2016-02-10 00:12:00 +01:00
|
|
|
// Copyright (c) 2011-present, Facebook, Inc. All rights reserved.
|
2017-07-16 01:03:42 +02:00
|
|
|
// This source code is licensed under both the GPLv2 (found in the
|
|
|
|
// COPYING file in the root directory) and Apache 2.0 License
|
|
|
|
// (found in the LICENSE.Apache file in the root directory).
|
2015-05-29 23:36:35 +02:00
|
|
|
|
|
|
|
#ifndef ROCKSDB_LITE
|
|
|
|
|
2017-05-22 19:21:38 +02:00
|
|
|
#include <atomic>
|
2016-12-16 20:17:26 +01:00
|
|
|
#include <functional>
|
2015-05-29 23:36:35 +02:00
|
|
|
#include <string>
|
2016-02-05 19:44:13 +01:00
|
|
|
#include <utility>
|
|
|
|
#include <vector>
|
2015-05-29 23:36:35 +02:00
|
|
|
|
|
|
|
#include "db/db_impl.h"
|
|
|
|
#include "db/write_callback.h"
|
|
|
|
#include "rocksdb/db.h"
|
|
|
|
#include "rocksdb/write_batch.h"
|
2017-02-06 23:43:55 +01:00
|
|
|
#include "port/port.h"
|
2017-04-06 22:59:31 +02:00
|
|
|
#include "util/random.h"
|
2016-02-05 19:44:13 +01:00
|
|
|
#include "util/sync_point.h"
|
2015-05-29 23:36:35 +02:00
|
|
|
#include "util/testharness.h"
|
|
|
|
|
|
|
|
using std::string;
|
|
|
|
|
|
|
|
namespace rocksdb {
|
|
|
|
|
|
|
|
class WriteCallbackTest : public testing::Test {
|
|
|
|
public:
|
|
|
|
string dbname;
|
|
|
|
|
|
|
|
WriteCallbackTest() {
|
|
|
|
dbname = test::TmpDir() + "/write_callback_testdb";
|
|
|
|
}
|
|
|
|
};
|
|
|
|
|
|
|
|
class WriteCallbackTestWriteCallback1 : public WriteCallback {
|
|
|
|
public:
|
|
|
|
bool was_called = false;
|
|
|
|
|
|
|
|
Status Callback(DB *db) override {
|
|
|
|
was_called = true;
|
|
|
|
|
|
|
|
// Make sure db is a DBImpl
|
|
|
|
DBImpl* db_impl = dynamic_cast<DBImpl*> (db);
|
|
|
|
if (db_impl == nullptr) {
|
|
|
|
return Status::InvalidArgument("");
|
|
|
|
}
|
|
|
|
|
|
|
|
return Status::OK();
|
|
|
|
}
|
2016-02-05 19:44:13 +01:00
|
|
|
|
|
|
|
bool AllowWriteBatching() override { return true; }
|
2015-05-29 23:36:35 +02:00
|
|
|
};
|
|
|
|
|
|
|
|
class WriteCallbackTestWriteCallback2 : public WriteCallback {
|
|
|
|
public:
|
2017-07-22 03:13:59 +02:00
|
|
|
Status Callback(DB *db) override {
|
|
|
|
return Status::Busy();
|
|
|
|
}
|
2016-02-05 19:44:13 +01:00
|
|
|
bool AllowWriteBatching() override { return true; }
|
|
|
|
};
|
|
|
|
|
|
|
|
class MockWriteCallback : public WriteCallback {
|
|
|
|
public:
|
|
|
|
bool should_fail_ = false;
|
|
|
|
bool allow_batching_ = false;
|
2017-05-22 19:21:38 +02:00
|
|
|
std::atomic<bool> was_called_{false};
|
|
|
|
|
|
|
|
MockWriteCallback() {}
|
|
|
|
|
|
|
|
MockWriteCallback(const MockWriteCallback& other) {
|
|
|
|
should_fail_ = other.should_fail_;
|
|
|
|
allow_batching_ = other.allow_batching_;
|
|
|
|
was_called_.store(other.was_called_.load());
|
|
|
|
}
|
2016-02-05 19:44:13 +01:00
|
|
|
|
2017-07-22 03:13:59 +02:00
|
|
|
Status Callback(DB* db) override {
|
2017-05-22 19:21:38 +02:00
|
|
|
was_called_.store(true);
|
2016-02-05 19:44:13 +01:00
|
|
|
if (should_fail_) {
|
|
|
|
return Status::Busy();
|
|
|
|
} else {
|
|
|
|
return Status::OK();
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
bool AllowWriteBatching() override { return allow_batching_; }
|
2015-05-29 23:36:35 +02:00
|
|
|
};
|
|
|
|
|
2016-02-05 19:44:13 +01:00
|
|
|
TEST_F(WriteCallbackTest, WriteWithCallbackTest) {
|
|
|
|
struct WriteOP {
|
|
|
|
WriteOP(bool should_fail = false) { callback_.should_fail_ = should_fail; }
|
|
|
|
|
|
|
|
void Put(const string& key, const string& val) {
|
|
|
|
kvs_.push_back(std::make_pair(key, val));
|
|
|
|
write_batch_.Put(key, val);
|
|
|
|
}
|
|
|
|
|
|
|
|
void Clear() {
|
|
|
|
kvs_.clear();
|
|
|
|
write_batch_.Clear();
|
2017-05-22 19:21:38 +02:00
|
|
|
callback_.was_called_.store(false);
|
2016-02-05 19:44:13 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
MockWriteCallback callback_;
|
|
|
|
WriteBatch write_batch_;
|
|
|
|
std::vector<std::pair<string, string>> kvs_;
|
|
|
|
};
|
|
|
|
|
2017-07-26 01:40:21 +02:00
|
|
|
// In each scenario we'll launch multiple threads to write.
|
|
|
|
// The size of each array equals to number of threads, and
|
|
|
|
// each boolean in it denote whether callback of corresponding
|
|
|
|
// thread should succeed or fail.
|
2016-02-05 19:44:13 +01:00
|
|
|
std::vector<std::vector<WriteOP>> write_scenarios = {
|
|
|
|
{true},
|
|
|
|
{false},
|
|
|
|
{false, false},
|
|
|
|
{true, true},
|
|
|
|
{true, false},
|
|
|
|
{false, true},
|
|
|
|
{false, false, false},
|
|
|
|
{true, true, true},
|
|
|
|
{false, true, false},
|
|
|
|
{true, false, true},
|
|
|
|
{true, false, false, false, false},
|
|
|
|
{false, false, false, false, true},
|
|
|
|
{false, false, true, false, true},
|
|
|
|
};
|
|
|
|
|
2017-06-24 23:06:43 +02:00
|
|
|
for (auto& two_queues : {true, false}) {
|
|
|
|
for (auto& allow_parallel : {true, false}) {
|
|
|
|
for (auto& allow_batching : {true, false}) {
|
|
|
|
for (auto& enable_WAL : {true, false}) {
|
|
|
|
for (auto& enable_pipelined_write : {true, false}) {
|
|
|
|
for (auto& write_group : write_scenarios) {
|
|
|
|
Options options;
|
|
|
|
options.create_if_missing = true;
|
|
|
|
options.allow_concurrent_memtable_write = allow_parallel;
|
|
|
|
options.enable_pipelined_write = enable_pipelined_write;
|
2017-11-11 02:18:01 +01:00
|
|
|
options.two_write_queues = two_queues;
|
|
|
|
if (options.enable_pipelined_write && options.two_write_queues) {
|
2017-09-12 23:48:19 +02:00
|
|
|
// This combination is not supported
|
|
|
|
continue;
|
|
|
|
}
|
2017-06-24 23:06:43 +02:00
|
|
|
|
|
|
|
ReadOptions read_options;
|
|
|
|
DB* db;
|
|
|
|
DBImpl* db_impl;
|
|
|
|
|
|
|
|
DestroyDB(dbname, options);
|
|
|
|
ASSERT_OK(DB::Open(options, dbname, &db));
|
|
|
|
|
|
|
|
db_impl = dynamic_cast<DBImpl*>(db);
|
|
|
|
ASSERT_TRUE(db_impl);
|
|
|
|
|
2017-07-26 01:40:21 +02:00
|
|
|
// Writers that have called JoinBatchGroup.
|
|
|
|
std::atomic<uint64_t> threads_joining(0);
|
|
|
|
// Writers that have linked to the queue
|
|
|
|
std::atomic<uint64_t> threads_linked(0);
|
|
|
|
// Writers that pass WriteThread::JoinBatchGroup:Wait sync-point.
|
|
|
|
std::atomic<uint64_t> threads_verified(0);
|
|
|
|
|
2017-06-24 23:06:43 +02:00
|
|
|
std::atomic<uint64_t> seq(db_impl->GetLatestSequenceNumber());
|
|
|
|
ASSERT_EQ(db_impl->GetLatestSequenceNumber(), 0);
|
|
|
|
|
2017-07-26 01:40:21 +02:00
|
|
|
rocksdb::SyncPoint::GetInstance()->SetCallBack(
|
|
|
|
"WriteThread::JoinBatchGroup:Start", [&](void*) {
|
|
|
|
uint64_t cur_threads_joining = threads_joining.fetch_add(1);
|
|
|
|
// Wait for the last joined writer to link to the queue.
|
|
|
|
// In this way the writers link to the queue one by one.
|
|
|
|
// This allows us to confidently detect the first writer
|
|
|
|
// who increases threads_linked as the leader.
|
|
|
|
while (threads_linked.load() < cur_threads_joining) {
|
|
|
|
}
|
|
|
|
});
|
|
|
|
|
|
|
|
// Verification once writers call JoinBatchGroup.
|
2017-06-24 23:06:43 +02:00
|
|
|
rocksdb::SyncPoint::GetInstance()->SetCallBack(
|
|
|
|
"WriteThread::JoinBatchGroup:Wait", [&](void* arg) {
|
2017-07-26 01:40:21 +02:00
|
|
|
uint64_t cur_threads_linked = threads_linked.fetch_add(1);
|
2017-06-24 23:06:43 +02:00
|
|
|
bool is_leader = false;
|
|
|
|
bool is_last = false;
|
|
|
|
|
|
|
|
// who am i
|
2017-07-26 01:40:21 +02:00
|
|
|
is_leader = (cur_threads_linked == 0);
|
|
|
|
is_last = (cur_threads_linked == write_group.size() - 1);
|
2017-06-24 23:06:43 +02:00
|
|
|
|
|
|
|
// check my state
|
|
|
|
auto* writer = reinterpret_cast<WriteThread::Writer*>(arg);
|
|
|
|
|
|
|
|
if (is_leader) {
|
|
|
|
ASSERT_TRUE(writer->state ==
|
|
|
|
WriteThread::State::STATE_GROUP_LEADER);
|
|
|
|
} else {
|
|
|
|
ASSERT_TRUE(writer->state ==
|
|
|
|
WriteThread::State::STATE_INIT);
|
|
|
|
}
|
|
|
|
|
|
|
|
// (meta test) the first WriteOP should indeed be the first
|
|
|
|
// and the last should be the last (all others can be out of
|
|
|
|
// order)
|
|
|
|
if (is_leader) {
|
|
|
|
ASSERT_TRUE(writer->callback->Callback(nullptr).ok() ==
|
|
|
|
!write_group.front().callback_.should_fail_);
|
|
|
|
} else if (is_last) {
|
|
|
|
ASSERT_TRUE(writer->callback->Callback(nullptr).ok() ==
|
|
|
|
!write_group.back().callback_.should_fail_);
|
|
|
|
}
|
|
|
|
|
2017-07-26 01:40:21 +02:00
|
|
|
threads_verified.fetch_add(1);
|
|
|
|
// Wait here until all verification in this sync-point
|
|
|
|
// callback finish for all writers.
|
|
|
|
while (threads_verified.load() < write_group.size()) {
|
2017-06-24 23:06:43 +02:00
|
|
|
}
|
|
|
|
});
|
|
|
|
|
|
|
|
rocksdb::SyncPoint::GetInstance()->SetCallBack(
|
|
|
|
"WriteThread::JoinBatchGroup:DoneWaiting", [&](void* arg) {
|
|
|
|
// check my state
|
|
|
|
auto* writer = reinterpret_cast<WriteThread::Writer*>(arg);
|
|
|
|
|
|
|
|
if (!allow_batching) {
|
|
|
|
// no batching so everyone should be a leader
|
|
|
|
ASSERT_TRUE(writer->state ==
|
|
|
|
WriteThread::State::STATE_GROUP_LEADER);
|
|
|
|
} else if (!allow_parallel) {
|
|
|
|
ASSERT_TRUE(writer->state ==
|
|
|
|
WriteThread::State::STATE_COMPLETED ||
|
|
|
|
(enable_pipelined_write &&
|
|
|
|
writer->state ==
|
|
|
|
WriteThread::State::
|
|
|
|
STATE_MEMTABLE_WRITER_LEADER));
|
|
|
|
}
|
|
|
|
});
|
|
|
|
|
|
|
|
std::atomic<uint32_t> thread_num(0);
|
|
|
|
std::atomic<char> dummy_key(0);
|
2017-07-26 01:40:21 +02:00
|
|
|
|
|
|
|
// Each write thread create a random write batch and write to DB
|
|
|
|
// with a write callback.
|
2017-06-24 23:06:43 +02:00
|
|
|
std::function<void()> write_with_callback_func = [&]() {
|
|
|
|
uint32_t i = thread_num.fetch_add(1);
|
|
|
|
Random rnd(i);
|
|
|
|
|
|
|
|
// leaders gotta lead
|
2017-07-26 01:40:21 +02:00
|
|
|
while (i > 0 && threads_verified.load() < 1) {
|
2017-06-24 23:06:43 +02:00
|
|
|
}
|
2017-05-19 23:24:23 +02:00
|
|
|
|
2017-06-24 23:06:43 +02:00
|
|
|
// loser has to lose
|
|
|
|
while (i == write_group.size() - 1 &&
|
2017-07-26 01:40:21 +02:00
|
|
|
threads_verified.load() < write_group.size() - 1) {
|
2017-06-24 23:06:43 +02:00
|
|
|
}
|
2017-05-19 23:24:23 +02:00
|
|
|
|
2017-06-24 23:06:43 +02:00
|
|
|
auto& write_op = write_group.at(i);
|
|
|
|
write_op.Clear();
|
|
|
|
write_op.callback_.allow_batching_ = allow_batching;
|
2017-05-19 23:24:23 +02:00
|
|
|
|
2017-06-24 23:06:43 +02:00
|
|
|
// insert some keys
|
|
|
|
for (uint32_t j = 0; j < rnd.Next() % 50; j++) {
|
|
|
|
// grab unique key
|
2017-07-26 01:40:21 +02:00
|
|
|
char my_key = dummy_key.fetch_add(1);
|
2016-02-16 21:39:55 +01:00
|
|
|
|
2017-06-24 23:06:43 +02:00
|
|
|
string skey(5, my_key);
|
|
|
|
string sval(10, my_key);
|
|
|
|
write_op.Put(skey, sval);
|
2016-02-16 21:39:55 +01:00
|
|
|
|
2017-06-24 23:06:43 +02:00
|
|
|
if (!write_op.callback_.should_fail_) {
|
|
|
|
seq.fetch_add(1);
|
|
|
|
}
|
2016-02-16 21:39:55 +01:00
|
|
|
}
|
2016-02-05 19:44:13 +01:00
|
|
|
|
2017-06-24 23:06:43 +02:00
|
|
|
WriteOptions woptions;
|
|
|
|
woptions.disableWAL = !enable_WAL;
|
|
|
|
woptions.sync = enable_WAL;
|
|
|
|
Status s = db_impl->WriteWithCallback(
|
|
|
|
woptions, &write_op.write_batch_, &write_op.callback_);
|
2016-02-05 19:44:13 +01:00
|
|
|
|
2017-06-24 23:06:43 +02:00
|
|
|
if (write_op.callback_.should_fail_) {
|
|
|
|
ASSERT_TRUE(s.IsBusy());
|
|
|
|
} else {
|
|
|
|
ASSERT_OK(s);
|
|
|
|
}
|
|
|
|
};
|
2016-02-05 19:44:13 +01:00
|
|
|
|
2017-06-24 23:06:43 +02:00
|
|
|
rocksdb::SyncPoint::GetInstance()->EnableProcessing();
|
2016-02-05 19:44:13 +01:00
|
|
|
|
2017-06-24 23:06:43 +02:00
|
|
|
// do all the writes
|
|
|
|
std::vector<port::Thread> threads;
|
|
|
|
for (uint32_t i = 0; i < write_group.size(); i++) {
|
|
|
|
threads.emplace_back(write_with_callback_func);
|
|
|
|
}
|
|
|
|
for (auto& t : threads) {
|
|
|
|
t.join();
|
|
|
|
}
|
2016-02-05 19:44:13 +01:00
|
|
|
|
2017-06-24 23:06:43 +02:00
|
|
|
rocksdb::SyncPoint::GetInstance()->DisableProcessing();
|
2016-02-05 19:44:13 +01:00
|
|
|
|
2017-06-24 23:06:43 +02:00
|
|
|
// check for keys
|
|
|
|
string value;
|
|
|
|
for (auto& w : write_group) {
|
|
|
|
ASSERT_TRUE(w.callback_.was_called_.load());
|
|
|
|
for (auto& kvp : w.kvs_) {
|
|
|
|
if (w.callback_.should_fail_) {
|
|
|
|
ASSERT_TRUE(
|
|
|
|
db->Get(read_options, kvp.first, &value).IsNotFound());
|
|
|
|
} else {
|
|
|
|
ASSERT_OK(db->Get(read_options, kvp.first, &value));
|
|
|
|
ASSERT_EQ(value, kvp.second);
|
|
|
|
}
|
2017-05-19 23:24:23 +02:00
|
|
|
}
|
2016-02-16 21:39:55 +01:00
|
|
|
}
|
2016-02-05 19:44:13 +01:00
|
|
|
|
2017-06-24 23:06:43 +02:00
|
|
|
ASSERT_EQ(seq.load(), db_impl->GetLatestSequenceNumber());
|
2016-02-05 19:44:13 +01:00
|
|
|
|
2017-06-24 23:06:43 +02:00
|
|
|
delete db;
|
|
|
|
DestroyDB(dbname, options);
|
|
|
|
}
|
2017-05-19 23:24:23 +02:00
|
|
|
}
|
2016-02-16 21:39:55 +01:00
|
|
|
}
|
2016-02-05 19:44:13 +01:00
|
|
|
}
|
|
|
|
}
|
2017-06-24 23:06:43 +02:00
|
|
|
}
|
2016-02-05 19:44:13 +01:00
|
|
|
}
|
|
|
|
|
2015-05-29 23:36:35 +02:00
|
|
|
TEST_F(WriteCallbackTest, WriteCallBackTest) {
|
|
|
|
Options options;
|
|
|
|
WriteOptions write_options;
|
|
|
|
ReadOptions read_options;
|
|
|
|
string value;
|
|
|
|
DB* db;
|
|
|
|
DBImpl* db_impl;
|
|
|
|
|
2016-04-16 02:13:12 +02:00
|
|
|
DestroyDB(dbname, options);
|
|
|
|
|
2015-05-29 23:36:35 +02:00
|
|
|
options.create_if_missing = true;
|
|
|
|
Status s = DB::Open(options, dbname, &db);
|
|
|
|
ASSERT_OK(s);
|
|
|
|
|
|
|
|
db_impl = dynamic_cast<DBImpl*> (db);
|
|
|
|
ASSERT_TRUE(db_impl);
|
|
|
|
|
|
|
|
WriteBatch wb;
|
|
|
|
|
|
|
|
wb.Put("a", "value.a");
|
|
|
|
wb.Delete("x");
|
|
|
|
|
|
|
|
// Test a simple Write
|
|
|
|
s = db->Write(write_options, &wb);
|
|
|
|
ASSERT_OK(s);
|
|
|
|
|
|
|
|
s = db->Get(read_options, "a", &value);
|
|
|
|
ASSERT_OK(s);
|
|
|
|
ASSERT_EQ("value.a", value);
|
|
|
|
|
|
|
|
// Test WriteWithCallback
|
|
|
|
WriteCallbackTestWriteCallback1 callback1;
|
|
|
|
WriteBatch wb2;
|
|
|
|
|
|
|
|
wb2.Put("a", "value.a2");
|
|
|
|
|
|
|
|
s = db_impl->WriteWithCallback(write_options, &wb2, &callback1);
|
|
|
|
ASSERT_OK(s);
|
|
|
|
ASSERT_TRUE(callback1.was_called);
|
|
|
|
|
|
|
|
s = db->Get(read_options, "a", &value);
|
|
|
|
ASSERT_OK(s);
|
|
|
|
ASSERT_EQ("value.a2", value);
|
|
|
|
|
|
|
|
// Test WriteWithCallback for a callback that fails
|
|
|
|
WriteCallbackTestWriteCallback2 callback2;
|
|
|
|
WriteBatch wb3;
|
|
|
|
|
|
|
|
wb3.Put("a", "value.a3");
|
|
|
|
|
|
|
|
s = db_impl->WriteWithCallback(write_options, &wb3, &callback2);
|
|
|
|
ASSERT_NOK(s);
|
|
|
|
|
|
|
|
s = db->Get(read_options, "a", &value);
|
|
|
|
ASSERT_OK(s);
|
|
|
|
ASSERT_EQ("value.a2", value);
|
|
|
|
|
|
|
|
delete db;
|
|
|
|
DestroyDB(dbname, options);
|
|
|
|
}
|
|
|
|
|
|
|
|
} // namespace rocksdb
|
|
|
|
|
|
|
|
int main(int argc, char** argv) {
|
|
|
|
::testing::InitGoogleTest(&argc, argv);
|
|
|
|
return RUN_ALL_TESTS();
|
|
|
|
}
|
|
|
|
|
2015-07-20 19:54:15 +02:00
|
|
|
#else
|
|
|
|
#include <stdio.h>
|
|
|
|
|
|
|
|
int main(int argc, char** argv) {
|
|
|
|
fprintf(stderr,
|
|
|
|
"SKIPPED as WriteWithCallback is not supported in ROCKSDB_LITE\n");
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
#endif // !ROCKSDB_LITE
|