You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
rocksdb/db/db_wal_test.cc

2434 lines
86 KiB

// 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).
//
// Copyright (c) 2011 The LevelDB Authors. All rights reserved.
// Use of this source code is governed by a BSD-style license that can be
// found in the LICENSE file. See the AUTHORS file for names of contributors.
#include "db/db_test_util.h"
#include "options/options_helper.h"
#include "port/port.h"
#include "port/stack_trace.h"
#include "rocksdb/file_system.h"
#include "test_util/sync_point.h"
#include "utilities/fault_injection_env.h"
#include "utilities/fault_injection_fs.h"
namespace ROCKSDB_NAMESPACE {
class DBWALTestBase : public DBTestBase {
protected:
explicit DBWALTestBase(const std::string& dir_name)
: DBTestBase(dir_name, /*env_do_fsync=*/true) {}
#if defined(ROCKSDB_PLATFORM_POSIX)
public:
#if defined(ROCKSDB_FALLOCATE_PRESENT)
bool IsFallocateSupported() {
// Test fallocate support of running file system.
// Skip this test if fallocate is not supported.
std::string fname_test_fallocate = dbname_ + "/preallocate_testfile";
int fd = -1;
do {
fd = open(fname_test_fallocate.c_str(), O_CREAT | O_RDWR | O_TRUNC, 0644);
} while (fd < 0 && errno == EINTR);
assert(fd > 0);
int alloc_status = fallocate(fd, 0, 0, 1);
int err_number = errno;
close(fd);
assert(env_->DeleteFile(fname_test_fallocate) == Status::OK());
if (err_number == ENOSYS || err_number == EOPNOTSUPP) {
fprintf(stderr, "Skipped preallocated space check: %s\n",
errnoStr(err_number).c_str());
return false;
}
assert(alloc_status == 0);
return true;
}
#endif // ROCKSDB_FALLOCATE_PRESENT
uint64_t GetAllocatedFileSize(std::string file_name) {
struct stat sbuf;
int err = stat(file_name.c_str(), &sbuf);
assert(err == 0);
return sbuf.st_blocks * 512;
}
#endif // ROCKSDB_PLATFORM_POSIX
};
class DBWALTest : public DBWALTestBase {
public:
DBWALTest() : DBWALTestBase("/db_wal_test") {}
};
Skip deleted WALs during recovery Summary: This patch record min log number to keep to the manifest while flushing SST files to ignore them and any WAL older than them during recovery. This is to avoid scenarios when we have a gap between the WAL files are fed to the recovery procedure. The gap could happen by for example out-of-order WAL deletion. Such gap could cause problems in 2PC recovery where the prepared and commit entry are placed into two separate WAL and gap in the WALs could result into not processing the WAL with the commit entry and hence breaking the 2PC recovery logic. Before the commit, for 2PC case, we determined which log number to keep in FindObsoleteFiles(). We looked at the earliest logs with outstanding prepare entries, or prepare entries whose respective commit or abort are in memtable. With the commit, the same calculation is done while we apply the SST flush. Just before installing the flush file, we precompute the earliest log file to keep after the flush finishes using the same logic (but skipping the memtables just flushed), record this information to the manifest entry for this new flushed SST file. This pre-computed value is also remembered in memory, and will later be used to determine whether a log file can be deleted. This value is unlikely to change until next flush because the commit entry will stay in memtable. (In WritePrepared, we could have removed the older log files as soon as all prepared entries are committed. It's not yet done anyway. Even if we do it, the only thing we loss with this new approach is earlier log deletion between two flushes, which does not guarantee to happen anyway because the obsolete file clean-up function is only executed after flush or compaction) This min log number to keep is stored in the manifest using the safely-ignore customized field of AddFile entry, in order to guarantee that the DB generated using newer release can be opened by previous releases no older than 4.2. Closes https://github.com/facebook/rocksdb/pull/3765 Differential Revision: D7747618 Pulled By: siying fbshipit-source-id: d00c92105b4f83852e9754a1b70d6b64cb590729
7 years ago
// A SpecialEnv enriched to give more insight about deleted files
class EnrichedSpecialEnv : public SpecialEnv {
public:
explicit EnrichedSpecialEnv(Env* base) : SpecialEnv(base) {}
Status NewSequentialFile(const std::string& f,
std::unique_ptr<SequentialFile>* r,
Skip deleted WALs during recovery Summary: This patch record min log number to keep to the manifest while flushing SST files to ignore them and any WAL older than them during recovery. This is to avoid scenarios when we have a gap between the WAL files are fed to the recovery procedure. The gap could happen by for example out-of-order WAL deletion. Such gap could cause problems in 2PC recovery where the prepared and commit entry are placed into two separate WAL and gap in the WALs could result into not processing the WAL with the commit entry and hence breaking the 2PC recovery logic. Before the commit, for 2PC case, we determined which log number to keep in FindObsoleteFiles(). We looked at the earliest logs with outstanding prepare entries, or prepare entries whose respective commit or abort are in memtable. With the commit, the same calculation is done while we apply the SST flush. Just before installing the flush file, we precompute the earliest log file to keep after the flush finishes using the same logic (but skipping the memtables just flushed), record this information to the manifest entry for this new flushed SST file. This pre-computed value is also remembered in memory, and will later be used to determine whether a log file can be deleted. This value is unlikely to change until next flush because the commit entry will stay in memtable. (In WritePrepared, we could have removed the older log files as soon as all prepared entries are committed. It's not yet done anyway. Even if we do it, the only thing we loss with this new approach is earlier log deletion between two flushes, which does not guarantee to happen anyway because the obsolete file clean-up function is only executed after flush or compaction) This min log number to keep is stored in the manifest using the safely-ignore customized field of AddFile entry, in order to guarantee that the DB generated using newer release can be opened by previous releases no older than 4.2. Closes https://github.com/facebook/rocksdb/pull/3765 Differential Revision: D7747618 Pulled By: siying fbshipit-source-id: d00c92105b4f83852e9754a1b70d6b64cb590729
7 years ago
const EnvOptions& soptions) override {
InstrumentedMutexLock l(&env_mutex_);
if (f == skipped_wal) {
deleted_wal_reopened = true;
if (IsWAL(f) && largest_deleted_wal.size() != 0 &&
f.compare(largest_deleted_wal) <= 0) {
Skip deleted WALs during recovery Summary: This patch record min log number to keep to the manifest while flushing SST files to ignore them and any WAL older than them during recovery. This is to avoid scenarios when we have a gap between the WAL files are fed to the recovery procedure. The gap could happen by for example out-of-order WAL deletion. Such gap could cause problems in 2PC recovery where the prepared and commit entry are placed into two separate WAL and gap in the WALs could result into not processing the WAL with the commit entry and hence breaking the 2PC recovery logic. Before the commit, for 2PC case, we determined which log number to keep in FindObsoleteFiles(). We looked at the earliest logs with outstanding prepare entries, or prepare entries whose respective commit or abort are in memtable. With the commit, the same calculation is done while we apply the SST flush. Just before installing the flush file, we precompute the earliest log file to keep after the flush finishes using the same logic (but skipping the memtables just flushed), record this information to the manifest entry for this new flushed SST file. This pre-computed value is also remembered in memory, and will later be used to determine whether a log file can be deleted. This value is unlikely to change until next flush because the commit entry will stay in memtable. (In WritePrepared, we could have removed the older log files as soon as all prepared entries are committed. It's not yet done anyway. Even if we do it, the only thing we loss with this new approach is earlier log deletion between two flushes, which does not guarantee to happen anyway because the obsolete file clean-up function is only executed after flush or compaction) This min log number to keep is stored in the manifest using the safely-ignore customized field of AddFile entry, in order to guarantee that the DB generated using newer release can be opened by previous releases no older than 4.2. Closes https://github.com/facebook/rocksdb/pull/3765 Differential Revision: D7747618 Pulled By: siying fbshipit-source-id: d00c92105b4f83852e9754a1b70d6b64cb590729
7 years ago
gap_in_wals = true;
}
}
return SpecialEnv::NewSequentialFile(f, r, soptions);
}
Status DeleteFile(const std::string& fname) override {
if (IsWAL(fname)) {
deleted_wal_cnt++;
InstrumentedMutexLock l(&env_mutex_);
// If this is the first WAL, remember its name and skip deleting it. We
// remember its name partly because the application might attempt to
// delete the file again.
if (skipped_wal.size() != 0 && skipped_wal != fname) {
if (largest_deleted_wal.size() == 0 ||
largest_deleted_wal.compare(fname) < 0) {
largest_deleted_wal = fname;
Skip deleted WALs during recovery Summary: This patch record min log number to keep to the manifest while flushing SST files to ignore them and any WAL older than them during recovery. This is to avoid scenarios when we have a gap between the WAL files are fed to the recovery procedure. The gap could happen by for example out-of-order WAL deletion. Such gap could cause problems in 2PC recovery where the prepared and commit entry are placed into two separate WAL and gap in the WALs could result into not processing the WAL with the commit entry and hence breaking the 2PC recovery logic. Before the commit, for 2PC case, we determined which log number to keep in FindObsoleteFiles(). We looked at the earliest logs with outstanding prepare entries, or prepare entries whose respective commit or abort are in memtable. With the commit, the same calculation is done while we apply the SST flush. Just before installing the flush file, we precompute the earliest log file to keep after the flush finishes using the same logic (but skipping the memtables just flushed), record this information to the manifest entry for this new flushed SST file. This pre-computed value is also remembered in memory, and will later be used to determine whether a log file can be deleted. This value is unlikely to change until next flush because the commit entry will stay in memtable. (In WritePrepared, we could have removed the older log files as soon as all prepared entries are committed. It's not yet done anyway. Even if we do it, the only thing we loss with this new approach is earlier log deletion between two flushes, which does not guarantee to happen anyway because the obsolete file clean-up function is only executed after flush or compaction) This min log number to keep is stored in the manifest using the safely-ignore customized field of AddFile entry, in order to guarantee that the DB generated using newer release can be opened by previous releases no older than 4.2. Closes https://github.com/facebook/rocksdb/pull/3765 Differential Revision: D7747618 Pulled By: siying fbshipit-source-id: d00c92105b4f83852e9754a1b70d6b64cb590729
7 years ago
}
} else {
skipped_wal = fname;
return Status::OK();
}
}
return SpecialEnv::DeleteFile(fname);
}
bool IsWAL(const std::string& fname) {
// printf("iswal %s\n", fname.c_str());
return fname.compare(fname.size() - 3, 3, "log") == 0;
}
InstrumentedMutex env_mutex_;
// the wal whose actual delete was skipped by the env
std::string skipped_wal = "";
// the largest WAL that was requested to be deleted
std::string largest_deleted_wal = "";
Skip deleted WALs during recovery Summary: This patch record min log number to keep to the manifest while flushing SST files to ignore them and any WAL older than them during recovery. This is to avoid scenarios when we have a gap between the WAL files are fed to the recovery procedure. The gap could happen by for example out-of-order WAL deletion. Such gap could cause problems in 2PC recovery where the prepared and commit entry are placed into two separate WAL and gap in the WALs could result into not processing the WAL with the commit entry and hence breaking the 2PC recovery logic. Before the commit, for 2PC case, we determined which log number to keep in FindObsoleteFiles(). We looked at the earliest logs with outstanding prepare entries, or prepare entries whose respective commit or abort are in memtable. With the commit, the same calculation is done while we apply the SST flush. Just before installing the flush file, we precompute the earliest log file to keep after the flush finishes using the same logic (but skipping the memtables just flushed), record this information to the manifest entry for this new flushed SST file. This pre-computed value is also remembered in memory, and will later be used to determine whether a log file can be deleted. This value is unlikely to change until next flush because the commit entry will stay in memtable. (In WritePrepared, we could have removed the older log files as soon as all prepared entries are committed. It's not yet done anyway. Even if we do it, the only thing we loss with this new approach is earlier log deletion between two flushes, which does not guarantee to happen anyway because the obsolete file clean-up function is only executed after flush or compaction) This min log number to keep is stored in the manifest using the safely-ignore customized field of AddFile entry, in order to guarantee that the DB generated using newer release can be opened by previous releases no older than 4.2. Closes https://github.com/facebook/rocksdb/pull/3765 Differential Revision: D7747618 Pulled By: siying fbshipit-source-id: d00c92105b4f83852e9754a1b70d6b64cb590729
7 years ago
// number of WALs that were successfully deleted
std::atomic<size_t> deleted_wal_cnt = {0};
// the WAL whose delete from fs was skipped is reopened during recovery
std::atomic<bool> deleted_wal_reopened = {false};
// whether a gap in the WALs was detected during recovery
std::atomic<bool> gap_in_wals = {false};
};
class DBWALTestWithEnrichedEnv : public DBTestBase {
public:
DBWALTestWithEnrichedEnv()
: DBTestBase("db_wal_test", /*env_do_fsync=*/true) {
Skip deleted WALs during recovery Summary: This patch record min log number to keep to the manifest while flushing SST files to ignore them and any WAL older than them during recovery. This is to avoid scenarios when we have a gap between the WAL files are fed to the recovery procedure. The gap could happen by for example out-of-order WAL deletion. Such gap could cause problems in 2PC recovery where the prepared and commit entry are placed into two separate WAL and gap in the WALs could result into not processing the WAL with the commit entry and hence breaking the 2PC recovery logic. Before the commit, for 2PC case, we determined which log number to keep in FindObsoleteFiles(). We looked at the earliest logs with outstanding prepare entries, or prepare entries whose respective commit or abort are in memtable. With the commit, the same calculation is done while we apply the SST flush. Just before installing the flush file, we precompute the earliest log file to keep after the flush finishes using the same logic (but skipping the memtables just flushed), record this information to the manifest entry for this new flushed SST file. This pre-computed value is also remembered in memory, and will later be used to determine whether a log file can be deleted. This value is unlikely to change until next flush because the commit entry will stay in memtable. (In WritePrepared, we could have removed the older log files as soon as all prepared entries are committed. It's not yet done anyway. Even if we do it, the only thing we loss with this new approach is earlier log deletion between two flushes, which does not guarantee to happen anyway because the obsolete file clean-up function is only executed after flush or compaction) This min log number to keep is stored in the manifest using the safely-ignore customized field of AddFile entry, in order to guarantee that the DB generated using newer release can be opened by previous releases no older than 4.2. Closes https://github.com/facebook/rocksdb/pull/3765 Differential Revision: D7747618 Pulled By: siying fbshipit-source-id: d00c92105b4f83852e9754a1b70d6b64cb590729
7 years ago
enriched_env_ = new EnrichedSpecialEnv(env_->target());
auto options = CurrentOptions();
options.env = enriched_env_;
options.allow_2pc = true;
Skip deleted WALs during recovery Summary: This patch record min log number to keep to the manifest while flushing SST files to ignore them and any WAL older than them during recovery. This is to avoid scenarios when we have a gap between the WAL files are fed to the recovery procedure. The gap could happen by for example out-of-order WAL deletion. Such gap could cause problems in 2PC recovery where the prepared and commit entry are placed into two separate WAL and gap in the WALs could result into not processing the WAL with the commit entry and hence breaking the 2PC recovery logic. Before the commit, for 2PC case, we determined which log number to keep in FindObsoleteFiles(). We looked at the earliest logs with outstanding prepare entries, or prepare entries whose respective commit or abort are in memtable. With the commit, the same calculation is done while we apply the SST flush. Just before installing the flush file, we precompute the earliest log file to keep after the flush finishes using the same logic (but skipping the memtables just flushed), record this information to the manifest entry for this new flushed SST file. This pre-computed value is also remembered in memory, and will later be used to determine whether a log file can be deleted. This value is unlikely to change until next flush because the commit entry will stay in memtable. (In WritePrepared, we could have removed the older log files as soon as all prepared entries are committed. It's not yet done anyway. Even if we do it, the only thing we loss with this new approach is earlier log deletion between two flushes, which does not guarantee to happen anyway because the obsolete file clean-up function is only executed after flush or compaction) This min log number to keep is stored in the manifest using the safely-ignore customized field of AddFile entry, in order to guarantee that the DB generated using newer release can be opened by previous releases no older than 4.2. Closes https://github.com/facebook/rocksdb/pull/3765 Differential Revision: D7747618 Pulled By: siying fbshipit-source-id: d00c92105b4f83852e9754a1b70d6b64cb590729
7 years ago
Reopen(options);
delete env_;
// to be deleted by the parent class
env_ = enriched_env_;
}
protected:
EnrichedSpecialEnv* enriched_env_;
};
// Test that the recovery would successfully avoid the gaps between the logs.
// One known scenario that could cause this is that the application issue the
// WAL deletion out of order. For the sake of simplicity in the test, here we
// create the gap by manipulating the env to skip deletion of the first WAL but
// not the ones after it.
TEST_F(DBWALTestWithEnrichedEnv, SkipDeletedWALs) {
auto options = last_options_;
// To cause frequent WAL deletion
options.write_buffer_size = 128;
Reopen(options);
WriteOptions writeOpt = WriteOptions();
for (int i = 0; i < 128 * 5; i++) {
ASSERT_OK(dbfull()->Put(writeOpt, "foo", "v1"));
}
FlushOptions fo;
fo.wait = true;
ASSERT_OK(db_->Flush(fo));
// some wals are deleted
ASSERT_NE(0, enriched_env_->deleted_wal_cnt);
// but not the first one
ASSERT_NE(0, enriched_env_->skipped_wal.size());
// Test that the WAL that was not deleted will be skipped during recovery
options = last_options_;
Reopen(options);
ASSERT_FALSE(enriched_env_->deleted_wal_reopened);
ASSERT_FALSE(enriched_env_->gap_in_wals);
}
TEST_F(DBWALTest, WAL) {
do {
CreateAndReopenWithCF({"pikachu"}, CurrentOptions());
WriteOptions writeOpt = WriteOptions();
writeOpt.disableWAL = true;
ASSERT_OK(dbfull()->Put(writeOpt, handles_[1], "foo", "v1"));
ASSERT_OK(dbfull()->Put(writeOpt, handles_[1], "bar", "v1"));
ReopenWithColumnFamilies({"default", "pikachu"}, CurrentOptions());
ASSERT_EQ("v1", Get(1, "foo"));
ASSERT_EQ("v1", Get(1, "bar"));
writeOpt.disableWAL = false;
ASSERT_OK(dbfull()->Put(writeOpt, handles_[1], "bar", "v2"));
writeOpt.disableWAL = true;
ASSERT_OK(dbfull()->Put(writeOpt, handles_[1], "foo", "v2"));
ReopenWithColumnFamilies({"default", "pikachu"}, CurrentOptions());
// Both value's should be present.
ASSERT_EQ("v2", Get(1, "bar"));
ASSERT_EQ("v2", Get(1, "foo"));
writeOpt.disableWAL = true;
ASSERT_OK(dbfull()->Put(writeOpt, handles_[1], "bar", "v3"));
writeOpt.disableWAL = false;
ASSERT_OK(dbfull()->Put(writeOpt, handles_[1], "foo", "v3"));
ReopenWithColumnFamilies({"default", "pikachu"}, CurrentOptions());
// again both values should be present.
ASSERT_EQ("v3", Get(1, "foo"));
ASSERT_EQ("v3", Get(1, "bar"));
} while (ChangeWalOptions());
}
TEST_F(DBWALTest, RollLog) {
do {
CreateAndReopenWithCF({"pikachu"}, CurrentOptions());
ASSERT_OK(Put(1, "foo", "v1"));
ASSERT_OK(Put(1, "baz", "v5"));
ReopenWithColumnFamilies({"default", "pikachu"}, CurrentOptions());
for (int i = 0; i < 10; i++) {
ReopenWithColumnFamilies({"default", "pikachu"}, CurrentOptions());
}
ASSERT_OK(Put(1, "foo", "v4"));
for (int i = 0; i < 10; i++) {
ReopenWithColumnFamilies({"default", "pikachu"}, CurrentOptions());
}
} while (ChangeWalOptions());
}
TEST_F(DBWALTest, SyncWALNotBlockWrite) {
Options options = CurrentOptions();
options.max_write_buffer_number = 4;
DestroyAndReopen(options);
ASSERT_OK(Put("foo1", "bar1"));
ASSERT_OK(Put("foo5", "bar5"));
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->LoadDependency({
{"WritableFileWriter::SyncWithoutFlush:1",
"DBWALTest::SyncWALNotBlockWrite:1"},
{"DBWALTest::SyncWALNotBlockWrite:2",
"WritableFileWriter::SyncWithoutFlush:2"},
});
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->EnableProcessing();
ROCKSDB_NAMESPACE::port::Thread thread([&]() { ASSERT_OK(db_->SyncWAL()); });
TEST_SYNC_POINT("DBWALTest::SyncWALNotBlockWrite:1");
ASSERT_OK(Put("foo2", "bar2"));
ASSERT_OK(Put("foo3", "bar3"));
FlushOptions fo;
fo.wait = false;
ASSERT_OK(db_->Flush(fo));
ASSERT_OK(Put("foo4", "bar4"));
TEST_SYNC_POINT("DBWALTest::SyncWALNotBlockWrite:2");
thread.join();
ASSERT_EQ(Get("foo1"), "bar1");
ASSERT_EQ(Get("foo2"), "bar2");
ASSERT_EQ(Get("foo3"), "bar3");
ASSERT_EQ(Get("foo4"), "bar4");
ASSERT_EQ(Get("foo5"), "bar5");
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->DisableProcessing();
}
TEST_F(DBWALTest, SyncWALNotWaitWrite) {
ASSERT_OK(Put("foo1", "bar1"));
ASSERT_OK(Put("foo3", "bar3"));
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->LoadDependency({
{"SpecialEnv::WalFile::Append:1", "DBWALTest::SyncWALNotWaitWrite:1"},
{"DBWALTest::SyncWALNotWaitWrite:2", "SpecialEnv::WalFile::Append:2"},
});
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->EnableProcessing();
ROCKSDB_NAMESPACE::port::Thread thread(
[&]() { ASSERT_OK(Put("foo2", "bar2")); });
Optimize for serial commits in 2PC Summary: Throughput: 46k tps in our sysbench settings (filling the details later) The idea is to have the simplest change that gives us a reasonable boost in 2PC throughput. Major design changes: 1. The WAL file internal buffer is not flushed after each write. Instead it is flushed before critical operations (WAL copy via fs) or when FlushWAL is called by MySQL. Flushing the WAL buffer is also protected via mutex_. 2. Use two sequence numbers: last seq, and last seq for write. Last seq is the last visible sequence number for reads. Last seq for write is the next sequence number that should be used to write to WAL/memtable. This allows to have a memtable write be in parallel to WAL writes. 3. BatchGroup is not used for writes. This means that we can have parallel writers which changes a major assumption in the code base. To accommodate for that i) allow only 1 WriteImpl that intends to write to memtable via mem_mutex_--which is fine since in 2PC almost all of the memtable writes come via group commit phase which is serial anyway, ii) make all the parts in the code base that assumed to be the only writer (via EnterUnbatched) to also acquire mem_mutex_, iii) stat updates are protected via a stat_mutex_. Note: the first commit has the approach figured out but is not clean. Submitting the PR anyway to get the early feedback on the approach. If we are ok with the approach I will go ahead with this updates: 0) Rebase with Yi's pipelining changes 1) Currently batching is disabled by default to make sure that it will be consistent with all unit tests. Will make this optional via a config. 2) A couple of unit tests are disabled. They need to be updated with the serial commit of 2PC taken into account. 3) Replacing BatchGroup with mem_mutex_ got a bit ugly as it requires releasing mutex_ beforehand (the same way EnterUnbatched does). This needs to be cleaned up. Closes https://github.com/facebook/rocksdb/pull/2345 Differential Revision: D5210732 Pulled By: maysamyabandeh fbshipit-source-id: 78653bd95a35cd1e831e555e0e57bdfd695355a4
8 years ago
// Moving this to SyncWAL before the actual fsync
// TEST_SYNC_POINT("DBWALTest::SyncWALNotWaitWrite:1");
ASSERT_OK(db_->SyncWAL());
Optimize for serial commits in 2PC Summary: Throughput: 46k tps in our sysbench settings (filling the details later) The idea is to have the simplest change that gives us a reasonable boost in 2PC throughput. Major design changes: 1. The WAL file internal buffer is not flushed after each write. Instead it is flushed before critical operations (WAL copy via fs) or when FlushWAL is called by MySQL. Flushing the WAL buffer is also protected via mutex_. 2. Use two sequence numbers: last seq, and last seq for write. Last seq is the last visible sequence number for reads. Last seq for write is the next sequence number that should be used to write to WAL/memtable. This allows to have a memtable write be in parallel to WAL writes. 3. BatchGroup is not used for writes. This means that we can have parallel writers which changes a major assumption in the code base. To accommodate for that i) allow only 1 WriteImpl that intends to write to memtable via mem_mutex_--which is fine since in 2PC almost all of the memtable writes come via group commit phase which is serial anyway, ii) make all the parts in the code base that assumed to be the only writer (via EnterUnbatched) to also acquire mem_mutex_, iii) stat updates are protected via a stat_mutex_. Note: the first commit has the approach figured out but is not clean. Submitting the PR anyway to get the early feedback on the approach. If we are ok with the approach I will go ahead with this updates: 0) Rebase with Yi's pipelining changes 1) Currently batching is disabled by default to make sure that it will be consistent with all unit tests. Will make this optional via a config. 2) A couple of unit tests are disabled. They need to be updated with the serial commit of 2PC taken into account. 3) Replacing BatchGroup with mem_mutex_ got a bit ugly as it requires releasing mutex_ beforehand (the same way EnterUnbatched does). This needs to be cleaned up. Closes https://github.com/facebook/rocksdb/pull/2345 Differential Revision: D5210732 Pulled By: maysamyabandeh fbshipit-source-id: 78653bd95a35cd1e831e555e0e57bdfd695355a4
8 years ago
// Moving this to SyncWAL after actual fsync
// TEST_SYNC_POINT("DBWALTest::SyncWALNotWaitWrite:2");
thread.join();
ASSERT_EQ(Get("foo1"), "bar1");
ASSERT_EQ(Get("foo2"), "bar2");
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->DisableProcessing();
}
TEST_F(DBWALTest, Recover) {
do {
CreateAndReopenWithCF({"pikachu"}, CurrentOptions());
ASSERT_OK(Put(1, "foo", "v1"));
ASSERT_OK(Put(1, "baz", "v5"));
ReopenWithColumnFamilies({"default", "pikachu"}, CurrentOptions());
ASSERT_EQ("v1", Get(1, "foo"));
ASSERT_EQ("v1", Get(1, "foo"));
ASSERT_EQ("v5", Get(1, "baz"));
ASSERT_OK(Put(1, "bar", "v2"));
ASSERT_OK(Put(1, "foo", "v3"));
ReopenWithColumnFamilies({"default", "pikachu"}, CurrentOptions());
ASSERT_EQ("v3", Get(1, "foo"));
ASSERT_OK(Put(1, "foo", "v4"));
ASSERT_EQ("v4", Get(1, "foo"));
ASSERT_EQ("v2", Get(1, "bar"));
ASSERT_EQ("v5", Get(1, "baz"));
} while (ChangeWalOptions());
}
TEST_F(DBWALTest, RecoverWithTableHandle) {
do {
Options options = CurrentOptions();
options.create_if_missing = true;
options.disable_auto_compactions = true;
options.avoid_flush_during_recovery = false;
DestroyAndReopen(options);
CreateAndReopenWithCF({"pikachu"}, options);
ASSERT_OK(Put(1, "foo", "v1"));
ASSERT_OK(Put(1, "bar", "v2"));
ASSERT_OK(Flush(1));
ASSERT_OK(Put(1, "foo", "v3"));
ASSERT_OK(Put(1, "bar", "v4"));
ASSERT_OK(Flush(1));
ASSERT_OK(Put(1, "big", std::string(100, 'a')));
options = CurrentOptions();
const int kSmallMaxOpenFiles = 13;
if (option_config_ == kDBLogDir) {
// Use this option to check not preloading files
// Set the max open files to be small enough so no preload will
// happen.
options.max_open_files = kSmallMaxOpenFiles;
// RocksDB sanitize max open files to at least 20. Modify it back.
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->SetCallBack(
"SanitizeOptions::AfterChangeMaxOpenFiles", [&](void* arg) {
int* max_open_files = static_cast<int*>(arg);
*max_open_files = kSmallMaxOpenFiles;
});
} else if (option_config_ == kWalDirAndMmapReads) {
// Use this option to check always loading all files.
options.max_open_files = 100;
} else {
options.max_open_files = -1;
}
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->EnableProcessing();
ReopenWithColumnFamilies({"default", "pikachu"}, options);
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->DisableProcessing();
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->ClearAllCallBacks();
std::vector<std::vector<FileMetaData>> files;
dbfull()->TEST_GetFilesMetaData(handles_[1], &files);
size_t total_files = 0;
for (const auto& level : files) {
total_files += level.size();
}
ASSERT_EQ(total_files, 3);
for (const auto& level : files) {
for (const auto& file : level) {
if (options.max_open_files == kSmallMaxOpenFiles) {
ASSERT_TRUE(file.table_reader_handle == nullptr);
} else {
ASSERT_TRUE(file.table_reader_handle != nullptr);
}
}
}
} while (ChangeWalOptions());
}
TEST_F(DBWALTest, RecoverWithBlob) {
// Write a value that's below the prospective size limit for blobs and another
// one that's above. Note that blob files are not actually enabled at this
// point.
constexpr uint64_t min_blob_size = 10;
constexpr char short_value[] = "short";
static_assert(sizeof(short_value) - 1 < min_blob_size,
"short_value too long");
constexpr char long_value[] = "long_value";
static_assert(sizeof(long_value) - 1 >= min_blob_size,
"long_value too short");
ASSERT_OK(Put("key1", short_value));
ASSERT_OK(Put("key2", long_value));
// There should be no files just yet since we haven't flushed.
{
VersionSet* const versions = dbfull()->GetVersionSet();
ASSERT_NE(versions, nullptr);
ColumnFamilyData* const cfd = versions->GetColumnFamilySet()->GetDefault();
ASSERT_NE(cfd, nullptr);
Version* const current = cfd->current();
ASSERT_NE(current, nullptr);
const VersionStorageInfo* const storage_info = current->storage_info();
ASSERT_NE(storage_info, nullptr);
ASSERT_EQ(storage_info->num_non_empty_levels(), 0);
ASSERT_TRUE(storage_info->GetBlobFiles().empty());
}
// Reopen the database with blob files enabled. A new table file/blob file
// pair should be written during recovery.
Options options;
options.enable_blob_files = true;
options.min_blob_size = min_blob_size;
options.avoid_flush_during_recovery = false;
options.disable_auto_compactions = true;
options.env = env_;
Reopen(options);
ASSERT_EQ(Get("key1"), short_value);
ASSERT_EQ(Get("key2"), long_value);
VersionSet* const versions = dbfull()->GetVersionSet();
ASSERT_NE(versions, nullptr);
ColumnFamilyData* const cfd = versions->GetColumnFamilySet()->GetDefault();
ASSERT_NE(cfd, nullptr);
Version* const current = cfd->current();
ASSERT_NE(current, nullptr);
const VersionStorageInfo* const storage_info = current->storage_info();
ASSERT_NE(storage_info, nullptr);
const auto& l0_files = storage_info->LevelFiles(0);
ASSERT_EQ(l0_files.size(), 1);
const FileMetaData* const table_file = l0_files[0];
ASSERT_NE(table_file, nullptr);
const auto& blob_files = storage_info->GetBlobFiles();
ASSERT_EQ(blob_files.size(), 1);
Use a sorted vector instead of a map to store blob file metadata (#9526) Summary: The patch replaces `std::map` with a sorted `std::vector` for `VersionStorageInfo::blob_files_` and preallocates the space for the `vector` before saving the `BlobFileMetaData` into the new `VersionStorageInfo` in `VersionBuilder::Rep::SaveBlobFilesTo`. These changes reduce the time the DB mutex is held while saving new `Version`s, and using a sorted `vector` also makes lookups faster thanks to better memory locality. In addition, the patch introduces helper methods `VersionStorageInfo::GetBlobFileMetaData` and `VersionStorageInfo::GetBlobFileMetaDataLB` that can be used by clients to perform lookups in the `vector`, and does some general cleanup in the parts of code where blob file metadata are used. Pull Request resolved: https://github.com/facebook/rocksdb/pull/9526 Test Plan: Ran `make check` and the crash test script for a while. Performance was tested using a load-optimized benchmark (`fillseq` with vector memtable, no WAL) and small file sizes so that a significant number of files are produced: ``` numactl --interleave=all ./db_bench --benchmarks=fillseq --allow_concurrent_memtable_write=false --level0_file_num_compaction_trigger=4 --level0_slowdown_writes_trigger=20 --level0_stop_writes_trigger=30 --max_background_jobs=8 --max_write_buffer_number=8 --db=/data/ltamasi-dbbench --wal_dir=/data/ltamasi-dbbench --num=800000000 --num_levels=8 --key_size=20 --value_size=400 --block_size=8192 --cache_size=51539607552 --cache_numshardbits=6 --compression_max_dict_bytes=0 --compression_ratio=0.5 --compression_type=lz4 --bytes_per_sync=8388608 --cache_index_and_filter_blocks=1 --cache_high_pri_pool_ratio=0.5 --benchmark_write_rate_limit=0 --write_buffer_size=16777216 --target_file_size_base=16777216 --max_bytes_for_level_base=67108864 --verify_checksum=1 --delete_obsolete_files_period_micros=62914560 --max_bytes_for_level_multiplier=8 --statistics=0 --stats_per_interval=1 --stats_interval_seconds=20 --histogram=1 --memtablerep=skip_list --bloom_bits=10 --open_files=-1 --subcompactions=1 --compaction_style=0 --min_level_to_compress=3 --level_compaction_dynamic_level_bytes=true --pin_l0_filter_and_index_blocks_in_cache=1 --soft_pending_compaction_bytes_limit=167503724544 --hard_pending_compaction_bytes_limit=335007449088 --min_level_to_compress=0 --use_existing_db=0 --sync=0 --threads=1 --memtablerep=vector --allow_concurrent_memtable_write=false --disable_wal=1 --enable_blob_files=1 --blob_file_size=16777216 --min_blob_size=0 --blob_compression_type=lz4 --enable_blob_garbage_collection=1 --seed=<some value> ``` Final statistics before the patch: ``` Cumulative writes: 0 writes, 700M keys, 0 commit groups, 0.0 writes per commit group, ingest: 284.62 GB, 121.27 MB/s Interval writes: 0 writes, 334K keys, 0 commit groups, 0.0 writes per commit group, ingest: 139.28 MB, 72.46 MB/s ``` With the patch: ``` Cumulative writes: 0 writes, 760M keys, 0 commit groups, 0.0 writes per commit group, ingest: 308.66 GB, 131.52 MB/s Interval writes: 0 writes, 445K keys, 0 commit groups, 0.0 writes per commit group, ingest: 185.35 MB, 93.15 MB/s ``` Total time to complete the benchmark is 2611 seconds with the patch, down from 2986 secs. Reviewed By: riversand963 Differential Revision: D34082728 Pulled By: ltamasi fbshipit-source-id: fc598abf676dce436734d06bb9d2d99a26a004fc
3 years ago
const auto& blob_file = blob_files.front();
ASSERT_NE(blob_file, nullptr);
ASSERT_EQ(table_file->smallest.user_key(), "key1");
ASSERT_EQ(table_file->largest.user_key(), "key2");
ASSERT_EQ(table_file->fd.smallest_seqno, 1);
ASSERT_EQ(table_file->fd.largest_seqno, 2);
ASSERT_EQ(table_file->oldest_blob_file_number,
blob_file->GetBlobFileNumber());
ASSERT_EQ(blob_file->GetTotalBlobCount(), 1);
const InternalStats* const internal_stats = cfd->internal_stats();
ASSERT_NE(internal_stats, nullptr);
const auto& compaction_stats = internal_stats->TEST_GetCompactionStats();
ASSERT_FALSE(compaction_stats.empty());
ASSERT_EQ(compaction_stats[0].bytes_written, table_file->fd.GetFileSize());
ASSERT_EQ(compaction_stats[0].bytes_written_blob,
blob_file->GetTotalBlobBytes());
ASSERT_EQ(compaction_stats[0].num_output_files, 1);
ASSERT_EQ(compaction_stats[0].num_output_files_blob, 1);
const uint64_t* const cf_stats_value = internal_stats->TEST_GetCFStatsValue();
ASSERT_EQ(cf_stats_value[InternalStats::BYTES_FLUSHED],
compaction_stats[0].bytes_written +
compaction_stats[0].bytes_written_blob);
}
TEST_F(DBWALTest, RecoverWithBlobMultiSST) {
// Write several large (4 KB) values without flushing. Note that blob files
// are not actually enabled at this point.
std::string large_value(1 << 12, 'a');
constexpr int num_keys = 64;
for (int i = 0; i < num_keys; ++i) {
ASSERT_OK(Put(Key(i), large_value));
}
// There should be no files just yet since we haven't flushed.
{
VersionSet* const versions = dbfull()->GetVersionSet();
ASSERT_NE(versions, nullptr);
ColumnFamilyData* const cfd = versions->GetColumnFamilySet()->GetDefault();
ASSERT_NE(cfd, nullptr);
Version* const current = cfd->current();
ASSERT_NE(current, nullptr);
const VersionStorageInfo* const storage_info = current->storage_info();
ASSERT_NE(storage_info, nullptr);
ASSERT_EQ(storage_info->num_non_empty_levels(), 0);
ASSERT_TRUE(storage_info->GetBlobFiles().empty());
}
// Reopen the database with blob files enabled and write buffer size set to a
// smaller value. Multiple table files+blob files should be written and added
// to the Version during recovery.
Options options;
options.write_buffer_size = 1 << 16; // 64 KB
options.enable_blob_files = true;
options.avoid_flush_during_recovery = false;
options.disable_auto_compactions = true;
options.env = env_;
Reopen(options);
for (int i = 0; i < num_keys; ++i) {
ASSERT_EQ(Get(Key(i)), large_value);
}
VersionSet* const versions = dbfull()->GetVersionSet();
ASSERT_NE(versions, nullptr);
ColumnFamilyData* const cfd = versions->GetColumnFamilySet()->GetDefault();
ASSERT_NE(cfd, nullptr);
Version* const current = cfd->current();
ASSERT_NE(current, nullptr);
const VersionStorageInfo* const storage_info = current->storage_info();
ASSERT_NE(storage_info, nullptr);
const auto& l0_files = storage_info->LevelFiles(0);
ASSERT_GT(l0_files.size(), 1);
const auto& blob_files = storage_info->GetBlobFiles();
ASSERT_GT(blob_files.size(), 1);
ASSERT_EQ(l0_files.size(), blob_files.size());
}
TEST_F(DBWALTest, WALWithChecksumHandoff) {
#ifndef ROCKSDB_ASSERT_STATUS_CHECKED
if (mem_env_ || encrypted_env_) {
ROCKSDB_GTEST_SKIP("Test requires non-mem or non-encrypted environment");
return;
}
std::shared_ptr<FaultInjectionTestFS> fault_fs(
new FaultInjectionTestFS(FileSystem::Default()));
std::unique_ptr<Env> fault_fs_env(NewCompositeEnv(fault_fs));
do {
Options options = CurrentOptions();
options.checksum_handoff_file_types.Add(FileType::kWalFile);
options.env = fault_fs_env.get();
fault_fs->SetChecksumHandoffFuncType(ChecksumType::kCRC32c);
CreateAndReopenWithCF({"pikachu"}, options);
WriteOptions writeOpt = WriteOptions();
writeOpt.disableWAL = true;
ASSERT_OK(dbfull()->Put(writeOpt, handles_[1], "foo", "v1"));
ASSERT_OK(dbfull()->Put(writeOpt, handles_[1], "bar", "v1"));
ReopenWithColumnFamilies({"default", "pikachu"}, options);
ASSERT_EQ("v1", Get(1, "foo"));
ASSERT_EQ("v1", Get(1, "bar"));
writeOpt.disableWAL = false;
ASSERT_OK(dbfull()->Put(writeOpt, handles_[1], "bar", "v2"));
writeOpt.disableWAL = true;
ASSERT_OK(dbfull()->Put(writeOpt, handles_[1], "foo", "v2"));
ReopenWithColumnFamilies({"default", "pikachu"}, options);
// Both value's should be present.
ASSERT_EQ("v2", Get(1, "bar"));
ASSERT_EQ("v2", Get(1, "foo"));
writeOpt.disableWAL = true;
// This put, data is persisted by Flush
ASSERT_OK(dbfull()->Put(writeOpt, handles_[1], "bar", "v3"));
ReopenWithColumnFamilies({"default", "pikachu"}, options);
writeOpt.disableWAL = false;
// Data is persisted in the WAL
ASSERT_OK(dbfull()->Put(writeOpt, handles_[1], "zoo", "v3"));
// The hash does not match, write fails
fault_fs->SetChecksumHandoffFuncType(ChecksumType::kxxHash);
writeOpt.disableWAL = false;
ASSERT_NOK(dbfull()->Put(writeOpt, handles_[1], "foo", "v3"));
ReopenWithColumnFamilies({"default", "pikachu"}, options);
// Due to the write failure, Get should not find
ASSERT_NE("v3", Get(1, "foo"));
ASSERT_EQ("v3", Get(1, "zoo"));
ASSERT_EQ("v3", Get(1, "bar"));
fault_fs->SetChecksumHandoffFuncType(ChecksumType::kCRC32c);
// Each write will be similated as corrupted.
fault_fs->IngestDataCorruptionBeforeWrite();
writeOpt.disableWAL = true;
ASSERT_OK(dbfull()->Put(writeOpt, handles_[1], "bar", "v4"));
writeOpt.disableWAL = false;
ASSERT_NOK(dbfull()->Put(writeOpt, handles_[1], "foo", "v4"));
ReopenWithColumnFamilies({"default", "pikachu"}, options);
ASSERT_NE("v4", Get(1, "foo"));
ASSERT_NE("v4", Get(1, "bar"));
fault_fs->NoDataCorruptionBeforeWrite();
fault_fs->SetChecksumHandoffFuncType(ChecksumType::kNoChecksum);
// The file system does not provide checksum method and verification.
writeOpt.disableWAL = true;
ASSERT_OK(dbfull()->Put(writeOpt, handles_[1], "bar", "v5"));
writeOpt.disableWAL = false;
ASSERT_OK(dbfull()->Put(writeOpt, handles_[1], "foo", "v5"));
ReopenWithColumnFamilies({"default", "pikachu"}, options);
ASSERT_EQ("v5", Get(1, "foo"));
ASSERT_EQ("v5", Get(1, "bar"));
Destroy(options);
} while (ChangeWalOptions());
#endif // ROCKSDB_ASSERT_STATUS_CHECKED
}
Revise LockWAL/UnlockWAL implementation (#11020) Summary: RocksDB has two public APIs: `DB::LockWAL()`/`DB::UnlockWAL()`. The current implementation acquires and releases the internal `DBImpl::log_write_mutex_`. According to the comment on `DBImpl::log_write_mutex_`: https://github.com/facebook/rocksdb/blob/7.8.fb/db/db_impl/db_impl.h#L2287:L2288 > Note: to avoid dealock, if needed to acquire both log_write_mutex_ and mutex_, the order should be first mutex_ and then log_write_mutex_. This puts limitations on how applications can use the `LockWAL()` API. After `LockWAL()` returns ok, then application should not perform any operation that acquires `mutex_`. Currently, the use case of `LockWAL()` is MyRocks implementing the MySQL storage engine handlerton `lock_hton_log` interface. The operation that MyRocks performs after `LockWAL()` is `GetSortedWalFiless()` which not only acquires mutex_, but also `log_write_mutex_`. There are two issues: 1. Applications using these two APIs may hang if one thread calls `GetSortedWalFiles()` after calling `LockWAL()` because log_write_mutex is not recursive. 2. Two threads may dead lock due to lock order inversion. To fix these issues, we can modify the implementation of LockWAL so that it does not keep `log_write_mutex_` held until UnlockWAL. To achieve the goal of locking the WAL, we can instead manually inject a write stall so that all future writes will be stopped. Pull Request resolved: https://github.com/facebook/rocksdb/pull/11020 Test Plan: make check Reviewed By: ajkr Differential Revision: D41785203 Pulled By: riversand963 fbshipit-source-id: 5ccb7a9c6eb9a2c3fa80fd2c399cc2568b8f89ce
2 years ago
TEST_F(DBWALTest, LockWal) {
do {
Options options = CurrentOptions();
options.create_if_missing = true;
DestroyAndReopen(options);
ASSERT_OK(Put("foo", "v"));
ASSERT_OK(Put("bar", "v"));
ASSERT_OK(db_->LockWAL());
// Verify writes are stopped
WriteOptions wopts;
wopts.no_slowdown = true;
Status s = db_->Put(wopts, "foo", "dontcare");
ASSERT_TRUE(s.IsIncomplete());
{
VectorLogPtr wals;
ASSERT_OK(db_->GetSortedWalFiles(wals));
ASSERT_FALSE(wals.empty());
}
Cleanup, improve, stress test LockWAL() (#11143) Summary: The previous API comments for LockWAL didn't provide much about why you might want to use it, and didn't really meet what one would infer its contract was. Also, LockWAL was not in db_stress / crash test. In this change: * Implement a counting semantics for LockWAL()+UnlockWAL(), so that they can safely be used concurrently across threads or recursively within a thread. This should make the API much less bug-prone and easier to use. * Make sure no UnlockWAL() is needed after non-OK LockWAL() (to match RocksDB conventions) * Make UnlockWAL() reliably return non-OK when there's no matching LockWAL() (for debug-ability) * Clarify API comments on LockWAL(), UnlockWAL(), FlushWAL(), and SyncWAL(). Their exact meanings are not obvious, and I don't think it's appropriate to talk about implementation mutexes in the API comments, but about what operations might block each other. * Add LockWAL()/UnlockWAL() to db_stress and crash test, mostly to check for assertion failures, but also checks that latest seqno doesn't change while WAL is locked. This is simpler to add when LockWAL() is allowed in multiple threads. * Remove unnecessary use of sync points in test DBWALTest::LockWal. There was a bug during development of above changes that caused this test to fail sporadically, with and without this sync point change. Pull Request resolved: https://github.com/facebook/rocksdb/pull/11143 Test Plan: unit tests added / updated, added to stress/crash test Reviewed By: ajkr Differential Revision: D42848627 Pulled By: pdillinger fbshipit-source-id: 6d976c51791941a31fd8fbf28b0f82e888d9f4b4
2 years ago
port::Thread worker([&]() {
Status tmp_s = db_->Flush(FlushOptions());
ASSERT_OK(tmp_s);
});
Revise LockWAL/UnlockWAL implementation (#11020) Summary: RocksDB has two public APIs: `DB::LockWAL()`/`DB::UnlockWAL()`. The current implementation acquires and releases the internal `DBImpl::log_write_mutex_`. According to the comment on `DBImpl::log_write_mutex_`: https://github.com/facebook/rocksdb/blob/7.8.fb/db/db_impl/db_impl.h#L2287:L2288 > Note: to avoid dealock, if needed to acquire both log_write_mutex_ and mutex_, the order should be first mutex_ and then log_write_mutex_. This puts limitations on how applications can use the `LockWAL()` API. After `LockWAL()` returns ok, then application should not perform any operation that acquires `mutex_`. Currently, the use case of `LockWAL()` is MyRocks implementing the MySQL storage engine handlerton `lock_hton_log` interface. The operation that MyRocks performs after `LockWAL()` is `GetSortedWalFiless()` which not only acquires mutex_, but also `log_write_mutex_`. There are two issues: 1. Applications using these two APIs may hang if one thread calls `GetSortedWalFiles()` after calling `LockWAL()` because log_write_mutex is not recursive. 2. Two threads may dead lock due to lock order inversion. To fix these issues, we can modify the implementation of LockWAL so that it does not keep `log_write_mutex_` held until UnlockWAL. To achieve the goal of locking the WAL, we can instead manually inject a write stall so that all future writes will be stopped. Pull Request resolved: https://github.com/facebook/rocksdb/pull/11020 Test Plan: make check Reviewed By: ajkr Differential Revision: D41785203 Pulled By: riversand963 fbshipit-source-id: 5ccb7a9c6eb9a2c3fa80fd2c399cc2568b8f89ce
2 years ago
FlushOptions flush_opts;
flush_opts.wait = false;
s = db_->Flush(flush_opts);
ASSERT_TRUE(s.IsTryAgain());
ASSERT_OK(db_->UnlockWAL());
ASSERT_OK(db_->Put(WriteOptions(), "foo", "dontcare"));
worker.join();
} while (ChangeWalOptions());
}
class DBRecoveryTestBlobError
: public DBWALTest,
public testing::WithParamInterface<std::string> {
public:
DBRecoveryTestBlobError() : sync_point_(GetParam()) {}
std::string sync_point_;
};
INSTANTIATE_TEST_CASE_P(DBRecoveryTestBlobError, DBRecoveryTestBlobError,
::testing::ValuesIn(std::vector<std::string>{
"BlobFileBuilder::WriteBlobToFile:AddRecord",
"BlobFileBuilder::WriteBlobToFile:AppendFooter"}));
TEST_P(DBRecoveryTestBlobError, RecoverWithBlobError) {
// Write a value. Note that blob files are not actually enabled at this point.
ASSERT_OK(Put("key", "blob"));
// Reopen with blob files enabled but make blob file writing fail during
// recovery.
SyncPoint::GetInstance()->SetCallBack(sync_point_, [this](void* arg) {
Status* const s = static_cast<Status*>(arg);
assert(s);
(*s) = Status::IOError(sync_point_);
});
SyncPoint::GetInstance()->EnableProcessing();
Options options;
options.enable_blob_files = true;
options.avoid_flush_during_recovery = false;
options.disable_auto_compactions = true;
options.env = env_;
ASSERT_NOK(TryReopen(options));
SyncPoint::GetInstance()->DisableProcessing();
SyncPoint::GetInstance()->ClearAllCallBacks();
// Make sure the files generated by the failed recovery have been deleted.
std::vector<std::string> files;
ASSERT_OK(env_->GetChildren(dbname_, &files));
for (const auto& file : files) {
uint64_t number = 0;
FileType type = kTableFile;
if (!ParseFileName(file, &number, &type)) {
continue;
}
ASSERT_NE(type, kTableFile);
ASSERT_NE(type, kBlobFile);
}
}
TEST_F(DBWALTest, IgnoreRecoveredLog) {
std::string backup_logs = dbname_ + "/backup_logs";
do {
// delete old files in backup_logs directory
ASSERT_OK(env_->CreateDirIfMissing(backup_logs));
std::vector<std::string> old_files;
ASSERT_OK(env_->GetChildren(backup_logs, &old_files));
for (auto& file : old_files) {
ASSERT_OK(env_->DeleteFile(backup_logs + "/" + file));
}
Options options = CurrentOptions();
options.create_if_missing = true;
options.merge_operator = MergeOperators::CreateUInt64AddOperator();
options.wal_dir = dbname_ + "/logs";
DestroyAndReopen(options);
// fill up the DB
std::string one, two;
PutFixed64(&one, 1);
PutFixed64(&two, 2);
ASSERT_OK(db_->Merge(WriteOptions(), Slice("foo"), Slice(one)));
ASSERT_OK(db_->Merge(WriteOptions(), Slice("foo"), Slice(one)));
ASSERT_OK(db_->Merge(WriteOptions(), Slice("bar"), Slice(one)));
// copy the logs to backup
std::vector<std::string> logs;
ASSERT_OK(env_->GetChildren(options.wal_dir, &logs));
for (auto& log : logs) {
CopyFile(options.wal_dir + "/" + log, backup_logs + "/" + log);
}
// recover the DB
Reopen(options);
ASSERT_EQ(two, Get("foo"));
ASSERT_EQ(one, Get("bar"));
Close();
// copy the logs from backup back to wal dir
for (auto& log : logs) {
CopyFile(backup_logs + "/" + log, options.wal_dir + "/" + log);
}
// this should ignore the log files, recovery should not happen again
// if the recovery happens, the same merge operator would be called twice,
// leading to incorrect results
Reopen(options);
ASSERT_EQ(two, Get("foo"));
ASSERT_EQ(one, Get("bar"));
Close();
Destroy(options);
Reopen(options);
Close();
// copy the logs from backup back to wal dir
ASSERT_OK(env_->CreateDirIfMissing(options.wal_dir));
for (auto& log : logs) {
CopyFile(backup_logs + "/" + log, options.wal_dir + "/" + log);
}
// assert that we successfully recovered only from logs, even though we
// destroyed the DB
Reopen(options);
ASSERT_EQ(two, Get("foo"));
ASSERT_EQ(one, Get("bar"));
// Recovery will fail if DB directory doesn't exist.
Destroy(options);
// copy the logs from backup back to wal dir
ASSERT_OK(env_->CreateDirIfMissing(options.wal_dir));
for (auto& log : logs) {
CopyFile(backup_logs + "/" + log, options.wal_dir + "/" + log);
// we won't be needing this file no more
ASSERT_OK(env_->DeleteFile(backup_logs + "/" + log));
}
Status s = TryReopen(options);
ASSERT_NOK(s);
Destroy(options);
} while (ChangeWalOptions());
}
TEST_F(DBWALTest, RecoveryWithEmptyLog) {
do {
CreateAndReopenWithCF({"pikachu"}, CurrentOptions());
ASSERT_OK(Put(1, "foo", "v1"));
ASSERT_OK(Put(1, "foo", "v2"));
ReopenWithColumnFamilies({"default", "pikachu"}, CurrentOptions());
ReopenWithColumnFamilies({"default", "pikachu"}, CurrentOptions());
ASSERT_OK(Put(1, "foo", "v3"));
ReopenWithColumnFamilies({"default", "pikachu"}, CurrentOptions());
ASSERT_EQ("v3", Get(1, "foo"));
} while (ChangeWalOptions());
}
#if !(defined NDEBUG) || !defined(OS_WIN)
TEST_F(DBWALTest, PreallocateBlock) {
Options options = CurrentOptions();
options.write_buffer_size = 10 * 1000 * 1000;
options.max_total_wal_size = 0;
size_t expected_preallocation_size = static_cast<size_t>(
options.write_buffer_size + options.write_buffer_size / 10);
DestroyAndReopen(options);
std::atomic<int> called(0);
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->SetCallBack(
"DBTestWalFile.GetPreallocationStatus", [&](void* arg) {
ASSERT_TRUE(arg != nullptr);
size_t preallocation_size = *(static_cast<size_t*>(arg));
ASSERT_EQ(expected_preallocation_size, preallocation_size);
called.fetch_add(1);
});
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->EnableProcessing();
ASSERT_OK(Put("", ""));
ASSERT_OK(Flush());
ASSERT_OK(Put("", ""));
Close();
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->DisableProcessing();
ASSERT_EQ(2, called.load());
options.max_total_wal_size = 1000 * 1000;
expected_preallocation_size = static_cast<size_t>(options.max_total_wal_size);
Reopen(options);
called.store(0);
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->SetCallBack(
"DBTestWalFile.GetPreallocationStatus", [&](void* arg) {
ASSERT_TRUE(arg != nullptr);
size_t preallocation_size = *(static_cast<size_t*>(arg));
ASSERT_EQ(expected_preallocation_size, preallocation_size);
called.fetch_add(1);
});
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->EnableProcessing();
ASSERT_OK(Put("", ""));
ASSERT_OK(Flush());
ASSERT_OK(Put("", ""));
Close();
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->DisableProcessing();
ASSERT_EQ(2, called.load());
options.db_write_buffer_size = 800 * 1000;
expected_preallocation_size =
static_cast<size_t>(options.db_write_buffer_size);
Reopen(options);
called.store(0);
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->SetCallBack(
"DBTestWalFile.GetPreallocationStatus", [&](void* arg) {
ASSERT_TRUE(arg != nullptr);
size_t preallocation_size = *(static_cast<size_t*>(arg));
ASSERT_EQ(expected_preallocation_size, preallocation_size);
called.fetch_add(1);
});
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->EnableProcessing();
ASSERT_OK(Put("", ""));
ASSERT_OK(Flush());
ASSERT_OK(Put("", ""));
Close();
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->DisableProcessing();
ASSERT_EQ(2, called.load());
expected_preallocation_size = 700 * 1000;
std::shared_ptr<WriteBufferManager> write_buffer_manager =
std::make_shared<WriteBufferManager>(static_cast<uint64_t>(700 * 1000));
options.write_buffer_manager = write_buffer_manager;
Reopen(options);
called.store(0);
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->SetCallBack(
"DBTestWalFile.GetPreallocationStatus", [&](void* arg) {
ASSERT_TRUE(arg != nullptr);
size_t preallocation_size = *(static_cast<size_t*>(arg));
ASSERT_EQ(expected_preallocation_size, preallocation_size);
called.fetch_add(1);
});
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->EnableProcessing();
ASSERT_OK(Put("", ""));
ASSERT_OK(Flush());
ASSERT_OK(Put("", ""));
Close();
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->DisableProcessing();
ASSERT_EQ(2, called.load());
}
#endif // !(defined NDEBUG) || !defined(OS_WIN)
TEST_F(DBWALTest, DISABLED_FullPurgePreservesRecycledLog) {
// TODO(ajkr): Disabled until WAL recycling is fixed for
// `kPointInTimeRecovery`.
// For github issue #1303
for (int i = 0; i < 2; ++i) {
Options options = CurrentOptions();
options.create_if_missing = true;
options.recycle_log_file_num = 2;
if (i != 0) {
options.wal_dir = alternative_wal_dir_;
}
DestroyAndReopen(options);
ASSERT_OK(Put("foo", "v1"));
VectorLogPtr log_files;
ASSERT_OK(dbfull()->GetSortedWalFiles(log_files));
ASSERT_GT(log_files.size(), 0);
ASSERT_OK(Flush());
// Now the original WAL is in log_files[0] and should be marked for
// recycling.
// Verify full purge cannot remove this file.
JobContext job_context(0);
dbfull()->TEST_LockMutex();
dbfull()->FindObsoleteFiles(&job_context, true /* force */);
dbfull()->TEST_UnlockMutex();
dbfull()->PurgeObsoleteFiles(job_context);
if (i == 0) {
ASSERT_OK(
env_->FileExists(LogFileName(dbname_, log_files[0]->LogNumber())));
} else {
ASSERT_OK(env_->FileExists(
LogFileName(alternative_wal_dir_, log_files[0]->LogNumber())));
}
}
}
TEST_F(DBWALTest, DISABLED_FullPurgePreservesLogPendingReuse) {
// TODO(ajkr): Disabled until WAL recycling is fixed for
// `kPointInTimeRecovery`.
// Ensures full purge cannot delete a WAL while it's in the process of being
// recycled. In particular, we force the full purge after a file has been
// chosen for reuse, but before it has been renamed.
for (int i = 0; i < 2; ++i) {
Options options = CurrentOptions();
options.recycle_log_file_num = 1;
if (i != 0) {
options.wal_dir = alternative_wal_dir_;
}
DestroyAndReopen(options);
// The first flush creates a second log so writes can continue before the
// flush finishes.
ASSERT_OK(Put("foo", "bar"));
ASSERT_OK(Flush());
// The second flush can recycle the first log. Sync points enforce the
// full purge happens after choosing the log to recycle and before it is
// renamed.
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->LoadDependency({
{"DBImpl::CreateWAL:BeforeReuseWritableFile1",
"DBWALTest::FullPurgePreservesLogPendingReuse:PreFullPurge"},
{"DBWALTest::FullPurgePreservesLogPendingReuse:PostFullPurge",
"DBImpl::CreateWAL:BeforeReuseWritableFile2"},
});
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->EnableProcessing();
ROCKSDB_NAMESPACE::port::Thread thread([&]() {
TEST_SYNC_POINT(
"DBWALTest::FullPurgePreservesLogPendingReuse:PreFullPurge");
ASSERT_OK(db_->EnableFileDeletions(true));
TEST_SYNC_POINT(
"DBWALTest::FullPurgePreservesLogPendingReuse:PostFullPurge");
});
ASSERT_OK(Put("foo", "bar"));
ASSERT_OK(Flush());
thread.join();
}
}
TEST_F(DBWALTest, GetSortedWalFiles) {
do {
CreateAndReopenWithCF({"pikachu"}, CurrentOptions());
VectorLogPtr log_files;
ASSERT_OK(dbfull()->GetSortedWalFiles(log_files));
ASSERT_EQ(0, log_files.size());
ASSERT_OK(Put(1, "foo", "v1"));
ASSERT_OK(dbfull()->GetSortedWalFiles(log_files));
ASSERT_EQ(1, log_files.size());
} while (ChangeWalOptions());
}
TEST_F(DBWALTest, GetCurrentWalFile) {
do {
CreateAndReopenWithCF({"pikachu"}, CurrentOptions());
std::unique_ptr<LogFile>* bad_log_file = nullptr;
ASSERT_NOK(dbfull()->GetCurrentWalFile(bad_log_file));
std::unique_ptr<LogFile> log_file;
ASSERT_OK(dbfull()->GetCurrentWalFile(&log_file));
// nothing has been written to the log yet
ASSERT_EQ(log_file->StartSequence(), 0);
ASSERT_EQ(log_file->SizeFileBytes(), 0);
ASSERT_EQ(log_file->Type(), kAliveLogFile);
ASSERT_GT(log_file->LogNumber(), 0);
// add some data and verify that the file size actually moves foward
ASSERT_OK(Put(0, "foo", "v1"));
ASSERT_OK(Put(0, "foo2", "v2"));
ASSERT_OK(Put(0, "foo3", "v3"));
ASSERT_OK(dbfull()->GetCurrentWalFile(&log_file));
ASSERT_EQ(log_file->StartSequence(), 0);
ASSERT_GT(log_file->SizeFileBytes(), 0);
ASSERT_EQ(log_file->Type(), kAliveLogFile);
ASSERT_GT(log_file->LogNumber(), 0);
// force log files to cycle and add some more data, then check if
// log number moves forward
ReopenWithColumnFamilies({"default", "pikachu"}, CurrentOptions());
for (int i = 0; i < 10; i++) {
ReopenWithColumnFamilies({"default", "pikachu"}, CurrentOptions());
}
ASSERT_OK(Put(0, "foo4", "v4"));
ASSERT_OK(Put(0, "foo5", "v5"));
ASSERT_OK(Put(0, "foo6", "v6"));
ASSERT_OK(dbfull()->GetCurrentWalFile(&log_file));
ASSERT_EQ(log_file->StartSequence(), 0);
ASSERT_GT(log_file->SizeFileBytes(), 0);
ASSERT_EQ(log_file->Type(), kAliveLogFile);
ASSERT_GT(log_file->LogNumber(), 0);
} while (ChangeWalOptions());
}
TEST_F(DBWALTest, RecoveryWithLogDataForSomeCFs) {
// Test for regression of WAL cleanup missing files that don't contain data
// for every column family.
do {
CreateAndReopenWithCF({"pikachu"}, CurrentOptions());
ASSERT_OK(Put(1, "foo", "v1"));
ASSERT_OK(Put(1, "foo", "v2"));
uint64_t earliest_log_nums[2];
for (int i = 0; i < 2; ++i) {
if (i > 0) {
ReopenWithColumnFamilies({"default", "pikachu"}, CurrentOptions());
}
VectorLogPtr log_files;
ASSERT_OK(dbfull()->GetSortedWalFiles(log_files));
if (log_files.size() > 0) {
earliest_log_nums[i] = log_files[0]->LogNumber();
} else {
earliest_log_nums[i] = std::numeric_limits<uint64_t>::max();
}
}
// Check at least the first WAL was cleaned up during the recovery.
ASSERT_LT(earliest_log_nums[0], earliest_log_nums[1]);
} while (ChangeWalOptions());
}
TEST_F(DBWALTest, RecoverWithLargeLog) {
do {
{
Options options = CurrentOptions();
CreateAndReopenWithCF({"pikachu"}, options);
ASSERT_OK(Put(1, "big1", std::string(200000, '1')));
ASSERT_OK(Put(1, "big2", std::string(200000, '2')));
ASSERT_OK(Put(1, "small3", std::string(10, '3')));
ASSERT_OK(Put(1, "small4", std::string(10, '4')));
ASSERT_EQ(NumTableFilesAtLevel(0, 1), 0);
}
// Make sure that if we re-open with a small write buffer size that
// we flush table files in the middle of a large log file.
Options options;
options.write_buffer_size = 100000;
options = CurrentOptions(options);
ReopenWithColumnFamilies({"default", "pikachu"}, options);
ASSERT_EQ(NumTableFilesAtLevel(0, 1), 3);
ASSERT_EQ(std::string(200000, '1'), Get(1, "big1"));
ASSERT_EQ(std::string(200000, '2'), Get(1, "big2"));
ASSERT_EQ(std::string(10, '3'), Get(1, "small3"));
ASSERT_EQ(std::string(10, '4'), Get(1, "small4"));
ASSERT_GT(NumTableFilesAtLevel(0, 1), 1);
} while (ChangeWalOptions());
}
// In https://reviews.facebook.net/D20661 we change
// recovery behavior: previously for each log file each column family
// memtable was flushed, even it was empty. Now it's changed:
// we try to create the smallest number of table files by merging
// updates from multiple logs
TEST_F(DBWALTest, RecoverCheckFileAmountWithSmallWriteBuffer) {
Options options = CurrentOptions();
options.write_buffer_size = 5000000;
CreateAndReopenWithCF({"pikachu", "dobrynia", "nikitich"}, options);
// Since we will reopen DB with smaller write_buffer_size,
// each key will go to new SST file
ASSERT_OK(Put(1, Key(10), DummyString(1000000)));
ASSERT_OK(Put(1, Key(10), DummyString(1000000)));
ASSERT_OK(Put(1, Key(10), DummyString(1000000)));
ASSERT_OK(Put(1, Key(10), DummyString(1000000)));
ASSERT_OK(Put(3, Key(10), DummyString(1)));
// Make 'dobrynia' to be flushed and new WAL file to be created
ASSERT_OK(Put(2, Key(10), DummyString(7500000)));
ASSERT_OK(Put(2, Key(1), DummyString(1)));
ASSERT_OK(dbfull()->TEST_WaitForFlushMemTable(handles_[2]));
{
auto tables = ListTableFiles(env_, dbname_);
ASSERT_EQ(tables.size(), static_cast<size_t>(1));
// Make sure 'dobrynia' was flushed: check sst files amount
ASSERT_EQ(GetNumberOfSstFilesForColumnFamily(db_, "dobrynia"),
static_cast<uint64_t>(1));
}
// New WAL file
ASSERT_OK(Put(1, Key(1), DummyString(1)));
ASSERT_OK(Put(1, Key(1), DummyString(1)));
ASSERT_OK(Put(3, Key(10), DummyString(1)));
ASSERT_OK(Put(3, Key(10), DummyString(1)));
ASSERT_OK(Put(3, Key(10), DummyString(1)));
options.write_buffer_size = 4096;
options.arena_block_size = 4096;
ReopenWithColumnFamilies({"default", "pikachu", "dobrynia", "nikitich"},
options);
{
// No inserts => default is empty
ASSERT_EQ(GetNumberOfSstFilesForColumnFamily(db_, "default"),
static_cast<uint64_t>(0));
// First 4 keys goes to separate SSTs + 1 more SST for 2 smaller keys
ASSERT_EQ(GetNumberOfSstFilesForColumnFamily(db_, "pikachu"),
static_cast<uint64_t>(5));
// 1 SST for big key + 1 SST for small one
ASSERT_EQ(GetNumberOfSstFilesForColumnFamily(db_, "dobrynia"),
static_cast<uint64_t>(2));
// 1 SST for all keys
ASSERT_EQ(GetNumberOfSstFilesForColumnFamily(db_, "nikitich"),
static_cast<uint64_t>(1));
}
}
// In https://reviews.facebook.net/D20661 we change
// recovery behavior: previously for each log file each column family
// memtable was flushed, even it wasn't empty. Now it's changed:
// we try to create the smallest number of table files by merging
// updates from multiple logs
TEST_F(DBWALTest, RecoverCheckFileAmount) {
Options options = CurrentOptions();
options.write_buffer_size = 100000;
options.arena_block_size = 4 * 1024;
options.avoid_flush_during_recovery = false;
CreateAndReopenWithCF({"pikachu", "dobrynia", "nikitich"}, options);
ASSERT_OK(Put(0, Key(1), DummyString(1)));
ASSERT_OK(Put(1, Key(1), DummyString(1)));
ASSERT_OK(Put(2, Key(1), DummyString(1)));
// Make 'nikitich' memtable to be flushed
ASSERT_OK(Put(3, Key(10), DummyString(1002400)));
ASSERT_OK(Put(3, Key(1), DummyString(1)));
ASSERT_OK(dbfull()->TEST_WaitForFlushMemTable(handles_[3]));
// 4 memtable are not flushed, 1 sst file
{
auto tables = ListTableFiles(env_, dbname_);
ASSERT_EQ(tables.size(), static_cast<size_t>(1));
ASSERT_EQ(GetNumberOfSstFilesForColumnFamily(db_, "nikitich"),
static_cast<uint64_t>(1));
}
// Memtable for 'nikitich' has flushed, new WAL file has opened
// 4 memtable still not flushed
// Write to new WAL file
ASSERT_OK(Put(0, Key(1), DummyString(1)));
ASSERT_OK(Put(1, Key(1), DummyString(1)));
ASSERT_OK(Put(2, Key(1), DummyString(1)));
// Fill up 'nikitich' one more time
ASSERT_OK(Put(3, Key(10), DummyString(1002400)));
// make it flush
ASSERT_OK(Put(3, Key(1), DummyString(1)));
ASSERT_OK(dbfull()->TEST_WaitForFlushMemTable(handles_[3]));
// There are still 4 memtable not flushed, and 2 sst tables
ASSERT_OK(Put(0, Key(1), DummyString(1)));
ASSERT_OK(Put(1, Key(1), DummyString(1)));
ASSERT_OK(Put(2, Key(1), DummyString(1)));
{
auto tables = ListTableFiles(env_, dbname_);
ASSERT_EQ(tables.size(), static_cast<size_t>(2));
ASSERT_EQ(GetNumberOfSstFilesForColumnFamily(db_, "nikitich"),
static_cast<uint64_t>(2));
}
ReopenWithColumnFamilies({"default", "pikachu", "dobrynia", "nikitich"},
options);
{
std::vector<uint64_t> table_files = ListTableFiles(env_, dbname_);
// Check, that records for 'default', 'dobrynia' and 'pikachu' from
// first, second and third WALs went to the same SST.
// So, there is 6 SSTs: three for 'nikitich', one for 'default', one for
// 'dobrynia', one for 'pikachu'
ASSERT_EQ(GetNumberOfSstFilesForColumnFamily(db_, "default"),
static_cast<uint64_t>(1));
ASSERT_EQ(GetNumberOfSstFilesForColumnFamily(db_, "nikitich"),
static_cast<uint64_t>(3));
ASSERT_EQ(GetNumberOfSstFilesForColumnFamily(db_, "dobrynia"),
static_cast<uint64_t>(1));
ASSERT_EQ(GetNumberOfSstFilesForColumnFamily(db_, "pikachu"),
static_cast<uint64_t>(1));
}
}
TEST_F(DBWALTest, SyncMultipleLogs) {
const uint64_t kNumBatches = 2;
const int kBatchSize = 1000;
Options options = CurrentOptions();
options.create_if_missing = true;
options.write_buffer_size = 4096;
Reopen(options);
WriteBatch batch;
WriteOptions wo;
wo.sync = true;
for (uint64_t b = 0; b < kNumBatches; b++) {
batch.Clear();
for (int i = 0; i < kBatchSize; i++) {
ASSERT_OK(batch.Put(Key(i), DummyString(128)));
}
ASSERT_OK(dbfull()->Write(wo, &batch));
}
ASSERT_OK(dbfull()->SyncWAL());
}
// Github issue 1339. Prior the fix we read sequence id from the first log to
// a local variable, then keep increase the variable as we replay logs,
// ignoring actual sequence id of the records. This is incorrect if some writes
// come with WAL disabled.
TEST_F(DBWALTest, PartOfWritesWithWALDisabled) {
std::unique_ptr<FaultInjectionTestEnv> fault_env(
new FaultInjectionTestEnv(env_));
Options options = CurrentOptions();
options.env = fault_env.get();
options.disable_auto_compactions = true;
WriteOptions wal_on, wal_off;
wal_on.sync = true;
wal_on.disableWAL = false;
wal_off.disableWAL = true;
CreateAndReopenWithCF({"dummy"}, options);
ASSERT_OK(Put(1, "dummy", "d1", wal_on)); // seq id 1
ASSERT_OK(Put(1, "dummy", "d2", wal_off));
ASSERT_OK(Put(1, "dummy", "d3", wal_off));
ASSERT_OK(Put(0, "key", "v4", wal_on)); // seq id 4
ASSERT_OK(Flush(0));
ASSERT_OK(Put(0, "key", "v5", wal_on)); // seq id 5
ASSERT_EQ("v5", Get(0, "key"));
ASSERT_OK(dbfull()->FlushWAL(false));
// Simulate a crash.
fault_env->SetFilesystemActive(false);
Close();
fault_env->ResetState();
ReopenWithColumnFamilies({"default", "dummy"}, options);
// Prior to the fix, we may incorrectly recover "v5" with sequence id = 3.
ASSERT_EQ("v5", Get(0, "key"));
// Destroy DB before destruct fault_env.
Destroy(options);
}
//
// Test WAL recovery for the various modes available
//
class RecoveryTestHelper {
public:
// Number of WAL files to generate
static constexpr int kWALFilesCount = 10;
// Starting number for the WAL file name like 00010.log
static constexpr int kWALFileOffset = 10;
// Keys to be written per WAL file
static constexpr int kKeysPerWALFile = 133;
// Size of the value
static constexpr int kValueSize = 96;
// Create WAL files with values filled in
static void FillData(DBWALTestBase* test, const Options& options,
const size_t wal_count, size_t* count) {
// Calling internal functions requires sanitized options.
Options sanitized_options = SanitizeOptions(test->dbname_, options);
const ImmutableDBOptions db_options(sanitized_options);
*count = 0;
std::shared_ptr<Cache> table_cache = NewLRUCache(50, 0);
FileOptions file_options;
WriteBufferManager write_buffer_manager(db_options.db_write_buffer_size);
std::unique_ptr<VersionSet> versions;
std::unique_ptr<WalManager> wal_manager;
WriteController write_controller;
versions.reset(new VersionSet(
test->dbname_, &db_options, file_options, table_cache.get(),
&write_buffer_manager, &write_controller,
/*block_cache_tracer=*/nullptr,
/*io_tracer=*/nullptr, /*db_id*/ "", /*db_session_id*/ ""));
wal_manager.reset(
new WalManager(db_options, file_options, /*io_tracer=*/nullptr));
std::unique_ptr<log::Writer> current_log_writer;
for (size_t j = kWALFileOffset; j < wal_count + kWALFileOffset; j++) {
uint64_t current_log_number = j;
std::string fname = LogFileName(test->dbname_, current_log_number);
std::unique_ptr<WritableFileWriter> file_writer;
ASSERT_OK(WritableFileWriter::Create(db_options.env->GetFileSystem(),
fname, file_options, &file_writer,
nullptr));
log::Writer* log_writer =
new log::Writer(std::move(file_writer), current_log_number,
db_options.recycle_log_file_num > 0, false,
db_options.wal_compression);
ASSERT_OK(log_writer->AddCompressionTypeRecord());
current_log_writer.reset(log_writer);
WriteBatch batch;
for (int i = 0; i < kKeysPerWALFile; i++) {
std::string key = "key" + std::to_string((*count)++);
std::string value = test->DummyString(kValueSize);
ASSERT_NE(current_log_writer.get(), nullptr);
uint64_t seq = versions->LastSequence() + 1;
batch.Clear();
ASSERT_OK(batch.Put(key, value));
WriteBatchInternal::SetSequence(&batch, seq);
ASSERT_OK(current_log_writer->AddRecord(
WriteBatchInternal::Contents(&batch)));
versions->SetLastAllocatedSequence(seq);
versions->SetLastPublishedSequence(seq);
versions->SetLastSequence(seq);
}
}
}
// Recreate and fill the store with some data
static size_t FillData(DBWALTestBase* test, Options* options) {
options->create_if_missing = true;
test->DestroyAndReopen(*options);
test->Close();
size_t count = 0;
FillData(test, *options, kWALFilesCount, &count);
return count;
}
// Read back all the keys we wrote and return the number of keys found
static size_t GetData(DBWALTestBase* test) {
size_t count = 0;
for (size_t i = 0; i < kWALFilesCount * kKeysPerWALFile; i++) {
if (test->Get("key" + std::to_string(i)) != "NOT_FOUND") {
++count;
}
}
return count;
}
// Manuall corrupt the specified WAL
static void CorruptWAL(DBWALTestBase* test, const Options& options,
const double off, const double len,
const int wal_file_id, const bool trunc = false) {
Env* env = options.env;
std::string fname = LogFileName(test->dbname_, wal_file_id);
uint64_t size;
ASSERT_OK(env->GetFileSize(fname, &size));
ASSERT_GT(size, 0);
#ifdef OS_WIN
// Windows disk cache behaves differently. When we truncate
// the original content is still in the cache due to the original
// handle is still open. Generally, in Windows, one prohibits
// shared access to files and it is not needed for WAL but we allow
// it to induce corruption at various tests.
test->Close();
#endif
if (trunc) {
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566) Summary: This PR does a few things: 1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation. 2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed: - The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated - The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory). 3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10). I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged. Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently. Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :) Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566 Reviewed By: zhichao-cao Differential Revision: D24408980 Pulled By: jay-zhuang fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
4 years ago
ASSERT_OK(
test::TruncateFile(env, fname, static_cast<uint64_t>(size * off)));
} else {
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566) Summary: This PR does a few things: 1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation. 2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed: - The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated - The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory). 3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10). I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged. Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently. Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :) Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566 Reviewed By: zhichao-cao Differential Revision: D24408980 Pulled By: jay-zhuang fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
4 years ago
ASSERT_OK(test::CorruptFile(env, fname, static_cast<int>(size * off + 8),
static_cast<int>(size * len), false));
}
}
};
class DBWALTestWithParams : public DBWALTestBase,
public ::testing::WithParamInterface<
std::tuple<bool, int, int, CompressionType>> {
public:
DBWALTestWithParams() : DBWALTestBase("/db_wal_test_with_params") {}
};
INSTANTIATE_TEST_CASE_P(
Wal, DBWALTestWithParams,
::testing::Combine(::testing::Bool(), ::testing::Range(0, 4, 1),
::testing::Range(RecoveryTestHelper::kWALFileOffset,
RecoveryTestHelper::kWALFileOffset +
RecoveryTestHelper::kWALFilesCount,
1),
::testing::Values(CompressionType::kNoCompression,
CompressionType::kZSTD)));
class DBWALTestWithParamsVaryingRecoveryMode
: public DBWALTestBase,
public ::testing::WithParamInterface<
std::tuple<bool, int, int, WALRecoveryMode, CompressionType>> {
public:
DBWALTestWithParamsVaryingRecoveryMode()
: DBWALTestBase("/db_wal_test_with_params_mode") {}
};
INSTANTIATE_TEST_CASE_P(
Wal, DBWALTestWithParamsVaryingRecoveryMode,
::testing::Combine(
::testing::Bool(), ::testing::Range(0, 4, 1),
::testing::Range(RecoveryTestHelper::kWALFileOffset,
RecoveryTestHelper::kWALFileOffset +
RecoveryTestHelper::kWALFilesCount,
1),
::testing::Values(WALRecoveryMode::kTolerateCorruptedTailRecords,
WALRecoveryMode::kAbsoluteConsistency,
WALRecoveryMode::kPointInTimeRecovery,
WALRecoveryMode::kSkipAnyCorruptedRecords),
::testing::Values(CompressionType::kNoCompression,
CompressionType::kZSTD)));
// Test scope:
// - We expect to open the data store when there is incomplete trailing writes
// at the end of any of the logs
// - We do not expect to open the data store for corruption
TEST_P(DBWALTestWithParams, kTolerateCorruptedTailRecords) {
bool trunc = std::get<0>(GetParam()); // Corruption style
// Corruption offset position
int corrupt_offset = std::get<1>(GetParam());
int wal_file_id = std::get<2>(GetParam()); // WAL file
// Fill data for testing
Options options = CurrentOptions();
const size_t row_count = RecoveryTestHelper::FillData(this, &options);
// test checksum failure or parsing
RecoveryTestHelper::CorruptWAL(this, options, corrupt_offset * .3,
/*len%=*/.1, wal_file_id, trunc);
options.wal_recovery_mode = WALRecoveryMode::kTolerateCorruptedTailRecords;
if (trunc) {
options.create_if_missing = false;
ASSERT_OK(TryReopen(options));
const size_t recovered_row_count = RecoveryTestHelper::GetData(this);
ASSERT_TRUE(corrupt_offset == 0 || recovered_row_count > 0);
ASSERT_LT(recovered_row_count, row_count);
} else {
ASSERT_NOK(TryReopen(options));
}
}
// Test scope:
// We don't expect the data store to be opened if there is any corruption
// (leading, middle or trailing -- incomplete writes or corruption)
TEST_P(DBWALTestWithParams, kAbsoluteConsistency) {
// Verify clean slate behavior
Options options = CurrentOptions();
const size_t row_count = RecoveryTestHelper::FillData(this, &options);
options.create_if_missing = false;
ASSERT_OK(TryReopen(options));
ASSERT_EQ(RecoveryTestHelper::GetData(this), row_count);
bool trunc = std::get<0>(GetParam()); // Corruption style
// Corruption offset position
int corrupt_offset = std::get<1>(GetParam());
int wal_file_id = std::get<2>(GetParam()); // WAL file
// WAL compression type
CompressionType compression_type = std::get<3>(GetParam());
options.wal_compression = compression_type;
if (trunc && corrupt_offset == 0) {
return;
}
// fill with new date
RecoveryTestHelper::FillData(this, &options);
// corrupt the wal
RecoveryTestHelper::CorruptWAL(this, options, corrupt_offset * .33,
/*len%=*/.1, wal_file_id, trunc);
// verify
options.wal_recovery_mode = WALRecoveryMode::kAbsoluteConsistency;
options.create_if_missing = false;
ASSERT_NOK(TryReopen(options));
}
// Test scope:
// We don't expect the data store to be opened if there is any inconsistency
// between WAL and SST files
TEST_F(DBWALTest, kPointInTimeRecoveryCFConsistency) {
Options options = CurrentOptions();
options.avoid_flush_during_recovery = true;
// Create DB with multiple column families.
CreateAndReopenWithCF({"one", "two"}, options);
ASSERT_OK(Put(1, "key1", "val1"));
ASSERT_OK(Put(2, "key2", "val2"));
// Record the offset at this point
Env* env = options.env;
Skip deleted WALs during recovery Summary: This patch record min log number to keep to the manifest while flushing SST files to ignore them and any WAL older than them during recovery. This is to avoid scenarios when we have a gap between the WAL files are fed to the recovery procedure. The gap could happen by for example out-of-order WAL deletion. Such gap could cause problems in 2PC recovery where the prepared and commit entry are placed into two separate WAL and gap in the WALs could result into not processing the WAL with the commit entry and hence breaking the 2PC recovery logic. Before the commit, for 2PC case, we determined which log number to keep in FindObsoleteFiles(). We looked at the earliest logs with outstanding prepare entries, or prepare entries whose respective commit or abort are in memtable. With the commit, the same calculation is done while we apply the SST flush. Just before installing the flush file, we precompute the earliest log file to keep after the flush finishes using the same logic (but skipping the memtables just flushed), record this information to the manifest entry for this new flushed SST file. This pre-computed value is also remembered in memory, and will later be used to determine whether a log file can be deleted. This value is unlikely to change until next flush because the commit entry will stay in memtable. (In WritePrepared, we could have removed the older log files as soon as all prepared entries are committed. It's not yet done anyway. Even if we do it, the only thing we loss with this new approach is earlier log deletion between two flushes, which does not guarantee to happen anyway because the obsolete file clean-up function is only executed after flush or compaction) This min log number to keep is stored in the manifest using the safely-ignore customized field of AddFile entry, in order to guarantee that the DB generated using newer release can be opened by previous releases no older than 4.2. Closes https://github.com/facebook/rocksdb/pull/3765 Differential Revision: D7747618 Pulled By: siying fbshipit-source-id: d00c92105b4f83852e9754a1b70d6b64cb590729
7 years ago
uint64_t wal_file_id = dbfull()->TEST_LogfileNumber();
std::string fname = LogFileName(dbname_, wal_file_id);
uint64_t offset_to_corrupt;
ASSERT_OK(env->GetFileSize(fname, &offset_to_corrupt));
ASSERT_GT(offset_to_corrupt, 0);
ASSERT_OK(Put(1, "key3", "val3"));
// Corrupt WAL at location of key3
ASSERT_OK(test::CorruptFile(env, fname, static_cast<int>(offset_to_corrupt),
4, false));
ASSERT_OK(Put(2, "key4", "val4"));
ASSERT_OK(Put(1, "key5", "val5"));
ASSERT_OK(Flush(2));
// PIT recovery & verify
options.wal_recovery_mode = WALRecoveryMode::kPointInTimeRecovery;
ASSERT_NOK(TryReopenWithColumnFamilies({"default", "one", "two"}, options));
}
Fix a race condition in WAL tracking causing DB open failure (#9715) Summary: There is a race condition if WAL tracking in the MANIFEST is enabled in a database that disables 2PC. The race condition is between two background flush threads trying to install flush results to the MANIFEST. Consider an example database with two column families: "default" (cfd0) and "cf1" (cfd1). Initially, both column families have one mutable (active) memtable whose data backed by 6.log. 1. Trigger a manual flush for "cf1", creating a 7.log 2. Insert another key to "default", and trigger flush for "default", creating 8.log 3. BgFlushThread1 finishes writing 9.sst 4. BgFlushThread2 finishes writing 10.sst ``` Time BgFlushThread1 BgFlushThread2 | mutex_.Lock() | precompute min_wal_to_keep as 6 | mutex_.Unlock() | mutex_.Lock() | precompute min_wal_to_keep as 6 | join MANIFEST write queue and mutex_.Unlock() | write to MANIFEST | mutex_.Lock() | cfd1->log_number = 7 | Signal bg_flush_2 and mutex_.Unlock() | wake up and mutex_.Lock() | cfd0->log_number = 8 | FindObsoleteFiles() with job_context->log_number == 7 | mutex_.Unlock() | PurgeObsoleteFiles() deletes 6.log V ``` As shown in the above, BgFlushThread2 thinks that the min wal to keep is 6.log because "cf1" has unflushed data in 6.log (cf1.log_number=6). Similarly, BgThread1 thinks that min wal to keep is also 6.log because "default" has unflushed data (default.log_number=6). No WAL deletion will be written to MANIFEST because 6 is equal to `versions_->wals_.min_wal_number_to_keep`, due to https://github.com/facebook/rocksdb/blob/7.1.fb/db/memtable_list.cc#L513:L514. The bg flush thread that finishes last will perform file purging. `job_context.log_number` will be evaluated as 7, i.e. the min wal that contains unflushed data, causing 6.log to be deleted. However, MANIFEST thinks 6.log should still exist. If you close the db at this point, you won't be able to re-open it if `track_and_verify_wal_in_manifest` is true. We must handle the case of multiple bg flush threads, and it is difficult for one bg flush thread to know the correct min wal number until the other bg flush threads have finished committing to the manifest and updated the `cfd::log_number`. To fix this issue, we rename an existing variable `min_log_number_to_keep_2pc` to `min_log_number_to_keep`, and use it to track WAL file deletion in non-2pc mode as well. This variable is updated only 1) during recovery with mutex held, or 2) in the MANIFEST write thread. `min_log_number_to_keep` means RocksDB will delete WALs below it, although there may be WALs above it which are also obsolete. Formally, we will have [min_wal_to_keep, max_obsolete_wal]. During recovery, we make sure that only WALs above max_obsolete_wal are checked and added back to `alive_log_files_`. Pull Request resolved: https://github.com/facebook/rocksdb/pull/9715 Test Plan: ``` make check ``` Also ran stress test below (with asan) to make sure it completes successfully. ``` TEST_TMPDIR=/dev/shm/rocksdb OPT=-g ASAN_OPTIONS=disable_coredump=0 \ CRASH_TEST_EXT_ARGS=--compression_type=zstd SKIP_FORMAT_BUCK_CHECKS=1 \ make J=52 -j52 blackbox_asan_crash_test ``` Reviewed By: ltamasi Differential Revision: D34984412 Pulled By: riversand963 fbshipit-source-id: c7b21a8d84751bb55ea79c9f387103d21b231005
3 years ago
TEST_F(DBWALTest, RaceInstallFlushResultsWithWalObsoletion) {
Options options = CurrentOptions();
options.env = env_;
options.track_and_verify_wals_in_manifest = true;
// The following make sure there are two bg flush threads.
options.max_background_jobs = 8;
DestroyAndReopen(options);
Fix a race condition in WAL tracking causing DB open failure (#9715) Summary: There is a race condition if WAL tracking in the MANIFEST is enabled in a database that disables 2PC. The race condition is between two background flush threads trying to install flush results to the MANIFEST. Consider an example database with two column families: "default" (cfd0) and "cf1" (cfd1). Initially, both column families have one mutable (active) memtable whose data backed by 6.log. 1. Trigger a manual flush for "cf1", creating a 7.log 2. Insert another key to "default", and trigger flush for "default", creating 8.log 3. BgFlushThread1 finishes writing 9.sst 4. BgFlushThread2 finishes writing 10.sst ``` Time BgFlushThread1 BgFlushThread2 | mutex_.Lock() | precompute min_wal_to_keep as 6 | mutex_.Unlock() | mutex_.Lock() | precompute min_wal_to_keep as 6 | join MANIFEST write queue and mutex_.Unlock() | write to MANIFEST | mutex_.Lock() | cfd1->log_number = 7 | Signal bg_flush_2 and mutex_.Unlock() | wake up and mutex_.Lock() | cfd0->log_number = 8 | FindObsoleteFiles() with job_context->log_number == 7 | mutex_.Unlock() | PurgeObsoleteFiles() deletes 6.log V ``` As shown in the above, BgFlushThread2 thinks that the min wal to keep is 6.log because "cf1" has unflushed data in 6.log (cf1.log_number=6). Similarly, BgThread1 thinks that min wal to keep is also 6.log because "default" has unflushed data (default.log_number=6). No WAL deletion will be written to MANIFEST because 6 is equal to `versions_->wals_.min_wal_number_to_keep`, due to https://github.com/facebook/rocksdb/blob/7.1.fb/db/memtable_list.cc#L513:L514. The bg flush thread that finishes last will perform file purging. `job_context.log_number` will be evaluated as 7, i.e. the min wal that contains unflushed data, causing 6.log to be deleted. However, MANIFEST thinks 6.log should still exist. If you close the db at this point, you won't be able to re-open it if `track_and_verify_wal_in_manifest` is true. We must handle the case of multiple bg flush threads, and it is difficult for one bg flush thread to know the correct min wal number until the other bg flush threads have finished committing to the manifest and updated the `cfd::log_number`. To fix this issue, we rename an existing variable `min_log_number_to_keep_2pc` to `min_log_number_to_keep`, and use it to track WAL file deletion in non-2pc mode as well. This variable is updated only 1) during recovery with mutex held, or 2) in the MANIFEST write thread. `min_log_number_to_keep` means RocksDB will delete WALs below it, although there may be WALs above it which are also obsolete. Formally, we will have [min_wal_to_keep, max_obsolete_wal]. During recovery, we make sure that only WALs above max_obsolete_wal are checked and added back to `alive_log_files_`. Pull Request resolved: https://github.com/facebook/rocksdb/pull/9715 Test Plan: ``` make check ``` Also ran stress test below (with asan) to make sure it completes successfully. ``` TEST_TMPDIR=/dev/shm/rocksdb OPT=-g ASAN_OPTIONS=disable_coredump=0 \ CRASH_TEST_EXT_ARGS=--compression_type=zstd SKIP_FORMAT_BUCK_CHECKS=1 \ make J=52 -j52 blackbox_asan_crash_test ``` Reviewed By: ltamasi Differential Revision: D34984412 Pulled By: riversand963 fbshipit-source-id: c7b21a8d84751bb55ea79c9f387103d21b231005
3 years ago
const std::string cf1_name("cf1");
CreateAndReopenWithCF({cf1_name}, options);
assert(handles_.size() == 2);
{
dbfull()->TEST_LockMutex();
ASSERT_LE(2, dbfull()->GetBGJobLimits().max_flushes);
dbfull()->TEST_UnlockMutex();
}
ASSERT_OK(dbfull()->PauseBackgroundWork());
ASSERT_OK(db_->Put(WriteOptions(), handles_[1], "foo", "value"));
ASSERT_OK(db_->Put(WriteOptions(), "foo", "value"));
ASSERT_OK(dbfull()->TEST_FlushMemTable(
/*wait=*/false, /*allow_write_stall=*/true, handles_[1]));
Fix a race condition in WAL tracking causing DB open failure (#9715) Summary: There is a race condition if WAL tracking in the MANIFEST is enabled in a database that disables 2PC. The race condition is between two background flush threads trying to install flush results to the MANIFEST. Consider an example database with two column families: "default" (cfd0) and "cf1" (cfd1). Initially, both column families have one mutable (active) memtable whose data backed by 6.log. 1. Trigger a manual flush for "cf1", creating a 7.log 2. Insert another key to "default", and trigger flush for "default", creating 8.log 3. BgFlushThread1 finishes writing 9.sst 4. BgFlushThread2 finishes writing 10.sst ``` Time BgFlushThread1 BgFlushThread2 | mutex_.Lock() | precompute min_wal_to_keep as 6 | mutex_.Unlock() | mutex_.Lock() | precompute min_wal_to_keep as 6 | join MANIFEST write queue and mutex_.Unlock() | write to MANIFEST | mutex_.Lock() | cfd1->log_number = 7 | Signal bg_flush_2 and mutex_.Unlock() | wake up and mutex_.Lock() | cfd0->log_number = 8 | FindObsoleteFiles() with job_context->log_number == 7 | mutex_.Unlock() | PurgeObsoleteFiles() deletes 6.log V ``` As shown in the above, BgFlushThread2 thinks that the min wal to keep is 6.log because "cf1" has unflushed data in 6.log (cf1.log_number=6). Similarly, BgThread1 thinks that min wal to keep is also 6.log because "default" has unflushed data (default.log_number=6). No WAL deletion will be written to MANIFEST because 6 is equal to `versions_->wals_.min_wal_number_to_keep`, due to https://github.com/facebook/rocksdb/blob/7.1.fb/db/memtable_list.cc#L513:L514. The bg flush thread that finishes last will perform file purging. `job_context.log_number` will be evaluated as 7, i.e. the min wal that contains unflushed data, causing 6.log to be deleted. However, MANIFEST thinks 6.log should still exist. If you close the db at this point, you won't be able to re-open it if `track_and_verify_wal_in_manifest` is true. We must handle the case of multiple bg flush threads, and it is difficult for one bg flush thread to know the correct min wal number until the other bg flush threads have finished committing to the manifest and updated the `cfd::log_number`. To fix this issue, we rename an existing variable `min_log_number_to_keep_2pc` to `min_log_number_to_keep`, and use it to track WAL file deletion in non-2pc mode as well. This variable is updated only 1) during recovery with mutex held, or 2) in the MANIFEST write thread. `min_log_number_to_keep` means RocksDB will delete WALs below it, although there may be WALs above it which are also obsolete. Formally, we will have [min_wal_to_keep, max_obsolete_wal]. During recovery, we make sure that only WALs above max_obsolete_wal are checked and added back to `alive_log_files_`. Pull Request resolved: https://github.com/facebook/rocksdb/pull/9715 Test Plan: ``` make check ``` Also ran stress test below (with asan) to make sure it completes successfully. ``` TEST_TMPDIR=/dev/shm/rocksdb OPT=-g ASAN_OPTIONS=disable_coredump=0 \ CRASH_TEST_EXT_ARGS=--compression_type=zstd SKIP_FORMAT_BUCK_CHECKS=1 \ make J=52 -j52 blackbox_asan_crash_test ``` Reviewed By: ltamasi Differential Revision: D34984412 Pulled By: riversand963 fbshipit-source-id: c7b21a8d84751bb55ea79c9f387103d21b231005
3 years ago
ASSERT_OK(db_->Put(WriteOptions(), "foo", "value"));
ASSERT_OK(dbfull()->TEST_FlushMemTable(
/*wait=*/false, /*allow_write_stall=*/true, handles_[0]));
Fix a race condition in WAL tracking causing DB open failure (#9715) Summary: There is a race condition if WAL tracking in the MANIFEST is enabled in a database that disables 2PC. The race condition is between two background flush threads trying to install flush results to the MANIFEST. Consider an example database with two column families: "default" (cfd0) and "cf1" (cfd1). Initially, both column families have one mutable (active) memtable whose data backed by 6.log. 1. Trigger a manual flush for "cf1", creating a 7.log 2. Insert another key to "default", and trigger flush for "default", creating 8.log 3. BgFlushThread1 finishes writing 9.sst 4. BgFlushThread2 finishes writing 10.sst ``` Time BgFlushThread1 BgFlushThread2 | mutex_.Lock() | precompute min_wal_to_keep as 6 | mutex_.Unlock() | mutex_.Lock() | precompute min_wal_to_keep as 6 | join MANIFEST write queue and mutex_.Unlock() | write to MANIFEST | mutex_.Lock() | cfd1->log_number = 7 | Signal bg_flush_2 and mutex_.Unlock() | wake up and mutex_.Lock() | cfd0->log_number = 8 | FindObsoleteFiles() with job_context->log_number == 7 | mutex_.Unlock() | PurgeObsoleteFiles() deletes 6.log V ``` As shown in the above, BgFlushThread2 thinks that the min wal to keep is 6.log because "cf1" has unflushed data in 6.log (cf1.log_number=6). Similarly, BgThread1 thinks that min wal to keep is also 6.log because "default" has unflushed data (default.log_number=6). No WAL deletion will be written to MANIFEST because 6 is equal to `versions_->wals_.min_wal_number_to_keep`, due to https://github.com/facebook/rocksdb/blob/7.1.fb/db/memtable_list.cc#L513:L514. The bg flush thread that finishes last will perform file purging. `job_context.log_number` will be evaluated as 7, i.e. the min wal that contains unflushed data, causing 6.log to be deleted. However, MANIFEST thinks 6.log should still exist. If you close the db at this point, you won't be able to re-open it if `track_and_verify_wal_in_manifest` is true. We must handle the case of multiple bg flush threads, and it is difficult for one bg flush thread to know the correct min wal number until the other bg flush threads have finished committing to the manifest and updated the `cfd::log_number`. To fix this issue, we rename an existing variable `min_log_number_to_keep_2pc` to `min_log_number_to_keep`, and use it to track WAL file deletion in non-2pc mode as well. This variable is updated only 1) during recovery with mutex held, or 2) in the MANIFEST write thread. `min_log_number_to_keep` means RocksDB will delete WALs below it, although there may be WALs above it which are also obsolete. Formally, we will have [min_wal_to_keep, max_obsolete_wal]. During recovery, we make sure that only WALs above max_obsolete_wal are checked and added back to `alive_log_files_`. Pull Request resolved: https://github.com/facebook/rocksdb/pull/9715 Test Plan: ``` make check ``` Also ran stress test below (with asan) to make sure it completes successfully. ``` TEST_TMPDIR=/dev/shm/rocksdb OPT=-g ASAN_OPTIONS=disable_coredump=0 \ CRASH_TEST_EXT_ARGS=--compression_type=zstd SKIP_FORMAT_BUCK_CHECKS=1 \ make J=52 -j52 blackbox_asan_crash_test ``` Reviewed By: ltamasi Differential Revision: D34984412 Pulled By: riversand963 fbshipit-source-id: c7b21a8d84751bb55ea79c9f387103d21b231005
3 years ago
bool called = false;
std::atomic<int> bg_flush_threads{0};
std::atomic<bool> wal_synced{false};
Fix a race condition in WAL tracking causing DB open failure (#9715) Summary: There is a race condition if WAL tracking in the MANIFEST is enabled in a database that disables 2PC. The race condition is between two background flush threads trying to install flush results to the MANIFEST. Consider an example database with two column families: "default" (cfd0) and "cf1" (cfd1). Initially, both column families have one mutable (active) memtable whose data backed by 6.log. 1. Trigger a manual flush for "cf1", creating a 7.log 2. Insert another key to "default", and trigger flush for "default", creating 8.log 3. BgFlushThread1 finishes writing 9.sst 4. BgFlushThread2 finishes writing 10.sst ``` Time BgFlushThread1 BgFlushThread2 | mutex_.Lock() | precompute min_wal_to_keep as 6 | mutex_.Unlock() | mutex_.Lock() | precompute min_wal_to_keep as 6 | join MANIFEST write queue and mutex_.Unlock() | write to MANIFEST | mutex_.Lock() | cfd1->log_number = 7 | Signal bg_flush_2 and mutex_.Unlock() | wake up and mutex_.Lock() | cfd0->log_number = 8 | FindObsoleteFiles() with job_context->log_number == 7 | mutex_.Unlock() | PurgeObsoleteFiles() deletes 6.log V ``` As shown in the above, BgFlushThread2 thinks that the min wal to keep is 6.log because "cf1" has unflushed data in 6.log (cf1.log_number=6). Similarly, BgThread1 thinks that min wal to keep is also 6.log because "default" has unflushed data (default.log_number=6). No WAL deletion will be written to MANIFEST because 6 is equal to `versions_->wals_.min_wal_number_to_keep`, due to https://github.com/facebook/rocksdb/blob/7.1.fb/db/memtable_list.cc#L513:L514. The bg flush thread that finishes last will perform file purging. `job_context.log_number` will be evaluated as 7, i.e. the min wal that contains unflushed data, causing 6.log to be deleted. However, MANIFEST thinks 6.log should still exist. If you close the db at this point, you won't be able to re-open it if `track_and_verify_wal_in_manifest` is true. We must handle the case of multiple bg flush threads, and it is difficult for one bg flush thread to know the correct min wal number until the other bg flush threads have finished committing to the manifest and updated the `cfd::log_number`. To fix this issue, we rename an existing variable `min_log_number_to_keep_2pc` to `min_log_number_to_keep`, and use it to track WAL file deletion in non-2pc mode as well. This variable is updated only 1) during recovery with mutex held, or 2) in the MANIFEST write thread. `min_log_number_to_keep` means RocksDB will delete WALs below it, although there may be WALs above it which are also obsolete. Formally, we will have [min_wal_to_keep, max_obsolete_wal]. During recovery, we make sure that only WALs above max_obsolete_wal are checked and added back to `alive_log_files_`. Pull Request resolved: https://github.com/facebook/rocksdb/pull/9715 Test Plan: ``` make check ``` Also ran stress test below (with asan) to make sure it completes successfully. ``` TEST_TMPDIR=/dev/shm/rocksdb OPT=-g ASAN_OPTIONS=disable_coredump=0 \ CRASH_TEST_EXT_ARGS=--compression_type=zstd SKIP_FORMAT_BUCK_CHECKS=1 \ make J=52 -j52 blackbox_asan_crash_test ``` Reviewed By: ltamasi Differential Revision: D34984412 Pulled By: riversand963 fbshipit-source-id: c7b21a8d84751bb55ea79c9f387103d21b231005
3 years ago
SyncPoint::GetInstance()->DisableProcessing();
SyncPoint::GetInstance()->ClearAllCallBacks();
SyncPoint::GetInstance()->SetCallBack(
"DBImpl::BackgroundCallFlush:start", [&](void* /*arg*/) {
int cur = bg_flush_threads.load();
int desired = cur + 1;
if (cur > 0 ||
!bg_flush_threads.compare_exchange_strong(cur, desired)) {
while (!wal_synced.load()) {
// Wait until the other bg flush thread finishes committing WAL sync
// operation to the MANIFEST.
}
}
});
SyncPoint::GetInstance()->SetCallBack(
"DBImpl::FlushMemTableToOutputFile:CommitWal:1",
[&](void* /*arg*/) { wal_synced.store(true); });
Fix a race condition in WAL tracking causing DB open failure (#9715) Summary: There is a race condition if WAL tracking in the MANIFEST is enabled in a database that disables 2PC. The race condition is between two background flush threads trying to install flush results to the MANIFEST. Consider an example database with two column families: "default" (cfd0) and "cf1" (cfd1). Initially, both column families have one mutable (active) memtable whose data backed by 6.log. 1. Trigger a manual flush for "cf1", creating a 7.log 2. Insert another key to "default", and trigger flush for "default", creating 8.log 3. BgFlushThread1 finishes writing 9.sst 4. BgFlushThread2 finishes writing 10.sst ``` Time BgFlushThread1 BgFlushThread2 | mutex_.Lock() | precompute min_wal_to_keep as 6 | mutex_.Unlock() | mutex_.Lock() | precompute min_wal_to_keep as 6 | join MANIFEST write queue and mutex_.Unlock() | write to MANIFEST | mutex_.Lock() | cfd1->log_number = 7 | Signal bg_flush_2 and mutex_.Unlock() | wake up and mutex_.Lock() | cfd0->log_number = 8 | FindObsoleteFiles() with job_context->log_number == 7 | mutex_.Unlock() | PurgeObsoleteFiles() deletes 6.log V ``` As shown in the above, BgFlushThread2 thinks that the min wal to keep is 6.log because "cf1" has unflushed data in 6.log (cf1.log_number=6). Similarly, BgThread1 thinks that min wal to keep is also 6.log because "default" has unflushed data (default.log_number=6). No WAL deletion will be written to MANIFEST because 6 is equal to `versions_->wals_.min_wal_number_to_keep`, due to https://github.com/facebook/rocksdb/blob/7.1.fb/db/memtable_list.cc#L513:L514. The bg flush thread that finishes last will perform file purging. `job_context.log_number` will be evaluated as 7, i.e. the min wal that contains unflushed data, causing 6.log to be deleted. However, MANIFEST thinks 6.log should still exist. If you close the db at this point, you won't be able to re-open it if `track_and_verify_wal_in_manifest` is true. We must handle the case of multiple bg flush threads, and it is difficult for one bg flush thread to know the correct min wal number until the other bg flush threads have finished committing to the manifest and updated the `cfd::log_number`. To fix this issue, we rename an existing variable `min_log_number_to_keep_2pc` to `min_log_number_to_keep`, and use it to track WAL file deletion in non-2pc mode as well. This variable is updated only 1) during recovery with mutex held, or 2) in the MANIFEST write thread. `min_log_number_to_keep` means RocksDB will delete WALs below it, although there may be WALs above it which are also obsolete. Formally, we will have [min_wal_to_keep, max_obsolete_wal]. During recovery, we make sure that only WALs above max_obsolete_wal are checked and added back to `alive_log_files_`. Pull Request resolved: https://github.com/facebook/rocksdb/pull/9715 Test Plan: ``` make check ``` Also ran stress test below (with asan) to make sure it completes successfully. ``` TEST_TMPDIR=/dev/shm/rocksdb OPT=-g ASAN_OPTIONS=disable_coredump=0 \ CRASH_TEST_EXT_ARGS=--compression_type=zstd SKIP_FORMAT_BUCK_CHECKS=1 \ make J=52 -j52 blackbox_asan_crash_test ``` Reviewed By: ltamasi Differential Revision: D34984412 Pulled By: riversand963 fbshipit-source-id: c7b21a8d84751bb55ea79c9f387103d21b231005
3 years ago
// This callback will be called when the first bg flush thread reaches the
// point before entering the MANIFEST write queue after flushing the SST
// file.
// The purpose of the sync points here is to ensure both bg flush threads
// finish computing `min_wal_number_to_keep` before any of them updates the
// `log_number` for the column family that's being flushed.
SyncPoint::GetInstance()->SetCallBack(
"MemTableList::TryInstallMemtableFlushResults:AfterComputeMinWalToKeep",
[&](void* /*arg*/) {
dbfull()->mutex()->AssertHeld();
if (!called) {
// We are the first bg flush thread in the MANIFEST write queue.
// We set up the dependency between sync points for two threads that
// will be executing the same code.
// For the interleaving of events, see
// https://github.com/facebook/rocksdb/pull/9715.
// bg flush thread1 will release the db mutex while in the MANIFEST
// write queue. In the meantime, bg flush thread2 locks db mutex and
// computes the min_wal_number_to_keep (before thread1 writes to
// MANIFEST thus before cf1->log_number is updated). Bg thread2 joins
// the MANIFEST write queue afterwards and bg flush thread1 proceeds
// with writing to MANIFEST.
called = true;
SyncPoint::GetInstance()->LoadDependency({
{"VersionSet::LogAndApply:WriteManifestStart",
"DBWALTest::RaceInstallFlushResultsWithWalObsoletion:BgFlush2"},
{"DBWALTest::RaceInstallFlushResultsWithWalObsoletion:BgFlush2",
"VersionSet::LogAndApply:WriteManifest"},
});
} else {
// The other bg flush thread has already been in the MANIFEST write
// queue, and we are after.
TEST_SYNC_POINT(
"DBWALTest::RaceInstallFlushResultsWithWalObsoletion:BgFlush2");
}
});
SyncPoint::GetInstance()->EnableProcessing();
ASSERT_OK(dbfull()->ContinueBackgroundWork());
ASSERT_OK(dbfull()->TEST_WaitForFlushMemTable(handles_[0]));
ASSERT_OK(dbfull()->TEST_WaitForFlushMemTable(handles_[1]));
ASSERT_TRUE(called);
Close();
SyncPoint::GetInstance()->DisableProcessing();
SyncPoint::GetInstance()->ClearAllCallBacks();
DB* db1 = nullptr;
Status s = DB::OpenForReadOnly(options, dbname_, &db1);
ASSERT_OK(s);
assert(db1);
delete db1;
}
Fix missing WAL in new manifest by rolling over the WAL deletion record from prev manifest (#10892) Summary: **Context** `Options::track_and_verify_wals_in_manifest = true` verifies each of the WALs tracked in manifest indeed presents in the WAL folder. If not, a corruption "Missing WAL with log number" will be thrown. `DB::SyncWAL()` called at a specific timing (i.e, at the `TEST_SYNC_POINT("FindObsoleteFiles::PostMutexUnlock")`) can record in a new manifest the WAL addition of a WAL file that already had a WAL deletion recorded in the previous manifest. And the WAL deletion record is not rollover-ed to the new manifest. So the new manifest creates the illusion of such WAL never gets deleted and should presents at db re/open. - Such WAL deletion record can be caused by flushing the memtable associated with that WAL and such WAL deletion can actually happen in` PurgeObsoleteFiles()`. As a consequence, upon `DB::Reopen()`, this WAL file can be deleted while manifest still has its WAL addition record , which causes a false alarm of corruption "Missing WAL with log number" to be thrown. **Summary** This PR fixes this false alarm by rolling over the WAL deletion record from prev manifest to the new manifest by adding the WAL deletion record to the new manifest. **Test** - Make check - Added new unit test `TEST_F(DBWALTest, FixSyncWalOnObseletedWalWithNewManifestCausingMissingWAL)` that failed before the fix and passed after - [Ongoing]CI stress test + aggressive value as in https://github.com/facebook/rocksdb/pull/10761 , which is how this false alarm was first surfaced, to confirm such false alarm disappears - [Ongoing]Regular CI stress test to confirm such fix didn't harm anything Pull Request resolved: https://github.com/facebook/rocksdb/pull/10892 Reviewed By: ajkr Differential Revision: D40778965 Pulled By: hx235 fbshipit-source-id: a512364bfdeb0b1a55c171890e60d856c528f37f
2 years ago
TEST_F(DBWALTest, FixSyncWalOnObseletedWalWithNewManifestCausingMissingWAL) {
Options options = CurrentOptions();
// Small size to force manifest creation
options.max_manifest_file_size = 1;
Fix missing WAL in new manifest by rolling over the WAL deletion record from prev manifest (#10892) Summary: **Context** `Options::track_and_verify_wals_in_manifest = true` verifies each of the WALs tracked in manifest indeed presents in the WAL folder. If not, a corruption "Missing WAL with log number" will be thrown. `DB::SyncWAL()` called at a specific timing (i.e, at the `TEST_SYNC_POINT("FindObsoleteFiles::PostMutexUnlock")`) can record in a new manifest the WAL addition of a WAL file that already had a WAL deletion recorded in the previous manifest. And the WAL deletion record is not rollover-ed to the new manifest. So the new manifest creates the illusion of such WAL never gets deleted and should presents at db re/open. - Such WAL deletion record can be caused by flushing the memtable associated with that WAL and such WAL deletion can actually happen in` PurgeObsoleteFiles()`. As a consequence, upon `DB::Reopen()`, this WAL file can be deleted while manifest still has its WAL addition record , which causes a false alarm of corruption "Missing WAL with log number" to be thrown. **Summary** This PR fixes this false alarm by rolling over the WAL deletion record from prev manifest to the new manifest by adding the WAL deletion record to the new manifest. **Test** - Make check - Added new unit test `TEST_F(DBWALTest, FixSyncWalOnObseletedWalWithNewManifestCausingMissingWAL)` that failed before the fix and passed after - [Ongoing]CI stress test + aggressive value as in https://github.com/facebook/rocksdb/pull/10761 , which is how this false alarm was first surfaced, to confirm such false alarm disappears - [Ongoing]Regular CI stress test to confirm such fix didn't harm anything Pull Request resolved: https://github.com/facebook/rocksdb/pull/10892 Reviewed By: ajkr Differential Revision: D40778965 Pulled By: hx235 fbshipit-source-id: a512364bfdeb0b1a55c171890e60d856c528f37f
2 years ago
options.track_and_verify_wals_in_manifest = true;
DestroyAndReopen(options);
// Accumulate memtable m1 and create the 1st wal (i.e, 4.log)
ASSERT_OK(Put(Key(1), ""));
ASSERT_OK(Put(Key(2), ""));
ASSERT_OK(Put(Key(3), ""));
const std::string wal_file_path = db_->GetName() + "/000004.log";
// Coerce the following sequence of events:
// (1) Flush() marks 4.log to be obsoleted, 8.log to be the latest (i.e,
// active) log and release the lock
// (2) SyncWAL() proceeds with the lock. It
// creates a new manifest and syncs all the inactive wals before the latest
// (i.e, active log), which is 4.log. Note that SyncWAL() is not aware of the
// fact that 4.log has marked as to be obseleted. Such wal
// sync will then add a WAL addition record of 4.log to the new manifest
// without any special treatment. Prior to the fix, there is no WAL deletion
// record to offset it. (3) BackgroundFlush() will eventually purge 4.log.
bool wal_synced = false;
Fix missing WAL in new manifest by rolling over the WAL deletion record from prev manifest (#10892) Summary: **Context** `Options::track_and_verify_wals_in_manifest = true` verifies each of the WALs tracked in manifest indeed presents in the WAL folder. If not, a corruption "Missing WAL with log number" will be thrown. `DB::SyncWAL()` called at a specific timing (i.e, at the `TEST_SYNC_POINT("FindObsoleteFiles::PostMutexUnlock")`) can record in a new manifest the WAL addition of a WAL file that already had a WAL deletion recorded in the previous manifest. And the WAL deletion record is not rollover-ed to the new manifest. So the new manifest creates the illusion of such WAL never gets deleted and should presents at db re/open. - Such WAL deletion record can be caused by flushing the memtable associated with that WAL and such WAL deletion can actually happen in` PurgeObsoleteFiles()`. As a consequence, upon `DB::Reopen()`, this WAL file can be deleted while manifest still has its WAL addition record , which causes a false alarm of corruption "Missing WAL with log number" to be thrown. **Summary** This PR fixes this false alarm by rolling over the WAL deletion record from prev manifest to the new manifest by adding the WAL deletion record to the new manifest. **Test** - Make check - Added new unit test `TEST_F(DBWALTest, FixSyncWalOnObseletedWalWithNewManifestCausingMissingWAL)` that failed before the fix and passed after - [Ongoing]CI stress test + aggressive value as in https://github.com/facebook/rocksdb/pull/10761 , which is how this false alarm was first surfaced, to confirm such false alarm disappears - [Ongoing]Regular CI stress test to confirm such fix didn't harm anything Pull Request resolved: https://github.com/facebook/rocksdb/pull/10892 Reviewed By: ajkr Differential Revision: D40778965 Pulled By: hx235 fbshipit-source-id: a512364bfdeb0b1a55c171890e60d856c528f37f
2 years ago
SyncPoint::GetInstance()->SetCallBack(
"FindObsoleteFiles::PostMutexUnlock", [&](void*) {
ASSERT_OK(env_->FileExists(wal_file_path));
uint64_t pre_sync_wal_manifest_no =
dbfull()->TEST_Current_Manifest_FileNo();
Fix missing WAL in new manifest by rolling over the WAL deletion record from prev manifest (#10892) Summary: **Context** `Options::track_and_verify_wals_in_manifest = true` verifies each of the WALs tracked in manifest indeed presents in the WAL folder. If not, a corruption "Missing WAL with log number" will be thrown. `DB::SyncWAL()` called at a specific timing (i.e, at the `TEST_SYNC_POINT("FindObsoleteFiles::PostMutexUnlock")`) can record in a new manifest the WAL addition of a WAL file that already had a WAL deletion recorded in the previous manifest. And the WAL deletion record is not rollover-ed to the new manifest. So the new manifest creates the illusion of such WAL never gets deleted and should presents at db re/open. - Such WAL deletion record can be caused by flushing the memtable associated with that WAL and such WAL deletion can actually happen in` PurgeObsoleteFiles()`. As a consequence, upon `DB::Reopen()`, this WAL file can be deleted while manifest still has its WAL addition record , which causes a false alarm of corruption "Missing WAL with log number" to be thrown. **Summary** This PR fixes this false alarm by rolling over the WAL deletion record from prev manifest to the new manifest by adding the WAL deletion record to the new manifest. **Test** - Make check - Added new unit test `TEST_F(DBWALTest, FixSyncWalOnObseletedWalWithNewManifestCausingMissingWAL)` that failed before the fix and passed after - [Ongoing]CI stress test + aggressive value as in https://github.com/facebook/rocksdb/pull/10761 , which is how this false alarm was first surfaced, to confirm such false alarm disappears - [Ongoing]Regular CI stress test to confirm such fix didn't harm anything Pull Request resolved: https://github.com/facebook/rocksdb/pull/10892 Reviewed By: ajkr Differential Revision: D40778965 Pulled By: hx235 fbshipit-source-id: a512364bfdeb0b1a55c171890e60d856c528f37f
2 years ago
ASSERT_OK(db_->SyncWAL());
uint64_t post_sync_wal_manifest_no =
dbfull()->TEST_Current_Manifest_FileNo();
bool new_manifest_created =
post_sync_wal_manifest_no == pre_sync_wal_manifest_no + 1;
ASSERT_TRUE(new_manifest_created);
wal_synced = true;
Fix missing WAL in new manifest by rolling over the WAL deletion record from prev manifest (#10892) Summary: **Context** `Options::track_and_verify_wals_in_manifest = true` verifies each of the WALs tracked in manifest indeed presents in the WAL folder. If not, a corruption "Missing WAL with log number" will be thrown. `DB::SyncWAL()` called at a specific timing (i.e, at the `TEST_SYNC_POINT("FindObsoleteFiles::PostMutexUnlock")`) can record in a new manifest the WAL addition of a WAL file that already had a WAL deletion recorded in the previous manifest. And the WAL deletion record is not rollover-ed to the new manifest. So the new manifest creates the illusion of such WAL never gets deleted and should presents at db re/open. - Such WAL deletion record can be caused by flushing the memtable associated with that WAL and such WAL deletion can actually happen in` PurgeObsoleteFiles()`. As a consequence, upon `DB::Reopen()`, this WAL file can be deleted while manifest still has its WAL addition record , which causes a false alarm of corruption "Missing WAL with log number" to be thrown. **Summary** This PR fixes this false alarm by rolling over the WAL deletion record from prev manifest to the new manifest by adding the WAL deletion record to the new manifest. **Test** - Make check - Added new unit test `TEST_F(DBWALTest, FixSyncWalOnObseletedWalWithNewManifestCausingMissingWAL)` that failed before the fix and passed after - [Ongoing]CI stress test + aggressive value as in https://github.com/facebook/rocksdb/pull/10761 , which is how this false alarm was first surfaced, to confirm such false alarm disappears - [Ongoing]Regular CI stress test to confirm such fix didn't harm anything Pull Request resolved: https://github.com/facebook/rocksdb/pull/10892 Reviewed By: ajkr Differential Revision: D40778965 Pulled By: hx235 fbshipit-source-id: a512364bfdeb0b1a55c171890e60d856c528f37f
2 years ago
});
SyncPoint::GetInstance()->EnableProcessing();
ASSERT_OK(Flush());
ASSERT_OK(dbfull()->TEST_WaitForBackgroundWork());
Fix missing WAL in new manifest by rolling over the WAL deletion record from prev manifest (#10892) Summary: **Context** `Options::track_and_verify_wals_in_manifest = true` verifies each of the WALs tracked in manifest indeed presents in the WAL folder. If not, a corruption "Missing WAL with log number" will be thrown. `DB::SyncWAL()` called at a specific timing (i.e, at the `TEST_SYNC_POINT("FindObsoleteFiles::PostMutexUnlock")`) can record in a new manifest the WAL addition of a WAL file that already had a WAL deletion recorded in the previous manifest. And the WAL deletion record is not rollover-ed to the new manifest. So the new manifest creates the illusion of such WAL never gets deleted and should presents at db re/open. - Such WAL deletion record can be caused by flushing the memtable associated with that WAL and such WAL deletion can actually happen in` PurgeObsoleteFiles()`. As a consequence, upon `DB::Reopen()`, this WAL file can be deleted while manifest still has its WAL addition record , which causes a false alarm of corruption "Missing WAL with log number" to be thrown. **Summary** This PR fixes this false alarm by rolling over the WAL deletion record from prev manifest to the new manifest by adding the WAL deletion record to the new manifest. **Test** - Make check - Added new unit test `TEST_F(DBWALTest, FixSyncWalOnObseletedWalWithNewManifestCausingMissingWAL)` that failed before the fix and passed after - [Ongoing]CI stress test + aggressive value as in https://github.com/facebook/rocksdb/pull/10761 , which is how this false alarm was first surfaced, to confirm such false alarm disappears - [Ongoing]Regular CI stress test to confirm such fix didn't harm anything Pull Request resolved: https://github.com/facebook/rocksdb/pull/10892 Reviewed By: ajkr Differential Revision: D40778965 Pulled By: hx235 fbshipit-source-id: a512364bfdeb0b1a55c171890e60d856c528f37f
2 years ago
ASSERT_TRUE(wal_synced);
Fix missing WAL in new manifest by rolling over the WAL deletion record from prev manifest (#10892) Summary: **Context** `Options::track_and_verify_wals_in_manifest = true` verifies each of the WALs tracked in manifest indeed presents in the WAL folder. If not, a corruption "Missing WAL with log number" will be thrown. `DB::SyncWAL()` called at a specific timing (i.e, at the `TEST_SYNC_POINT("FindObsoleteFiles::PostMutexUnlock")`) can record in a new manifest the WAL addition of a WAL file that already had a WAL deletion recorded in the previous manifest. And the WAL deletion record is not rollover-ed to the new manifest. So the new manifest creates the illusion of such WAL never gets deleted and should presents at db re/open. - Such WAL deletion record can be caused by flushing the memtable associated with that WAL and such WAL deletion can actually happen in` PurgeObsoleteFiles()`. As a consequence, upon `DB::Reopen()`, this WAL file can be deleted while manifest still has its WAL addition record , which causes a false alarm of corruption "Missing WAL with log number" to be thrown. **Summary** This PR fixes this false alarm by rolling over the WAL deletion record from prev manifest to the new manifest by adding the WAL deletion record to the new manifest. **Test** - Make check - Added new unit test `TEST_F(DBWALTest, FixSyncWalOnObseletedWalWithNewManifestCausingMissingWAL)` that failed before the fix and passed after - [Ongoing]CI stress test + aggressive value as in https://github.com/facebook/rocksdb/pull/10761 , which is how this false alarm was first surfaced, to confirm such false alarm disappears - [Ongoing]Regular CI stress test to confirm such fix didn't harm anything Pull Request resolved: https://github.com/facebook/rocksdb/pull/10892 Reviewed By: ajkr Differential Revision: D40778965 Pulled By: hx235 fbshipit-source-id: a512364bfdeb0b1a55c171890e60d856c528f37f
2 years ago
// BackgroundFlush() purged 4.log
// because the memtable associated with the WAL was flushed and new WAL was
// created (i.e, 8.log)
ASSERT_TRUE(env_->FileExists(wal_file_path).IsNotFound());
SyncPoint::GetInstance()->ClearAllCallBacks();
SyncPoint::GetInstance()->DisableProcessing();
// To verify the corruption of "Missing WAL with log number: 4" under
// `options.track_and_verify_wals_in_manifest = true` is fixed.
//
// Before the fix, `db_->SyncWAL()` will sync and record WAL addtion of the
// obseleted WAL 4.log in a new manifest without any special treament.
// This will result in missing-wal corruption in DB::Reopen().
Status s = TryReopen(options);
EXPECT_OK(s);
}
// Test scope:
// - We expect to open data store under all circumstances
// - We expect only data upto the point where the first error was encountered
TEST_P(DBWALTestWithParams, kPointInTimeRecovery) {
const int maxkeys =
RecoveryTestHelper::kWALFilesCount * RecoveryTestHelper::kKeysPerWALFile;
bool trunc = std::get<0>(GetParam()); // Corruption style
// Corruption offset position
int corrupt_offset = std::get<1>(GetParam());
int wal_file_id = std::get<2>(GetParam()); // WAL file
// WAL compression type
CompressionType compression_type = std::get<3>(GetParam());
// Fill data for testing
Options options = CurrentOptions();
options.wal_compression = compression_type;
const size_t row_count = RecoveryTestHelper::FillData(this, &options);
// Corrupt the wal
// The offset here was 0.3 which cuts off right at the end of a
// valid fragment after wal zstd compression checksum is enabled,
// so changed the value to 0.33.
RecoveryTestHelper::CorruptWAL(this, options, corrupt_offset * .33,
/*len%=*/.1, wal_file_id, trunc);
// Verify
options.wal_recovery_mode = WALRecoveryMode::kPointInTimeRecovery;
options.create_if_missing = false;
ASSERT_OK(TryReopen(options));
// Probe data for invariants
size_t recovered_row_count = RecoveryTestHelper::GetData(this);
ASSERT_LT(recovered_row_count, row_count);
// Verify a prefix of keys were recovered. But not in the case of full WAL
// truncation, because we have no way to know there was a corruption when
// truncation happened on record boundaries (preventing recovery holes in
// that case requires using `track_and_verify_wals_in_manifest`).
if (!trunc || corrupt_offset != 0) {
bool expect_data = true;
for (size_t k = 0; k < maxkeys; ++k) {
bool found = Get("key" + std::to_string(k)) != "NOT_FOUND";
if (expect_data && !found) {
expect_data = false;
}
ASSERT_EQ(found, expect_data);
}
}
const size_t min = RecoveryTestHelper::kKeysPerWALFile *
(wal_file_id - RecoveryTestHelper::kWALFileOffset);
ASSERT_GE(recovered_row_count, min);
if (!trunc && corrupt_offset != 0) {
const size_t max = RecoveryTestHelper::kKeysPerWALFile *
(wal_file_id - RecoveryTestHelper::kWALFileOffset + 1);
ASSERT_LE(recovered_row_count, max);
}
}
// Test scope:
// - We expect to open the data store under all scenarios
// - We expect to have recovered records past the corruption zone
TEST_P(DBWALTestWithParams, kSkipAnyCorruptedRecords) {
bool trunc = std::get<0>(GetParam()); // Corruption style
// Corruption offset position
int corrupt_offset = std::get<1>(GetParam());
int wal_file_id = std::get<2>(GetParam()); // WAL file
// WAL compression type
CompressionType compression_type = std::get<3>(GetParam());
// Fill data for testing
Options options = CurrentOptions();
options.wal_compression = compression_type;
const size_t row_count = RecoveryTestHelper::FillData(this, &options);
// Corrupt the WAL
RecoveryTestHelper::CorruptWAL(this, options, corrupt_offset * .3,
/*len%=*/.1, wal_file_id, trunc);
// Verify behavior
options.wal_recovery_mode = WALRecoveryMode::kSkipAnyCorruptedRecords;
options.create_if_missing = false;
ASSERT_OK(TryReopen(options));
// Probe data for invariants
size_t recovered_row_count = RecoveryTestHelper::GetData(this);
ASSERT_LT(recovered_row_count, row_count);
if (!trunc) {
ASSERT_TRUE(corrupt_offset != 0 || recovered_row_count > 0);
}
}
TEST_F(DBWALTest, AvoidFlushDuringRecovery) {
Options options = CurrentOptions();
options.disable_auto_compactions = true;
options.avoid_flush_during_recovery = false;
// Test with flush after recovery.
Reopen(options);
ASSERT_OK(Put("foo", "v1"));
ASSERT_OK(Put("bar", "v2"));
ASSERT_OK(Flush());
ASSERT_OK(Put("foo", "v3"));
ASSERT_OK(Put("bar", "v4"));
ASSERT_EQ(1, TotalTableFiles());
// Reopen DB. Check if WAL logs flushed.
Reopen(options);
ASSERT_EQ("v3", Get("foo"));
ASSERT_EQ("v4", Get("bar"));
ASSERT_EQ(2, TotalTableFiles());
// Test without flush after recovery.
options.avoid_flush_during_recovery = true;
DestroyAndReopen(options);
ASSERT_OK(Put("foo", "v5"));
ASSERT_OK(Put("bar", "v6"));
ASSERT_OK(Flush());
ASSERT_OK(Put("foo", "v7"));
ASSERT_OK(Put("bar", "v8"));
ASSERT_EQ(1, TotalTableFiles());
// Reopen DB. WAL logs should not be flushed this time.
Reopen(options);
ASSERT_EQ("v7", Get("foo"));
ASSERT_EQ("v8", Get("bar"));
ASSERT_EQ(1, TotalTableFiles());
// Force flush with allow_2pc.
options.avoid_flush_during_recovery = true;
options.allow_2pc = true;
ASSERT_OK(Put("foo", "v9"));
ASSERT_OK(Put("bar", "v10"));
ASSERT_OK(Flush());
ASSERT_OK(Put("foo", "v11"));
ASSERT_OK(Put("bar", "v12"));
Reopen(options);
ASSERT_EQ("v11", Get("foo"));
ASSERT_EQ("v12", Get("bar"));
ASSERT_EQ(3, TotalTableFiles());
}
TEST_F(DBWALTest, WalCleanupAfterAvoidFlushDuringRecovery) {
// Verifies WAL files that were present during recovery, but not flushed due
// to avoid_flush_during_recovery, will be considered for deletion at a later
// stage. We check at least one such file is deleted during Flush().
Options options = CurrentOptions();
options.disable_auto_compactions = true;
options.avoid_flush_during_recovery = true;
Reopen(options);
ASSERT_OK(Put("foo", "v1"));
Reopen(options);
for (int i = 0; i < 2; ++i) {
if (i > 0) {
// Flush() triggers deletion of obsolete tracked files
ASSERT_OK(Flush());
}
VectorLogPtr log_files;
ASSERT_OK(dbfull()->GetSortedWalFiles(log_files));
if (i == 0) {
ASSERT_GT(log_files.size(), 0);
} else {
ASSERT_EQ(0, log_files.size());
}
}
}
TEST_F(DBWALTest, RecoverWithoutFlush) {
Options options = CurrentOptions();
options.avoid_flush_during_recovery = true;
options.create_if_missing = false;
options.disable_auto_compactions = true;
options.write_buffer_size = 64 * 1024 * 1024;
size_t count = RecoveryTestHelper::FillData(this, &options);
auto validateData = [this, count]() {
for (size_t i = 0; i < count; i++) {
ASSERT_NE(Get("key" + std::to_string(i)), "NOT_FOUND");
}
};
Reopen(options);
validateData();
// Insert some data without flush
ASSERT_OK(Put("foo", "foo_v1"));
ASSERT_OK(Put("bar", "bar_v1"));
Reopen(options);
validateData();
ASSERT_EQ(Get("foo"), "foo_v1");
ASSERT_EQ(Get("bar"), "bar_v1");
// Insert again and reopen
ASSERT_OK(Put("foo", "foo_v2"));
ASSERT_OK(Put("bar", "bar_v2"));
Reopen(options);
validateData();
ASSERT_EQ(Get("foo"), "foo_v2");
ASSERT_EQ(Get("bar"), "bar_v2");
// manual flush and insert again
ASSERT_OK(Flush());
ASSERT_EQ(Get("foo"), "foo_v2");
ASSERT_EQ(Get("bar"), "bar_v2");
ASSERT_OK(Put("foo", "foo_v3"));
ASSERT_OK(Put("bar", "bar_v3"));
Reopen(options);
validateData();
ASSERT_EQ(Get("foo"), "foo_v3");
ASSERT_EQ(Get("bar"), "bar_v3");
}
TEST_F(DBWALTest, RecoverWithoutFlushMultipleCF) {
const std::string kSmallValue = "v";
const std::string kLargeValue = DummyString(1024);
Options options = CurrentOptions();
options.avoid_flush_during_recovery = true;
options.create_if_missing = false;
options.disable_auto_compactions = true;
auto countWalFiles = [this]() {
VectorLogPtr log_files;
if (!dbfull()->GetSortedWalFiles(log_files).ok()) {
return size_t{0};
}
return log_files.size();
};
// Create DB with multiple column families and multiple log files.
CreateAndReopenWithCF({"one", "two"}, options);
ASSERT_OK(Put(0, "key1", kSmallValue));
ASSERT_OK(Put(1, "key2", kLargeValue));
ASSERT_OK(Flush(1));
ASSERT_EQ(1, countWalFiles());
ASSERT_OK(Put(0, "key3", kSmallValue));
ASSERT_OK(Put(2, "key4", kLargeValue));
ASSERT_OK(Flush(2));
ASSERT_EQ(2, countWalFiles());
// Reopen, insert and flush.
options.db_write_buffer_size = 64 * 1024 * 1024;
ReopenWithColumnFamilies({"default", "one", "two"}, options);
ASSERT_EQ(Get(0, "key1"), kSmallValue);
ASSERT_EQ(Get(1, "key2"), kLargeValue);
ASSERT_EQ(Get(0, "key3"), kSmallValue);
ASSERT_EQ(Get(2, "key4"), kLargeValue);
// Insert more data.
ASSERT_OK(Put(0, "key5", kLargeValue));
ASSERT_OK(Put(1, "key6", kLargeValue));
ASSERT_EQ(3, countWalFiles());
ASSERT_OK(Flush(1));
ASSERT_OK(Put(2, "key7", kLargeValue));
ASSERT_OK(dbfull()->FlushWAL(false));
ASSERT_EQ(4, countWalFiles());
// Reopen twice and validate.
for (int i = 0; i < 2; i++) {
ReopenWithColumnFamilies({"default", "one", "two"}, options);
ASSERT_EQ(Get(0, "key1"), kSmallValue);
ASSERT_EQ(Get(1, "key2"), kLargeValue);
ASSERT_EQ(Get(0, "key3"), kSmallValue);
ASSERT_EQ(Get(2, "key4"), kLargeValue);
ASSERT_EQ(Get(0, "key5"), kLargeValue);
ASSERT_EQ(Get(1, "key6"), kLargeValue);
ASSERT_EQ(Get(2, "key7"), kLargeValue);
ASSERT_EQ(4, countWalFiles());
}
}
// In this test we are trying to do the following:
// 1. Create a DB with corrupted WAL log;
// 2. Open with avoid_flush_during_recovery = true;
// 3. Append more data without flushing, which creates new WAL log.
// 4. Open again. See if it can correctly handle previous corruption.
TEST_P(DBWALTestWithParamsVaryingRecoveryMode,
RecoverFromCorruptedWALWithoutFlush) {
const int kAppendKeys = 100;
Options options = CurrentOptions();
options.avoid_flush_during_recovery = true;
options.create_if_missing = false;
options.disable_auto_compactions = true;
options.write_buffer_size = 64 * 1024 * 1024;
auto getAll = [this]() {
std::vector<std::pair<std::string, std::string>> data;
ReadOptions ropt;
Iterator* iter = dbfull()->NewIterator(ropt);
for (iter->SeekToFirst(); iter->Valid(); iter->Next()) {
data.push_back(
std::make_pair(iter->key().ToString(), iter->value().ToString()));
}
delete iter;
return data;
};
bool trunc = std::get<0>(GetParam()); // Corruption style
// Corruption offset position
int corrupt_offset = std::get<1>(GetParam());
int wal_file_id = std::get<2>(GetParam()); // WAL file
WALRecoveryMode recovery_mode = std::get<3>(GetParam());
// WAL compression type
CompressionType compression_type = std::get<4>(GetParam());
options.wal_recovery_mode = recovery_mode;
options.wal_compression = compression_type;
// Create corrupted WAL
RecoveryTestHelper::FillData(this, &options);
RecoveryTestHelper::CorruptWAL(this, options, corrupt_offset * .3,
/*len%=*/.1, wal_file_id, trunc);
// Skip the test if DB won't open.
if (!TryReopen(options).ok()) {
ASSERT_TRUE(options.wal_recovery_mode ==
WALRecoveryMode::kAbsoluteConsistency ||
(!trunc && options.wal_recovery_mode ==
WALRecoveryMode::kTolerateCorruptedTailRecords));
return;
}
ASSERT_OK(TryReopen(options));
// Append some more data.
for (int k = 0; k < kAppendKeys; k++) {
std::string key = "extra_key" + std::to_string(k);
std::string value = DummyString(RecoveryTestHelper::kValueSize);
ASSERT_OK(Put(key, value));
}
// Save data for comparison.
auto data = getAll();
// Reopen. Verify data.
ASSERT_OK(TryReopen(options));
auto actual_data = getAll();
ASSERT_EQ(data, actual_data);
}
// Tests that total log size is recovered if we set
// avoid_flush_during_recovery=true.
// Flush should trigger if max_total_wal_size is reached.
TEST_F(DBWALTest, RestoreTotalLogSizeAfterRecoverWithoutFlush) {
auto test_listener = std::make_shared<FlushCounterListener>();
test_listener->expected_flush_reason = FlushReason::kWalFull;
constexpr size_t kKB = 1024;
constexpr size_t kMB = 1024 * 1024;
Options options = CurrentOptions();
options.avoid_flush_during_recovery = true;
options.max_total_wal_size = 1 * kMB;
options.listeners.push_back(test_listener);
// Have to open DB in multi-CF mode to trigger flush when
// max_total_wal_size is reached.
CreateAndReopenWithCF({"one"}, options);
// Write some keys and we will end up with one log file which is slightly
// smaller than 1MB.
std::string value_100k(100 * kKB, 'v');
std::string value_300k(300 * kKB, 'v');
ASSERT_OK(Put(0, "foo", "v1"));
for (int i = 0; i < 9; i++) {
ASSERT_OK(Put(1, "key" + std::to_string(i), value_100k));
}
// Get log files before reopen.
VectorLogPtr log_files_before;
ASSERT_OK(dbfull()->GetSortedWalFiles(log_files_before));
ASSERT_EQ(1, log_files_before.size());
uint64_t log_size_before = log_files_before[0]->SizeFileBytes();
ASSERT_GT(log_size_before, 900 * kKB);
ASSERT_LT(log_size_before, 1 * kMB);
ReopenWithColumnFamilies({"default", "one"}, options);
// Write one more value to make log larger than 1MB.
ASSERT_OK(Put(1, "bar", value_300k));
// Get log files again. A new log file will be opened.
VectorLogPtr log_files_after_reopen;
ASSERT_OK(dbfull()->GetSortedWalFiles(log_files_after_reopen));
ASSERT_EQ(2, log_files_after_reopen.size());
ASSERT_EQ(log_files_before[0]->LogNumber(),
log_files_after_reopen[0]->LogNumber());
ASSERT_GT(log_files_after_reopen[0]->SizeFileBytes() +
log_files_after_reopen[1]->SizeFileBytes(),
1 * kMB);
// Write one more key to trigger flush.
ASSERT_OK(Put(0, "foo", "v2"));
for (auto* h : handles_) {
ASSERT_OK(dbfull()->TEST_WaitForFlushMemTable(h));
}
// Flushed two column families.
ASSERT_EQ(2, test_listener->count.load());
}
#if defined(ROCKSDB_PLATFORM_POSIX)
#if defined(ROCKSDB_FALLOCATE_PRESENT)
// Tests that we will truncate the preallocated space of the last log from
// previous.
TEST_F(DBWALTest, TruncateLastLogAfterRecoverWithoutFlush) {
constexpr size_t kKB = 1024;
Options options = CurrentOptions();
options.env = env_;
options.avoid_flush_during_recovery = true;
if (mem_env_) {
ROCKSDB_GTEST_SKIP("Test requires non-mem environment");
Fix many tests to run with MEM_ENV and ENCRYPTED_ENV; Introduce a MemoryFileSystem class (#7566) Summary: This PR does a few things: 1. The MockFileSystem class was split out from the MockEnv. This change would theoretically allow a MockFileSystem to be used by other Environments as well (if we created a means of constructing one). The MockFileSystem implements a FileSystem in its entirety and does not rely on any Wrapper implementation. 2. Make the RocksDB test suite work when MOCK_ENV=1 and ENCRYPTED_ENV=1 are set. To accomplish this, a few things were needed: - The tests that tried to use the "wrong" environment (Env::Default() instead of env_) were updated - The MockFileSystem was changed to support the features it was missing or mishandled (such as recursively deleting files in a directory or supporting renaming of a directory). 3. Updated the test framework to have a ROCKSDB_GTEST_SKIP macro. This can be used to flag tests that are skipped. Currently, this defaults to doing nothing (marks the test as SUCCESS) but will mark the tests as SKIPPED when RocksDB is upgraded to a version of gtest that supports this (gtest-1.10). I have run a full "make check" with MEM_ENV, ENCRYPTED_ENV, both, and neither under both MacOS and RedHat. A few tests were disabled/skipped for the MEM/ENCRYPTED cases. The error_handler_fs_test fails/hangs for MEM_ENV (presumably a timing problem) and I will introduce another PR/issue to track that problem. (I will also push a change to disable those tests soon). There is one more test in DBTest2 that also fails which I need to investigate or skip before this PR is merged. Theoretically, this PR should also allow the test suite to run against an Env loaded from the registry, though I do not have one to try it with currently. Finally, once this is accepted, it would be nice if there was a CircleCI job to run these tests on a checkin so this effort does not become stale. I do not know how to do that, so if someone could write that job, it would be appreciated :) Pull Request resolved: https://github.com/facebook/rocksdb/pull/7566 Reviewed By: zhichao-cao Differential Revision: D24408980 Pulled By: jay-zhuang fbshipit-source-id: 911b1554a4d0da06fd51feca0c090a4abdcb4a5f
4 years ago
return;
}
if (!IsFallocateSupported()) {
return;
}
DestroyAndReopen(options);
size_t preallocated_size =
dbfull()->TEST_GetWalPreallocateBlockSize(options.write_buffer_size);
ASSERT_OK(Put("foo", "v1"));
VectorLogPtr log_files_before;
ASSERT_OK(dbfull()->GetSortedWalFiles(log_files_before));
ASSERT_EQ(1, log_files_before.size());
auto& file_before = log_files_before[0];
ASSERT_LT(file_before->SizeFileBytes(), 1 * kKB);
// The log file has preallocated space.
ASSERT_GE(GetAllocatedFileSize(dbname_ + file_before->PathName()),
preallocated_size);
Reopen(options);
VectorLogPtr log_files_after;
ASSERT_OK(dbfull()->GetSortedWalFiles(log_files_after));
ASSERT_EQ(1, log_files_after.size());
ASSERT_LT(log_files_after[0]->SizeFileBytes(), 1 * kKB);
// The preallocated space should be truncated.
ASSERT_LT(GetAllocatedFileSize(dbname_ + file_before->PathName()),
preallocated_size);
}
// Tests that we will truncate the preallocated space of the last log from
// previous.
TEST_F(DBWALTest, TruncateLastLogAfterRecoverWithFlush) {
constexpr size_t kKB = 1024;
Options options = CurrentOptions();
options.env = env_;
options.avoid_flush_during_recovery = false;
options.avoid_flush_during_shutdown = true;
if (mem_env_) {
ROCKSDB_GTEST_SKIP("Test requires non-mem environment");
return;
}
if (!IsFallocateSupported()) {
return;
}
DestroyAndReopen(options);
size_t preallocated_size =
dbfull()->TEST_GetWalPreallocateBlockSize(options.write_buffer_size);
ASSERT_OK(Put("foo", "v1"));
VectorLogPtr log_files_before;
ASSERT_OK(dbfull()->GetSortedWalFiles(log_files_before));
ASSERT_EQ(1, log_files_before.size());
auto& file_before = log_files_before[0];
ASSERT_LT(file_before->SizeFileBytes(), 1 * kKB);
ASSERT_GE(GetAllocatedFileSize(dbname_ + file_before->PathName()),
preallocated_size);
// The log file has preallocated space.
Close();
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->LoadDependency(
{{"DBImpl::PurgeObsoleteFiles:Begin",
"DBWALTest::TruncateLastLogAfterRecoverWithFlush:AfterRecover"},
{"DBWALTest::TruncateLastLogAfterRecoverWithFlush:AfterTruncate",
"DBImpl::DeleteObsoleteFileImpl::BeforeDeletion"}});
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->EnableProcessing();
port::Thread reopen_thread([&]() { Reopen(options); });
TEST_SYNC_POINT(
"DBWALTest::TruncateLastLogAfterRecoverWithFlush:AfterRecover");
// After the flush during Open, the log file should get deleted. However,
// if the process is in a crash loop, the log file may not get
// deleted and thte preallocated space will keep accumulating. So we need
// to ensure it gets trtuncated.
EXPECT_LT(GetAllocatedFileSize(dbname_ + file_before->PathName()),
preallocated_size);
TEST_SYNC_POINT(
"DBWALTest::TruncateLastLogAfterRecoverWithFlush:AfterTruncate");
reopen_thread.join();
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->DisableProcessing();
}
TEST_F(DBWALTest, TruncateLastLogAfterRecoverWALEmpty) {
Options options = CurrentOptions();
options.env = env_;
options.avoid_flush_during_recovery = false;
if (mem_env_ || encrypted_env_) {
ROCKSDB_GTEST_SKIP("Test requires non-mem/non-encrypted environment");
return;
}
if (!IsFallocateSupported()) {
return;
}
DestroyAndReopen(options);
size_t preallocated_size =
dbfull()->TEST_GetWalPreallocateBlockSize(options.write_buffer_size);
Close();
std::vector<std::string> filenames;
std::string last_log;
uint64_t last_log_num = 0;
ASSERT_OK(env_->GetChildren(dbname_, &filenames));
for (auto fname : filenames) {
uint64_t number;
FileType type;
if (ParseFileName(fname, &number, &type, nullptr)) {
if (type == kWalFile && number > last_log_num) {
last_log = fname;
}
}
}
ASSERT_NE(last_log, "");
last_log = dbname_ + '/' + last_log;
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->LoadDependency(
{{"DBImpl::PurgeObsoleteFiles:Begin",
"DBWALTest::TruncateLastLogAfterRecoverWithFlush:AfterRecover"},
{"DBWALTest::TruncateLastLogAfterRecoverWithFlush:AfterTruncate",
"DBImpl::DeleteObsoleteFileImpl::BeforeDeletion"}});
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->SetCallBack(
"PosixWritableFile::Close",
[](void* arg) { *(reinterpret_cast<size_t*>(arg)) = 0; });
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->EnableProcessing();
// Preallocate space for the empty log file. This could happen if WAL data
// was buffered in memory and the process crashed.
std::unique_ptr<WritableFile> log_file;
ASSERT_OK(env_->ReopenWritableFile(last_log, &log_file, EnvOptions()));
log_file->SetPreallocationBlockSize(preallocated_size);
log_file->PrepareWrite(0, 4096);
log_file.reset();
ASSERT_GE(GetAllocatedFileSize(last_log), preallocated_size);
port::Thread reopen_thread([&]() { Reopen(options); });
TEST_SYNC_POINT(
"DBWALTest::TruncateLastLogAfterRecoverWithFlush:AfterRecover");
// The preallocated space should be truncated.
EXPECT_LT(GetAllocatedFileSize(last_log), preallocated_size);
TEST_SYNC_POINT(
"DBWALTest::TruncateLastLogAfterRecoverWithFlush:AfterTruncate");
reopen_thread.join();
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->DisableProcessing();
ROCKSDB_NAMESPACE::SyncPoint::GetInstance()->ClearAllCallBacks();
}
Do not truncate WAL if in read_only mode (#8313) Summary: I noticed ```openat``` system call with ```O_WRONLY``` flag and ```sync_file_range``` and ```truncate``` on WAL file when using ```rocksdb::DB::OpenForReadOnly``` by way of ```db_bench --readonly=true --benchmarks=readseq --use_existing_db=1 --num=1 ...``` Noticed in ```strace``` after seeing the last modification time of the WAL file change after each run (with ```--readonly=true```). I think introduced by https://github.com/facebook/rocksdb/commit/7d7f14480e135a4939ed6903f46b3f7056aa837a from https://github.com/facebook/rocksdb/pull/8122 I added a test to catch the WAL file being truncated and the modification time on it changing. I am not sure if a mock filesystem with mock clock could be used to avoid having to sleep 1.1s. The test could also check the set of files is the same and that the sizes are also unchanged. Before: ``` [ RUN ] DBBasicTest.ReadOnlyReopenMtimeUnchanged db/db_basic_test.cc:182: Failure Expected equality of these values: file_mtime_after_readonly_reopen Which is: 1621611136 file_mtime_before_readonly_reopen Which is: 1621611135 file is: 000010.log [ FAILED ] DBBasicTest.ReadOnlyReopenMtimeUnchanged (1108 ms) ``` After: ``` [ RUN ] DBBasicTest.ReadOnlyReopenMtimeUnchanged [ OK ] DBBasicTest.ReadOnlyReopenMtimeUnchanged (1108 ms) ``` Pull Request resolved: https://github.com/facebook/rocksdb/pull/8313 Reviewed By: pdillinger Differential Revision: D28656925 Pulled By: jay-zhuang fbshipit-source-id: ea9e215cb53e7c830e76bc5fc75c45e21f12a1d6
4 years ago
TEST_F(DBWALTest, ReadOnlyRecoveryNoTruncate) {
constexpr size_t kKB = 1024;
Options options = CurrentOptions();
options.env = env_;
options.avoid_flush_during_recovery = true;
if (mem_env_) {
ROCKSDB_GTEST_SKIP("Test requires non-mem environment");
return;
}
if (!IsFallocateSupported()) {
return;
}
// create DB and close with file truncate disabled
std::atomic_bool enable_truncate{false};
SyncPoint::GetInstance()->SetCallBack(
"PosixWritableFile::Close", [&](void* arg) {
if (!enable_truncate) {
*(reinterpret_cast<size_t*>(arg)) = 0;
}
});
SyncPoint::GetInstance()->EnableProcessing();
DestroyAndReopen(options);
size_t preallocated_size =
dbfull()->TEST_GetWalPreallocateBlockSize(options.write_buffer_size);
ASSERT_OK(Put("foo", "v1"));
VectorLogPtr log_files_before;
ASSERT_OK(dbfull()->GetSortedWalFiles(log_files_before));
ASSERT_EQ(1, log_files_before.size());
auto& file_before = log_files_before[0];
ASSERT_LT(file_before->SizeFileBytes(), 1 * kKB);
// The log file has preallocated space.
auto db_size = GetAllocatedFileSize(dbname_ + file_before->PathName());
ASSERT_GE(db_size, preallocated_size);
Close();
// enable truncate and open DB as readonly, the file should not be truncated
// and DB size is not changed.
enable_truncate = true;
ASSERT_OK(ReadOnlyReopen(options));
VectorLogPtr log_files_after;
ASSERT_OK(dbfull()->GetSortedWalFiles(log_files_after));
ASSERT_EQ(1, log_files_after.size());
ASSERT_LT(log_files_after[0]->SizeFileBytes(), 1 * kKB);
ASSERT_EQ(log_files_after[0]->PathName(), file_before->PathName());
// The preallocated space should NOT be truncated.
// the DB size is almost the same.
ASSERT_NEAR(GetAllocatedFileSize(dbname_ + file_before->PathName()), db_size,
db_size / 100);
SyncPoint::GetInstance()->DisableProcessing();
SyncPoint::GetInstance()->ClearAllCallBacks();
}
#endif // ROCKSDB_FALLOCATE_PRESENT
#endif // ROCKSDB_PLATFORM_POSIX
TEST_F(DBWALTest, WalInManifestButNotInSortedWals) {
Options options = CurrentOptions();
options.track_and_verify_wals_in_manifest = true;
options.wal_recovery_mode = WALRecoveryMode::kAbsoluteConsistency;
// Build a way to make wal files selectively go missing
bool wals_go_missing = false;
struct MissingWalFs : public FileSystemWrapper {
MissingWalFs(const std::shared_ptr<FileSystem>& t,
bool* _wals_go_missing_flag)
: FileSystemWrapper(t), wals_go_missing_flag(_wals_go_missing_flag) {}
bool* wals_go_missing_flag;
IOStatus GetChildren(const std::string& dir, const IOOptions& io_opts,
std::vector<std::string>* r,
IODebugContext* dbg) override {
IOStatus s = target_->GetChildren(dir, io_opts, r, dbg);
if (s.ok() && *wals_go_missing_flag) {
for (size_t i = 0; i < r->size();) {
if (EndsWith(r->at(i), ".log")) {
r->erase(r->begin() + i);
} else {
++i;
}
}
}
return s;
}
const char* Name() const override { return "MissingWalFs"; }
};
auto my_fs =
std::make_shared<MissingWalFs>(env_->GetFileSystem(), &wals_go_missing);
std::unique_ptr<Env> my_env(NewCompositeEnv(my_fs));
options.env = my_env.get();
CreateAndReopenWithCF({"blah"}, options);
// Currently necessary to get a WAL tracked in manifest; see
// https://github.com/facebook/rocksdb/issues/10080
ASSERT_OK(Put(0, "x", "y"));
ASSERT_OK(db_->SyncWAL());
ASSERT_OK(Put(1, "x", "y"));
ASSERT_OK(db_->SyncWAL());
ASSERT_OK(Flush(1));
ASSERT_FALSE(dbfull()->GetVersionSet()->GetWalSet().GetWals().empty());
std::vector<std::unique_ptr<LogFile>> wals;
ASSERT_OK(db_->GetSortedWalFiles(wals));
wals_go_missing = true;
ASSERT_NOK(db_->GetSortedWalFiles(wals));
wals_go_missing = false;
Close();
}
TEST_F(DBWALTest, WalTermTest) {
Options options = CurrentOptions();
options.env = env_;
CreateAndReopenWithCF({"pikachu"}, options);
ASSERT_OK(Put(1, "foo", "bar"));
WriteOptions wo;
wo.sync = true;
wo.disableWAL = false;
WriteBatch batch;
ASSERT_OK(batch.Put("foo", "bar"));
batch.MarkWalTerminationPoint();
ASSERT_OK(batch.Put("foo2", "bar2"));
ASSERT_OK(dbfull()->Write(wo, &batch));
// make sure we can re-open it.
ASSERT_OK(TryReopenWithColumnFamilies({"default", "pikachu"}, options));
ASSERT_EQ("bar", Get(1, "foo"));
ASSERT_EQ("NOT_FOUND", Get(1, "foo2"));
}
TEST_F(DBWALTest, GetCompressedWalsAfterSync) {
if (db_->GetOptions().wal_compression == kNoCompression) {
ROCKSDB_GTEST_BYPASS("stream compression not present");
return;
}
Options options = GetDefaultOptions();
options.wal_recovery_mode = WALRecoveryMode::kPointInTimeRecovery;
options.create_if_missing = true;
options.env = env_;
options.avoid_flush_during_recovery = true;
options.track_and_verify_wals_in_manifest = true;
// Enable WAL compression so that the newly-created WAL will be non-empty
// after DB open, even if point-in-time WAL recovery encounters no
// corruption.
options.wal_compression = kZSTD;
DestroyAndReopen(options);
// Write something to memtable and WAL so that log_empty_ will be false after
// next DB::Open().
ASSERT_OK(Put("a", "v"));
Reopen(options);
// New WAL is created, thanks to !log_empty_.
ASSERT_OK(dbfull()->TEST_SwitchWAL());
ASSERT_OK(Put("b", "v"));
ASSERT_OK(db_->SyncWAL());
VectorLogPtr wals;
Status s = dbfull()->GetSortedWalFiles(wals);
ASSERT_OK(s);
}
TEST_F(DBWALTest, EmptyWalReopenTest) {
Options options = CurrentOptions();
options.env = env_;
CreateAndReopenWithCF({"pikachu"}, options);
// make sure we can re-open it.
ASSERT_OK(TryReopenWithColumnFamilies({"default", "pikachu"}, options));
{
std::vector<std::string> files;
int num_wal_files = 0;
ASSERT_OK(env_->GetChildren(dbname_, &files));
for (const auto& file : files) {
uint64_t number = 0;
FileType type = kWalFile;
if (ParseFileName(file, &number, &type) && type == kWalFile) {
num_wal_files++;
}
}
ASSERT_EQ(num_wal_files, 1);
}
}
} // namespace ROCKSDB_NAMESPACE
int main(int argc, char** argv) {
ROCKSDB_NAMESPACE::port::InstallStackTraceHandler();
::testing::InitGoogleTest(&argc, argv);
return RUN_ALL_TESTS();
}