rocksdb/db/job_context.h

240 lines
7.5 KiB
C
Raw Normal View History

// Copyright (c) 2011-present, Facebook, Inc. All rights reserved.
// This source code is licensed under both the GPLv2 (found in the
// COPYING file in the root directory) and Apache 2.0 License
// (found in the LICENSE.Apache file in the root directory).
//
// 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.
#pragma once
#include <string>
#include <vector>
#include "db/column_family.h"
#include "db/log_writer.h"
#include "db/version_set.h"
namespace ROCKSDB_NAMESPACE {
class MemTable;
struct SuperVersion;
struct SuperVersionContext {
struct WriteStallNotification {
WriteStallInfo write_stall_info;
const ImmutableOptions* immutable_options;
};
autovector<SuperVersion*> superversions_to_free;
#ifndef ROCKSDB_DISABLE_STALL_NOTIFICATION
autovector<WriteStallNotification> write_stall_notifications;
#endif
std::unique_ptr<SuperVersion>
new_superversion; // if nullptr no new superversion
explicit SuperVersionContext(bool create_superversion = false)
: new_superversion(create_superversion ? new SuperVersion() : nullptr) {}
Add manifest fix-up utility for file temperatures (#9683) Summary: The goal of this change is to allow changes to the "current" (in FileSystem) file temperatures to feed back into DB metadata, so that they can inform decisions and stats reporting. In part because of modular code factoring, it doesn't seem easy to do this automagically, where opening an SST file and observing current Temperature different from expected would trigger a change in metadata and DB manifest write (essentially giving the deep read path access to the write path). It is also difficult to do this while the DB is open because of the limitations of LogAndApply. This change allows updating file temperature metadata on a closed DB using an experimental utility function UpdateManifestForFilesState() or `ldb update_manifest --update_temperatures`. This should suffice for "migration" scenarios where outside tooling has placed or re-arranged DB files into a (different) tiered configuration without going through RocksDB itself (currently, only compaction can change temperature metadata). Some details: * Refactored and added unit test for `ldb unsafe_remove_sst_file` because of shared functionality * Pulled in autovector.h changes from https://github.com/facebook/rocksdb/issues/9546 to fix SuperVersionContext move constructor (related to an older draft of this change) Possible follow-up work: * Support updating manifest with file checksums, such as when a new checksum function is used and want existing DB metadata updated for it. * It's possible that for some repair scenarios, lighter weight than full repair, we might want to support UpdateManifestForFilesState() to modify critical file details like size or checksum using same algorithm. But let's make sure these are differentiated from modifying file details in ways that don't suspect corruption (or require extreme trust). Pull Request resolved: https://github.com/facebook/rocksdb/pull/9683 Test Plan: unit tests added Reviewed By: jay-zhuang Differential Revision: D34798828 Pulled By: pdillinger fbshipit-source-id: cfd83e8fb10761d8c9e7f9c020d68c9106a95554
2022-03-18 23:35:51 +00:00
explicit SuperVersionContext(SuperVersionContext&& other) noexcept
: superversions_to_free(std::move(other.superversions_to_free)),
#ifndef ROCKSDB_DISABLE_STALL_NOTIFICATION
write_stall_notifications(std::move(other.write_stall_notifications)),
#endif
new_superversion(std::move(other.new_superversion)) {
}
Add manifest fix-up utility for file temperatures (#9683) Summary: The goal of this change is to allow changes to the "current" (in FileSystem) file temperatures to feed back into DB metadata, so that they can inform decisions and stats reporting. In part because of modular code factoring, it doesn't seem easy to do this automagically, where opening an SST file and observing current Temperature different from expected would trigger a change in metadata and DB manifest write (essentially giving the deep read path access to the write path). It is also difficult to do this while the DB is open because of the limitations of LogAndApply. This change allows updating file temperature metadata on a closed DB using an experimental utility function UpdateManifestForFilesState() or `ldb update_manifest --update_temperatures`. This should suffice for "migration" scenarios where outside tooling has placed or re-arranged DB files into a (different) tiered configuration without going through RocksDB itself (currently, only compaction can change temperature metadata). Some details: * Refactored and added unit test for `ldb unsafe_remove_sst_file` because of shared functionality * Pulled in autovector.h changes from https://github.com/facebook/rocksdb/issues/9546 to fix SuperVersionContext move constructor (related to an older draft of this change) Possible follow-up work: * Support updating manifest with file checksums, such as when a new checksum function is used and want existing DB metadata updated for it. * It's possible that for some repair scenarios, lighter weight than full repair, we might want to support UpdateManifestForFilesState() to modify critical file details like size or checksum using same algorithm. But let's make sure these are differentiated from modifying file details in ways that don't suspect corruption (or require extreme trust). Pull Request resolved: https://github.com/facebook/rocksdb/pull/9683 Test Plan: unit tests added Reviewed By: jay-zhuang Differential Revision: D34798828 Pulled By: pdillinger fbshipit-source-id: cfd83e8fb10761d8c9e7f9c020d68c9106a95554
2022-03-18 23:35:51 +00:00
// No copies
SuperVersionContext(const SuperVersionContext& other) = delete;
void operator=(const SuperVersionContext& other) = delete;
void NewSuperVersion() {
new_superversion = std::unique_ptr<SuperVersion>(new SuperVersion());
}
inline bool HaveSomethingToDelete() const {
#ifndef ROCKSDB_DISABLE_STALL_NOTIFICATION
return !superversions_to_free.empty() ||
!write_stall_notifications.empty();
#else
return !superversions_to_free.empty();
#endif
}
void PushWriteStallNotification(WriteStallCondition old_cond,
WriteStallCondition new_cond,
const std::string& name,
const ImmutableOptions* ioptions) {
#if !defined(ROCKSDB_LITE) && !defined(ROCKSDB_DISABLE_STALL_NOTIFICATION)
WriteStallNotification notif;
notif.write_stall_info.cf_name = name;
notif.write_stall_info.condition.prev = old_cond;
notif.write_stall_info.condition.cur = new_cond;
notif.immutable_options = ioptions;
write_stall_notifications.push_back(notif);
#else
(void)old_cond;
(void)new_cond;
(void)name;
(void)ioptions;
#endif // !defined(ROCKSDB_LITE) && !defined(ROCKSDB_DISABLE_STALL_NOTIFICATION)
}
void Clean() {
#if !defined(ROCKSDB_LITE) && !defined(ROCKSDB_DISABLE_STALL_NOTIFICATION)
// notify listeners on changed write stall conditions
for (auto& notif : write_stall_notifications) {
for (auto& listener : notif.immutable_options->listeners) {
listener->OnStallConditionsChanged(notif.write_stall_info);
}
}
write_stall_notifications.clear();
#endif // !ROCKSDB_LITE
// free superversions
for (auto s : superversions_to_free) {
delete s;
}
superversions_to_free.clear();
}
~SuperVersionContext() {
#ifndef ROCKSDB_DISABLE_STALL_NOTIFICATION
assert(write_stall_notifications.empty());
#endif
assert(superversions_to_free.empty());
}
};
struct JobContext {
inline bool HaveSomethingToDelete() const {
return !(full_scan_candidate_files.empty() && sst_delete_files.empty() &&
blob_delete_files.empty() && log_delete_files.empty() &&
manifest_delete_files.empty());
}
inline bool HaveSomethingToClean() const {
bool sv_have_sth = false;
for (const auto& sv_ctx : superversion_contexts) {
if (sv_ctx.HaveSomethingToDelete()) {
sv_have_sth = true;
break;
}
}
return memtables_to_free.size() > 0 || logs_to_free.size() > 0 ||
Deflake write-prepared and write-unprepared tests (#8696) Summary: The `JobContext::job_snapshot` referenced DB state but could have been deleted by a BG thread after the signal/unlock allowing shutdown to proceed. Then we would see an error like this (valgrind): ``` ==354104== Thread 2: ==354104== Invalid read of size 8 ==354104== at 0x694C4D: rocksdb::ManagedSnapshot::~ManagedSnapshot() (snapshot_impl.cc:20) ==354104== by 0x58F5BA: operator() (unique_ptr.h:81) ==354104== by 0x58F5BA: operator() (unique_ptr.h:75) ==354104== by 0x58F5BA: ~unique_ptr (unique_ptr.h:292) ==354104== by 0x58F5BA: rocksdb::JobContext::~JobContext() (job_context.h:221) ==354104== by 0x5F155E: rocksdb::DBImpl::BackgroundCallCompaction(rocksdb::DBImpl::PrepickedCompaction*, rocksdb::Env::Priority) (db_impl_compaction_flush.cc:2696) ==354104== by 0x5F1BC2: rocksdb::DBImpl::BGWorkCompaction(void*) (db_impl_compaction_flush.cc:2468) ==354104== by 0x83707A: operator() (std_function.h:688) ==354104== by 0x83707A: rocksdb::ThreadPoolImpl::Impl::BGThread(unsigned long) (threadpool_imp.cc:266) ==354104== by 0x8373ED: rocksdb::ThreadPoolImpl::Impl::BGThreadWrapper(void*) (threadpool_imp.cc:307) ==354104== by 0x492A800: execute_native_thread_routine (in /usr/local/fbcode/platform009/lib/libstdc++.so.6.0.28) ==354104== by 0x4A5020B: start_thread (in /usr/local/fbcode/platform009/lib/libpthread-2.30.so) ==354104== by 0x4CF281E: clone (in /usr/local/fbcode/platform009/lib/libc-2.30.so) ``` Pull Request resolved: https://github.com/facebook/rocksdb/pull/8696 Test Plan: unable to repro Reviewed By: pdillinger Differential Revision: D30505277 Pulled By: ajkr fbshipit-source-id: 5a99f34137cd14d06b0f624add6d37a70a61135d
2021-08-24 06:08:25 +00:00
job_snapshot != nullptr || sv_have_sth;
}
CompactionIterator sees consistent view of which keys are committed (#9830) Summary: **This PR does not affect the functionality of `DB` and write-committed transactions.** `CompactionIterator` uses `KeyCommitted(seq)` to determine if a key in the database is committed. As the name 'write-committed' implies, if write-committed policy is used, a key exists in the database only if it is committed. In fact, the implementation of `KeyCommitted()` is as follows: ``` inline bool KeyCommitted(SequenceNumber seq) { // For non-txn-db and write-committed, snapshot_checker_ is always nullptr. return snapshot_checker_ == nullptr || snapshot_checker_->CheckInSnapshot(seq, kMaxSequence) == SnapshotCheckerResult::kInSnapshot; } ``` With that being said, we focus on write-prepared/write-unprepared transactions. A few notes: - A key can exist in the db even if it's uncommitted. Therefore, we rely on `snapshot_checker_` to determine data visibility. We also require that all writes go through transaction API instead of the raw `WriteBatch` + `Write`, thus at most one uncommitted version of one user key can exist in the database. - `CompactionIterator` outputs a key as long as the key is uncommitted. Due to the above reasons, it is possible that `CompactionIterator` decides to output an uncommitted key without doing further checks on the key (`NextFromInput()`). By the time the key is being prepared for output, the key becomes committed because the `snapshot_checker_(seq, kMaxSequence)` becomes true in the implementation of `KeyCommitted()`. Then `CompactionIterator` will try to zero its sequence number and hit assertion error if the key is a tombstone. To fix this issue, we should make the `CompactionIterator` see a consistent view of the input keys. Note that for write-prepared/write-unprepared, the background flush/compaction jobs already take a "job snapshot" before starting processing keys. The job snapshot is released only after the entire flush/compaction finishes. We can use this snapshot to determine whether a key is committed or not with minor change to `KeyCommitted()`. ``` inline bool KeyCommitted(SequenceNumber sequence) { // For non-txn-db and write-committed, snapshot_checker_ is always nullptr. return snapshot_checker_ == nullptr || snapshot_checker_->CheckInSnapshot(sequence, job_snapshot_) == SnapshotCheckerResult::kInSnapshot; } ``` As a result, whether a key is committed or not will remain a constant throughout compaction, causing no trouble for `CompactionIterator`s assertions. Pull Request resolved: https://github.com/facebook/rocksdb/pull/9830 Test Plan: make check Reviewed By: ltamasi Differential Revision: D35561162 Pulled By: riversand963 fbshipit-source-id: 0e00d200c195240341cfe6d34cbc86798b315b9f
2022-04-14 18:11:04 +00:00
SequenceNumber GetJobSnapshotSequence() const {
if (job_snapshot) {
assert(job_snapshot->snapshot());
return job_snapshot->snapshot()->GetSequenceNumber();
}
return kMaxSequenceNumber;
}
// Structure to store information for candidate files to delete.
struct CandidateFileInfo {
std::string file_name;
std::string file_path;
CandidateFileInfo(std::string name, std::string path)
: file_name(std::move(name)), file_path(std::move(path)) {}
bool operator==(const CandidateFileInfo& other) const {
return file_name == other.file_name &&
file_path == other.file_path;
}
};
// Unique job id
int job_id;
// a list of all files that we'll consider deleting
// (every once in a while this is filled up with all files
// in the DB directory)
Speed up FindObsoleteFiles() Summary: There are two versions of FindObsoleteFiles(): * full scan, which is executed every 6 hours (and it's terribly slow) * no full scan, which is executed every time a background process finishes and iterator is deleted This diff is optimizing the second case (no full scan). Here's what we do before the diff: * Get the list of obsolete files (files with ref==0). Some files in obsolete_files set might actually be live. * Get the list of live files to avoid deleting files that are live. * Delete files that are in obsolete_files and not in live_files. After this diff: * The only files with ref==0 that are still live are files that have been part of move compaction. Don't include moved files in obsolete_files. * Get the list of obsolete files (which exclude moved files). * No need to get the list of live files, since all files in obsolete_files need to be deleted. I'll post the benchmark results, but you can get the feel of it here: https://reviews.facebook.net/D30123 This depends on D30123. P.S. We should do full scan only in failure scenarios, not every 6 hours. I'll do this in a follow-up diff. Test Plan: One new unit test. Made sure that unit test fails if we don't have a `if (!f->moved)` safeguard in ~Version. make check Big number of compactions and flushes: ./db_stress --threads=30 --ops_per_thread=20000000 --max_key=10000 --column_families=20 --clear_column_family_one_in=10000000 --verify_before_write=0 --reopen=15 --max_background_compactions=10 --max_background_flushes=10 --db=/fast-rocksdb-tmp/db_stress --prefixpercent=0 --iterpercent=0 --writepercent=75 --db_write_buffer_size=2000000 Reviewers: yhchiang, rven, sdong Reviewed By: sdong Subscribers: dhruba, leveldb Differential Revision: https://reviews.facebook.net/D30249
2014-12-22 11:04:45 +00:00
// (filled only if we're doing full scan)
std::vector<CandidateFileInfo> full_scan_candidate_files;
// the list of all live sst files that cannot be deleted
std::vector<uint64_t> sst_live;
// the list of sst files that we need to delete
std::vector<ObsoleteFileInfo> sst_delete_files;
// the list of all live blob files that cannot be deleted
std::vector<uint64_t> blob_live;
// the list of blob files that we need to delete
std::vector<ObsoleteBlobFileInfo> blob_delete_files;
// a list of log files that we need to delete
std::vector<uint64_t> log_delete_files;
// a list of log files that we need to preserve during full purge since they
// will be reused later
std::vector<uint64_t> log_recycle_files;
// a list of manifest files that we need to delete
std::vector<std::string> manifest_delete_files;
// a list of memtables to be free
autovector<MemTable*> memtables_to_free;
// contexts for installing superversions for multiple column families
std::vector<SuperVersionContext> superversion_contexts;
autovector<log::Writer*> logs_to_free;
// the current manifest_file_number, log_number and prev_log_number
// that corresponds to the set of files in 'live'.
uint64_t manifest_file_number;
uint64_t pending_manifest_file_number;
uint64_t log_number;
uint64_t prev_log_number;
uint64_t min_pending_output = 0;
uint64_t prev_total_log_size = 0;
size_t num_alive_log_files = 0;
uint64_t size_log_to_delete = 0;
// Snapshot taken before flush/compaction job.
std::unique_ptr<ManagedSnapshot> job_snapshot;
explicit JobContext(int _job_id, bool create_superversion = false) {
job_id = _job_id;
manifest_file_number = 0;
pending_manifest_file_number = 0;
log_number = 0;
prev_log_number = 0;
superversion_contexts.emplace_back(
SuperVersionContext(create_superversion));
}
// For non-empty JobContext Clean() has to be called at least once before
// before destruction (see asserts in ~JobContext()). Should be called with
// unlocked DB mutex. Destructor doesn't call Clean() to avoid accidentally
// doing potentially slow Clean() with locked DB mutex.
void Clean() {
// free superversions
for (auto& sv_context : superversion_contexts) {
sv_context.Clean();
}
// free pending memtables
for (auto m : memtables_to_free) {
delete m;
}
for (auto l : logs_to_free) {
delete l;
}
memtables_to_free.clear();
logs_to_free.clear();
job_snapshot.reset();
}
~JobContext() {
assert(memtables_to_free.size() == 0);
assert(logs_to_free.size() == 0);
}
};
} // namespace ROCKSDB_NAMESPACE