600c0448b4f4b2deb3f7677f95e3076b4ed
2020-10-27 01:20:43 +00:00
TEST_F(MergeTest, MergeWithCompactionAndFlush) {
const std::string dbname =
test::PerThreadDBPath("merge_with_compaction_and_flush");
{
auto db = OpenDb(dbname);
{
MergeBasedCounters counters(db, 0);
testCountersWithFlushAndCompaction(counters, db.get());
}
}
ASSERT_OK(DestroyDB(dbname, Options()));
Perform post-flush updates of memtable list in a callback (#6069) Summary: Currently, the following interleaving of events can lead to SuperVersion containing both immutable memtables as well as the resulting L0. This can cause Get to return incorrect result if there are merge operands. This may also affect other operations such as single deletes. ``` time main_thr bg_flush_thr bg_compact_thr compact_thr set_opts_thr 0 | WriteManifest:0 1 | issue compact 2 | wait 3 | Merge(counter) 4 | issue flush 5 | wait 6 | WriteManifest:1 7 | wake up 8 | write manifest 9 | wake up 10 | Get(counter) 11 | remove imm V ``` The reason behind is that: one bg flush thread's installing new `Version` can be batched and performed by another thread that is the "leader" MANIFEST writer. This bg thread removes the memtables from current super version only after `LogAndApply` returns. After the leader MANIFEST writer signals (releasing mutex) this bg flush thread, it is possible that another thread sees this cf with both memtables (whose data have been flushed to the newest L0) and the L0 before this bg flush thread removes the memtables. To address this issue, each bg flush thread can pass a callback function to `LogAndApply`. The callback is responsible for removing the memtables. Therefore, the leader MANIFEST writer can call this callback and remove the memtables before releasing the mutex. Test plan (devserver) ``` $make merge_test $./merge_test --gtest_filter=MergeTest.MergeWithCompactionAndFlush $make check ``` Pull Request resolved: https://github.com/facebook/rocksdb/pull/6069 Reviewed By: cheng-chang Differential Revision: D18790894 Pulled By: riversand963 fbshipit-source-id: e41bd600c0448b4f4b2deb3f7677f95e3076b4ed
2020-10-27 01:20:43 +00:00
}
Converted db/merge_test.cc to use gtest (#4114) Summary: Picked up a task to convert this to use the gtest framework. It can't be this simple, can it? It works, but should all the std::cout be removed? ``` [$] ~/git/rocksdb [gft !]: ./merge_test [==========] Running 2 tests from 1 test case. [----------] Global test environment set-up. [----------] 2 tests from MergeTest [ RUN ] MergeTest.MergeDbTest Test read-modify-write counters... a: 3 1 2 a: 3 b: 1225 3 Compaction started ... Compaction ended a: 3 b: 1225 Test merge-based counters... a: 3 1 2 a: 3 b: 1225 3 Test merge in memtable... a: 3 1 2 a: 3 b: 1225 3 Test Partial-Merge Test merge-operator not set after reopen [ OK ] MergeTest.MergeDbTest (93 ms) [ RUN ] MergeTest.MergeDbTtlTest Opening database with TTL Test read-modify-write counters... a: 3 1 2 a: 3 b: 1225 3 Compaction started ... Compaction ended a: 3 b: 1225 Test merge-based counters... a: 3 1 2 a: 3 b: 1225 3 Test merge in memtable... Opening database with TTL a: 3 1 2 a: 3 b: 1225 3 Test Partial-Merge Opening database with TTL Opening database with TTL Opening database with TTL Opening database with TTL Test merge-operator not set after reopen [ OK ] MergeTest.MergeDbTtlTest (97 ms) [----------] 2 tests from MergeTest (190 ms total) [----------] Global test environment tear-down [==========] 2 tests from 1 test case ran. (190 ms total) [ PASSED ] 2 tests. ``` Pull Request resolved: https://github.com/facebook/rocksdb/pull/4114 Differential Revision: D8822886 Pulled By: gfosco fbshipit-source-id: c299d008e883c3bb911d2b357a2e9e4423f8e91a
2018-07-13 21:07:53 +00:00
Introduce a wide column aware MergeOperator API (#11807) Summary: Pull Request resolved: https://github.com/facebook/rocksdb/pull/11807 For now, RocksDB has limited support for using merge with wide columns: when a bunch of merge operands have to be applied to a wide-column base value, RocksDB currently passes only the value of the default column to the application's `MergeOperator`, which means there is no way to update any other columns during a merge. As a first step in making this more general, the patch adds a new API `FullMergeV3` to `MergeOperator`. `FullMergeV3`'s interface enables applications to receive a plain, wide-column, or non-existent base value as merge input, and to produce a new plain value, a new wide-column value, or an existing operand as merge output. Note that there are no limitations on the column names and values if the merge result is a wide-column entity. Also, the interface is general in the sense that it makes it possible e.g. for a merge that takes a plain base value and some deltas to produce a wide-column entity as a result. For backward compatibility, the default implementation of `FullMergeV3` falls back to `FullMergeV2` and implements the current logic where merge operands are applied to the default column of the base entity and any other columns are unchanged. (Note that with `FullMergeV3` in the `MergeOperator` interface, this behavior will become customizable.) This patch just introduces the new API and the default backward compatible implementation. I plan to integrate `FullMergeV3` into the query and compaction logic in subsequent diffs. Reviewed By: jaykorean Differential Revision: D49117253 fbshipit-source-id: 109e016f25cd130fc504790818d927bae7fec6bd
2023-09-11 19:13:58 +00:00
TEST_F(MergeTest, FullMergeV3FallbackNewValue) {
// Test that the default FullMergeV3 implementation correctly handles the case
// when FullMergeV2 results in a new value.
const Slice key("foo");
const MergeOperator::MergeOperationInputV3::OperandList operands{
"first", "second", "third"};
constexpr Logger* logger = nullptr;
auto append_operator =
MergeOperators::CreateStringAppendOperator(std::string());
// No existing value
{
MergeOperator::MergeOperationInputV3::ExistingValue existing_value;
const MergeOperator::MergeOperationInputV3 merge_in(
key, std::move(existing_value), operands, logger);
MergeOperator::MergeOperationOutputV3 merge_out;
ASSERT_TRUE(append_operator->FullMergeV3(merge_in, &merge_out));
const auto& result = std::get<std::string>(merge_out.new_value);
ASSERT_EQ(result, operands[0].ToString() + operands[1].ToString() +
operands[2].ToString());
}
// Plain existing value
{
const Slice plain("plain");
MergeOperator::MergeOperationInputV3::ExistingValue existing_value(plain);
const MergeOperator::MergeOperationInputV3 merge_in(
key, std::move(existing_value), operands, logger);
MergeOperator::MergeOperationOutputV3 merge_out;
ASSERT_TRUE(append_operator->FullMergeV3(merge_in, &merge_out));
const auto& result = std::get<std::string>(merge_out.new_value);
ASSERT_EQ(result, plain.ToString() + operands[0].ToString() +
operands[1].ToString() + operands[2].ToString());
}
// Wide-column existing value with default column
{
const WideColumns entity{
{kDefaultWideColumnName, "default"}, {"one", "1"}, {"two", "2"}};
MergeOperator::MergeOperationInputV3::ExistingValue existing_value(entity);
const MergeOperator::MergeOperationInputV3 merge_in(
key, std::move(existing_value), operands, logger);
MergeOperator::MergeOperationOutputV3 merge_out;
ASSERT_TRUE(append_operator->FullMergeV3(merge_in, &merge_out));
const auto& result =
std::get<MergeOperator::MergeOperationOutputV3::NewColumns>(
merge_out.new_value);
ASSERT_EQ(result.size(), entity.size());
ASSERT_EQ(result[0].first, entity[0].name());
ASSERT_EQ(result[0].second,
entity[0].value().ToString() + operands[0].ToString() +
operands[1].ToString() + operands[2].ToString());
ASSERT_EQ(result[1].first, entity[1].name());
ASSERT_EQ(result[1].second, entity[1].value());
ASSERT_EQ(result[2].first, entity[2].name());
ASSERT_EQ(result[2].second, entity[2].value());
}
// Wide-column existing value without default column
{
const WideColumns entity{{"one", "1"}, {"two", "2"}};
MergeOperator::MergeOperationInputV3::ExistingValue existing_value(entity);
const MergeOperator::MergeOperationInputV3 merge_in(
key, std::move(existing_value), operands, logger);
MergeOperator::MergeOperationOutputV3 merge_out;
ASSERT_TRUE(append_operator->FullMergeV3(merge_in, &merge_out));
const auto& result =
std::get<MergeOperator::MergeOperationOutputV3::NewColumns>(
merge_out.new_value);
ASSERT_EQ(result.size(), entity.size() + 1);
ASSERT_EQ(result[0].first, kDefaultWideColumnName);
ASSERT_EQ(result[0].second, operands[0].ToString() +
operands[1].ToString() +
operands[2].ToString());
ASSERT_EQ(result[1].first, entity[0].name());
ASSERT_EQ(result[1].second, entity[0].value());
ASSERT_EQ(result[2].first, entity[1].name());
ASSERT_EQ(result[2].second, entity[1].value());
}
}
TEST_F(MergeTest, FullMergeV3FallbackExistingOperand) {
// Test that the default FullMergeV3 implementation correctly handles the case
// when FullMergeV2 results in an existing operand.
const Slice key("foo");
const MergeOperator::MergeOperationInputV3::OperandList operands{
"first", "second", "third"};
constexpr Logger* logger = nullptr;
auto put_operator = MergeOperators::CreatePutOperator();
// No existing value
{
MergeOperator::MergeOperationInputV3::ExistingValue existing_value;
const MergeOperator::MergeOperationInputV3 merge_in(
key, std::move(existing_value), operands, logger);
MergeOperator::MergeOperationOutputV3 merge_out;
ASSERT_TRUE(put_operator->FullMergeV3(merge_in, &merge_out));
const auto& result = std::get<Slice>(merge_out.new_value);
ASSERT_EQ(result.data(), operands.back().data());
ASSERT_EQ(result.size(), operands.back().size());
}
// Plain existing value
{
const Slice plain("plain");
MergeOperator::MergeOperationInputV3::ExistingValue existing_value(plain);
const MergeOperator::MergeOperationInputV3 merge_in(
key, std::move(existing_value), operands, logger);
MergeOperator::MergeOperationOutputV3 merge_out;
ASSERT_TRUE(put_operator->FullMergeV3(merge_in, &merge_out));
const auto& result = std::get<Slice>(merge_out.new_value);
ASSERT_EQ(result.data(), operands.back().data());
ASSERT_EQ(result.size(), operands.back().size());
}
// Wide-column existing value with default column
{
const WideColumns entity{
{kDefaultWideColumnName, "default"}, {"one", "1"}, {"two", "2"}};
MergeOperator::MergeOperationInputV3::ExistingValue existing_value(entity);
const MergeOperator::MergeOperationInputV3 merge_in(
key, std::move(existing_value), operands, logger);
MergeOperator::MergeOperationOutputV3 merge_out;
ASSERT_TRUE(put_operator->FullMergeV3(merge_in, &merge_out));
const auto& result =
std::get<MergeOperator::MergeOperationOutputV3::NewColumns>(
merge_out.new_value);
ASSERT_EQ(result.size(), entity.size());
ASSERT_EQ(result[0].first, entity[0].name());
ASSERT_EQ(result[0].second, operands.back());
ASSERT_EQ(result[1].first, entity[1].name());
ASSERT_EQ(result[1].second, entity[1].value());
ASSERT_EQ(result[2].first, entity[2].name());
ASSERT_EQ(result[2].second, entity[2].value());
}
// Wide-column existing value without default column
{
const WideColumns entity{{"one", "1"}, {"two", "2"}};
MergeOperator::MergeOperationInputV3::ExistingValue existing_value(entity);
const MergeOperator::MergeOperationInputV3 merge_in(
key, std::move(existing_value), operands, logger);
MergeOperator::MergeOperationOutputV3 merge_out;
ASSERT_TRUE(put_operator->FullMergeV3(merge_in, &merge_out));
const auto& result =
std::get<MergeOperator::MergeOperationOutputV3::NewColumns>(
merge_out.new_value);
ASSERT_EQ(result.size(), entity.size() + 1);
ASSERT_EQ(result[0].first, kDefaultWideColumnName);
ASSERT_EQ(result[0].second, operands.back());
ASSERT_EQ(result[1].first, entity[0].name());
ASSERT_EQ(result[1].second, entity[0].value());
ASSERT_EQ(result[2].first, entity[1].name());
ASSERT_EQ(result[2].second, entity[1].value());
}
}
TEST_F(MergeTest, FullMergeV3FallbackFailure) {
// Test that the default FullMergeV3 implementation correctly handles the case
// when FullMergeV2 fails.
const Slice key("foo");
const MergeOperator::MergeOperationInputV3::OperandList operands{
"first", "second", "third"};
constexpr Logger* logger = nullptr;
class FailMergeOperator : public MergeOperator {
public:
bool FullMergeV2(const MergeOperationInput& /* merge_in */,
MergeOperationOutput* merge_out) const override {
assert(merge_out);
merge_out->op_failure_scope = OpFailureScope::kMustMerge;
return false;
}
const char* Name() const override { return "FailMergeOperator"; }
};
FailMergeOperator fail_operator;
// No existing value
{
MergeOperator::MergeOperationInputV3::ExistingValue existing_value;
const MergeOperator::MergeOperationInputV3 merge_in(
key, std::move(existing_value), operands, logger);
MergeOperator::MergeOperationOutputV3 merge_out;
ASSERT_FALSE(fail_operator.FullMergeV3(merge_in, &merge_out));
ASSERT_EQ(merge_out.op_failure_scope,
MergeOperator::OpFailureScope::kMustMerge);
}
// Plain existing value
{
const Slice plain("plain");
MergeOperator::MergeOperationInputV3::ExistingValue existing_value(plain);
const MergeOperator::MergeOperationInputV3 merge_in(
key, std::move(existing_value), operands, logger);
MergeOperator::MergeOperationOutputV3 merge_out;
ASSERT_FALSE(fail_operator.FullMergeV3(merge_in, &merge_out));
ASSERT_EQ(merge_out.op_failure_scope,
MergeOperator::OpFailureScope::kMustMerge);
}
// Wide-column existing value with default column
{
const WideColumns entity{
{kDefaultWideColumnName, "default"}, {"one", "1"}, {"two", "2"}};
MergeOperator::MergeOperationInputV3::ExistingValue existing_value(entity);
const MergeOperator::MergeOperationInputV3 merge_in(
key, std::move(existing_value), operands, logger);
MergeOperator::MergeOperationOutputV3 merge_out;
ASSERT_FALSE(fail_operator.FullMergeV3(merge_in, &merge_out));
ASSERT_EQ(merge_out.op_failure_scope,
MergeOperator::OpFailureScope::kMustMerge);
}
// Wide-column existing value without default column
{
const WideColumns entity{{"one", "1"}, {"two", "2"}};
MergeOperator::MergeOperationInputV3::ExistingValue existing_value(entity);
const MergeOperator::MergeOperationInputV3 merge_in(
key, std::move(existing_value), operands, logger);
MergeOperator::MergeOperationOutputV3 merge_out;
ASSERT_FALSE(fail_operator.FullMergeV3(merge_in, &merge_out));
ASSERT_EQ(merge_out.op_failure_scope,
MergeOperator::OpFailureScope::kMustMerge);
}
}
} // namespace ROCKSDB_NAMESPACE
Converted db/merge_test.cc to use gtest (#4114) Summary: Picked up a task to convert this to use the gtest framework. It can't be this simple, can it? It works, but should all the std::cout be removed? ``` [$] ~/git/rocksdb [gft !]: ./merge_test [==========] Running 2 tests from 1 test case. [----------] Global test environment set-up. [----------] 2 tests from MergeTest [ RUN ] MergeTest.MergeDbTest Test read-modify-write counters... a: 3 1 2 a: 3 b: 1225 3 Compaction started ... Compaction ended a: 3 b: 1225 Test merge-based counters... a: 3 1 2 a: 3 b: 1225 3 Test merge in memtable... a: 3 1 2 a: 3 b: 1225 3 Test Partial-Merge Test merge-operator not set after reopen [ OK ] MergeTest.MergeDbTest (93 ms) [ RUN ] MergeTest.MergeDbTtlTest Opening database with TTL Test read-modify-write counters... a: 3 1 2 a: 3 b: 1225 3 Compaction started ... Compaction ended a: 3 b: 1225 Test merge-based counters... a: 3 1 2 a: 3 b: 1225 3 Test merge in memtable... Opening database with TTL a: 3 1 2 a: 3 b: 1225 3 Test Partial-Merge Opening database with TTL Opening database with TTL Opening database with TTL Opening database with TTL Test merge-operator not set after reopen [ OK ] MergeTest.MergeDbTtlTest (97 ms) [----------] 2 tests from MergeTest (190 ms total) [----------] Global test environment tear-down [==========] 2 tests from 1 test case ran. (190 ms total) [ PASSED ] 2 tests. ``` Pull Request resolved: https://github.com/facebook/rocksdb/pull/4114 Differential Revision: D8822886 Pulled By: gfosco fbshipit-source-id: c299d008e883c3bb911d2b357a2e9e4423f8e91a
2018-07-13 21:07:53 +00:00
int main(int argc, char** argv) {
ROCKSDB_NAMESPACE::use_compression = false;
Converted db/merge_test.cc to use gtest (#4114) Summary: Picked up a task to convert this to use the gtest framework. It can't be this simple, can it? It works, but should all the std::cout be removed? ``` [$] ~/git/rocksdb [gft !]: ./merge_test [==========] Running 2 tests from 1 test case. [----------] Global test environment set-up. [----------] 2 tests from MergeTest [ RUN ] MergeTest.MergeDbTest Test read-modify-write counters... a: 3 1 2 a: 3 b: 1225 3 Compaction started ... Compaction ended a: 3 b: 1225 Test merge-based counters... a: 3 1 2 a: 3 b: 1225 3 Test merge in memtable... a: 3 1 2 a: 3 b: 1225 3 Test Partial-Merge Test merge-operator not set after reopen [ OK ] MergeTest.MergeDbTest (93 ms) [ RUN ] MergeTest.MergeDbTtlTest Opening database with TTL Test read-modify-write counters... a: 3 1 2 a: 3 b: 1225 3 Compaction started ... Compaction ended a: 3 b: 1225 Test merge-based counters... a: 3 1 2 a: 3 b: 1225 3 Test merge in memtable... Opening database with TTL a: 3 1 2 a: 3 b: 1225 3 Test Partial-Merge Opening database with TTL Opening database with TTL Opening database with TTL Opening database with TTL Test merge-operator not set after reopen [ OK ] MergeTest.MergeDbTtlTest (97 ms) [----------] 2 tests from MergeTest (190 ms total) [----------] Global test environment tear-down [==========] 2 tests from 1 test case ran. (190 ms total) [ PASSED ] 2 tests. ``` Pull Request resolved: https://github.com/facebook/rocksdb/pull/4114 Differential Revision: D8822886 Pulled By: gfosco fbshipit-source-id: c299d008e883c3bb911d2b357a2e9e4423f8e91a
2018-07-13 21:07:53 +00:00
if (argc > 1) {
ROCKSDB_NAMESPACE::use_compression = true;
Converted db/merge_test.cc to use gtest (#4114) Summary: Picked up a task to convert this to use the gtest framework. It can't be this simple, can it? It works, but should all the std::cout be removed? ``` [$] ~/git/rocksdb [gft !]: ./merge_test [==========] Running 2 tests from 1 test case. [----------] Global test environment set-up. [----------] 2 tests from MergeTest [ RUN ] MergeTest.MergeDbTest Test read-modify-write counters... a: 3 1 2 a: 3 b: 1225 3 Compaction started ... Compaction ended a: 3 b: 1225 Test merge-based counters... a: 3 1 2 a: 3 b: 1225 3 Test merge in memtable... a: 3 1 2 a: 3 b: 1225 3 Test Partial-Merge Test merge-operator not set after reopen [ OK ] MergeTest.MergeDbTest (93 ms) [ RUN ] MergeTest.MergeDbTtlTest Opening database with TTL Test read-modify-write counters... a: 3 1 2 a: 3 b: 1225 3 Compaction started ... Compaction ended a: 3 b: 1225 Test merge-based counters... a: 3 1 2 a: 3 b: 1225 3 Test merge in memtable... Opening database with TTL a: 3 1 2 a: 3 b: 1225 3 Test Partial-Merge Opening database with TTL Opening database with TTL Opening database with TTL Opening database with TTL Test merge-operator not set after reopen [ OK ] MergeTest.MergeDbTtlTest (97 ms) [----------] 2 tests from MergeTest (190 ms total) [----------] Global test environment tear-down [==========] 2 tests from 1 test case ran. (190 ms total) [ PASSED ] 2 tests. ``` Pull Request resolved: https://github.com/facebook/rocksdb/pull/4114 Differential Revision: D8822886 Pulled By: gfosco fbshipit-source-id: c299d008e883c3bb911d2b357a2e9e4423f8e91a
2018-07-13 21:07:53 +00:00
}
ROCKSDB_NAMESPACE::port::InstallStackTraceHandler();
Converted db/merge_test.cc to use gtest (#4114) Summary: Picked up a task to convert this to use the gtest framework. It can't be this simple, can it? It works, but should all the std::cout be removed? ``` [$] ~/git/rocksdb [gft !]: ./merge_test [==========] Running 2 tests from 1 test case. [----------] Global test environment set-up. [----------] 2 tests from MergeTest [ RUN ] MergeTest.MergeDbTest Test read-modify-write counters... a: 3 1 2 a: 3 b: 1225 3 Compaction started ... Compaction ended a: 3 b: 1225 Test merge-based counters... a: 3 1 2 a: 3 b: 1225 3 Test merge in memtable... a: 3 1 2 a: 3 b: 1225 3 Test Partial-Merge Test merge-operator not set after reopen [ OK ] MergeTest.MergeDbTest (93 ms) [ RUN ] MergeTest.MergeDbTtlTest Opening database with TTL Test read-modify-write counters... a: 3 1 2 a: 3 b: 1225 3 Compaction started ... Compaction ended a: 3 b: 1225 Test merge-based counters... a: 3 1 2 a: 3 b: 1225 3 Test merge in memtable... Opening database with TTL a: 3 1 2 a: 3 b: 1225 3 Test Partial-Merge Opening database with TTL Opening database with TTL Opening database with TTL Opening database with TTL Test merge-operator not set after reopen [ OK ] MergeTest.MergeDbTtlTest (97 ms) [----------] 2 tests from MergeTest (190 ms total) [----------] Global test environment tear-down [==========] 2 tests from 1 test case ran. (190 ms total) [ PASSED ] 2 tests. ``` Pull Request resolved: https://github.com/facebook/rocksdb/pull/4114 Differential Revision: D8822886 Pulled By: gfosco fbshipit-source-id: c299d008e883c3bb911d2b357a2e9e4423f8e91a
2018-07-13 21:07:53 +00:00
::testing::InitGoogleTest(&argc, argv);
return RUN_ALL_TESTS();
}