rocksdb/db/repair.cc
Dhruba Borthakur 47c4191fe8 Reduce write amplification by merging files in L0 back into L0
Summary:
There is a new option called hybrid_mode which, when switched on,
causes HBase style compactions.  Files from L0 are
compacted back into L0. This meat of this compaction algorithm
is in PickCompactionHybrid().

All files reside in L0. That means all files have overlapping
keys. Each file has a time-bound, i.e. each file contains a
range of keys that were inserted around the same time. The
start-seqno and the end-seqno refers to the timeframe when
these keys were inserted.  Files that have contiguous seqno
are compacted together into a larger file. All files are
ordered from most recent to the oldest.

The current compaction algorithm starts to look for
candidate files starting from the most recent file. It continues to
add more files to the same compaction run as long as the
sum of the files chosen till now is smaller than the next
candidate file size. This logic needs to be debated
and validated.

The above logic should reduce write amplification to a
large extent... will publish numbers shortly.

Test Plan: dbstress runs for 6 hours with no data corruption (tested so far).

Differential Revision: https://reviews.facebook.net/D11289
2013-06-30 20:07:04 -07:00

386 lines
12 KiB
C++

// 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.
//
// We recover the contents of the descriptor from the other files we find.
// (1) Any log files are first converted to tables
// (2) We scan every table to compute
// (a) smallest/largest for the table
// (b) largest sequence number in the table
// (3) We generate descriptor contents:
// - log number is set to zero
// - next-file-number is set to 1 + largest file number we found
// - last-sequence-number is set to largest sequence# found across
// all tables (see 2c)
// - compaction pointers are cleared
// - every table file is added at level 0
//
// Possible optimization 1:
// (a) Compute total size and use to pick appropriate max-level M
// (b) Sort tables by largest sequence# in the table
// (c) For each table: if it overlaps earlier table, place in level-0,
// else place in level-M.
// Possible optimization 2:
// Store per-table metadata (smallest, largest, largest-seq#, ...)
// in the table's meta section to speed up ScanTable.
#include "db/builder.h"
#include "db/db_impl.h"
#include "db/dbformat.h"
#include "db/filename.h"
#include "db/log_reader.h"
#include "db/log_writer.h"
#include "db/memtable.h"
#include "db/table_cache.h"
#include "db/version_edit.h"
#include "db/write_batch_internal.h"
#include "leveldb/comparator.h"
#include "leveldb/db.h"
#include "leveldb/env.h"
namespace leveldb {
namespace {
class Repairer {
public:
Repairer(const std::string& dbname, const Options& options)
: dbname_(dbname),
env_(options.env),
icmp_(options.comparator),
ipolicy_(options.filter_policy),
options_(SanitizeOptions(dbname, &icmp_, &ipolicy_, options)),
next_file_number_(1) {
// TableCache can be small since we expect each table to be opened once.
table_cache_ = new TableCache(dbname_, &options_, storage_options_, 10);
edit_ = new VersionEdit(options.num_levels);
}
~Repairer() {
delete table_cache_;
delete edit_;
}
Status Run() {
Status status = FindFiles();
if (status.ok()) {
ConvertLogFilesToTables();
ExtractMetaData();
status = WriteDescriptor();
}
if (status.ok()) {
unsigned long long bytes = 0;
for (size_t i = 0; i < tables_.size(); i++) {
bytes += tables_[i].meta.file_size;
}
Log(options_.info_log,
"**** Repaired leveldb %s; "
"recovered %d files; %llu bytes. "
"Some data may have been lost. "
"****",
dbname_.c_str(),
static_cast<int>(tables_.size()),
bytes);
}
return status;
}
private:
struct TableInfo {
FileMetaData meta;
SequenceNumber min_sequence;
SequenceNumber max_sequence;
};
std::string const dbname_;
Env* const env_;
InternalKeyComparator const icmp_;
InternalFilterPolicy const ipolicy_;
Options const options_;
bool owns_cache_;
TableCache* table_cache_;
VersionEdit* edit_;
std::vector<std::string> manifests_;
std::vector<uint64_t> table_numbers_;
std::vector<uint64_t> logs_;
std::vector<TableInfo> tables_;
uint64_t next_file_number_;
const EnvOptions storage_options_;
Status FindFiles() {
std::vector<std::string> filenames;
Status status = env_->GetChildren(dbname_, &filenames);
if (!status.ok()) {
return status;
}
if (filenames.empty()) {
return Status::IOError(dbname_, "repair found no files");
}
uint64_t number;
FileType type;
for (size_t i = 0; i < filenames.size(); i++) {
if (ParseFileName(filenames[i], &number, &type)) {
if (type == kDescriptorFile) {
manifests_.push_back(filenames[i]);
} else {
if (number + 1 > next_file_number_) {
next_file_number_ = number + 1;
}
if (type == kLogFile) {
logs_.push_back(number);
} else if (type == kTableFile) {
table_numbers_.push_back(number);
} else {
// Ignore other files
}
}
}
}
return status;
}
void ConvertLogFilesToTables() {
for (size_t i = 0; i < logs_.size(); i++) {
std::string logname = LogFileName(dbname_, logs_[i]);
Status status = ConvertLogToTable(logs_[i]);
if (!status.ok()) {
Log(options_.info_log, "Log #%llu: ignoring conversion error: %s",
(unsigned long long) logs_[i],
status.ToString().c_str());
}
ArchiveFile(logname);
}
}
Status ConvertLogToTable(uint64_t log) {
struct LogReporter : public log::Reader::Reporter {
Env* env;
std::shared_ptr<Logger> info_log;
uint64_t lognum;
virtual void Corruption(size_t bytes, const Status& s) {
// We print error messages for corruption, but continue repairing.
Log(info_log, "Log #%llu: dropping %d bytes; %s",
(unsigned long long) lognum,
static_cast<int>(bytes),
s.ToString().c_str());
}
};
// Open the log file
std::string logname = LogFileName(dbname_, log);
unique_ptr<SequentialFile> lfile;
Status status = env_->NewSequentialFile(logname, &lfile, storage_options_);
if (!status.ok()) {
return status;
}
// Create the log reader.
LogReporter reporter;
reporter.env = env_;
reporter.info_log = options_.info_log;
reporter.lognum = log;
// We intentially make log::Reader do checksumming so that
// corruptions cause entire commits to be skipped instead of
// propagating bad information (like overly large sequence
// numbers).
log::Reader reader(std::move(lfile), &reporter, false/*do not checksum*/,
0/*initial_offset*/);
// Read all the records and add to a memtable
std::string scratch;
Slice record;
WriteBatch batch;
MemTable* mem = new MemTable(icmp_, options_.num_levels);
mem->Ref();
int counter = 0;
while (reader.ReadRecord(&record, &scratch)) {
if (record.size() < 12) {
reporter.Corruption(
record.size(), Status::Corruption("log record too small"));
continue;
}
WriteBatchInternal::SetContents(&batch, record);
status = WriteBatchInternal::InsertInto(&batch, mem);
if (status.ok()) {
counter += WriteBatchInternal::Count(&batch);
} else {
Log(options_.info_log, "Log #%llu: ignoring %s",
(unsigned long long) log,
status.ToString().c_str());
status = Status::OK(); // Keep going with rest of file
}
}
// Do not record a version edit for this conversion to a Table
// since ExtractMetaData() will also generate edits.
FileMetaData meta;
meta.number = next_file_number_++;
Iterator* iter = mem->NewIterator();
status = BuildTable(dbname_, env_, options_, storage_options_,
table_cache_, iter, &meta,
icmp_.user_comparator(), 0, 0);
delete iter;
mem->Unref();
mem = nullptr;
if (status.ok()) {
if (meta.file_size > 0) {
table_numbers_.push_back(meta.number);
}
}
Log(options_.info_log, "Log #%llu: %d ops saved to Table #%llu %s",
(unsigned long long) log,
counter,
(unsigned long long) meta.number,
status.ToString().c_str());
return status;
}
void ExtractMetaData() {
std::vector<TableInfo> kept;
for (size_t i = 0; i < table_numbers_.size(); i++) {
TableInfo t;
t.meta.number = table_numbers_[i];
Status status = ScanTable(&t);
if (!status.ok()) {
std::string fname = TableFileName(dbname_, table_numbers_[i]);
Log(options_.info_log, "Table #%llu: ignoring %s",
(unsigned long long) table_numbers_[i],
status.ToString().c_str());
ArchiveFile(fname);
} else {
tables_.push_back(t);
}
}
}
Status ScanTable(TableInfo* t) {
std::string fname = TableFileName(dbname_, t->meta.number);
int counter = 0;
Status status = env_->GetFileSize(fname, &t->meta.file_size);
if (status.ok()) {
Iterator* iter = table_cache_->NewIterator(
ReadOptions(), storage_options_, t->meta.number, t->meta.file_size);
bool empty = true;
ParsedInternalKey parsed;
t->min_sequence = 0;
t->max_sequence = 0;
for (iter->SeekToFirst(); iter->Valid(); iter->Next()) {
Slice key = iter->key();
if (!ParseInternalKey(key, &parsed)) {
Log(options_.info_log, "Table #%llu: unparsable key %s",
(unsigned long long) t->meta.number,
EscapeString(key).c_str());
continue;
}
counter++;
if (empty) {
empty = false;
t->meta.smallest.DecodeFrom(key);
}
t->meta.largest.DecodeFrom(key);
if (parsed.sequence < t->min_sequence) {
t->min_sequence = parsed.sequence;
}
if (parsed.sequence > t->max_sequence) {
t->max_sequence = parsed.sequence;
}
}
if (!iter->status().ok()) {
status = iter->status();
}
delete iter;
}
Log(options_.info_log, "Table #%llu: %d entries %s",
(unsigned long long) t->meta.number,
counter,
status.ToString().c_str());
return status;
}
Status WriteDescriptor() {
std::string tmp = TempFileName(dbname_, 1);
unique_ptr<WritableFile> file;
Status status = env_->NewWritableFile(tmp, &file, storage_options_);
if (!status.ok()) {
return status;
}
SequenceNumber max_sequence = 0;
for (size_t i = 0; i < tables_.size(); i++) {
if (max_sequence < tables_[i].max_sequence) {
max_sequence = tables_[i].max_sequence;
}
}
edit_->SetComparatorName(icmp_.user_comparator()->Name());
edit_->SetLogNumber(0);
edit_->SetNextFile(next_file_number_);
edit_->SetLastSequence(max_sequence);
for (size_t i = 0; i < tables_.size(); i++) {
// TODO(opt): separate out into multiple levels
const TableInfo& t = tables_[i];
edit_->AddFile(0, t.meta.number, t.meta.file_size,
t.meta.smallest, t.meta.largest,
t.min_sequence, t.max_sequence);
}
//fprintf(stderr, "NewDescriptor:\n%s\n", edit_.DebugString().c_str());
{
log::Writer log(std::move(file));
std::string record;
edit_->EncodeTo(&record);
status = log.AddRecord(record);
}
if (!status.ok()) {
env_->DeleteFile(tmp);
} else {
// Discard older manifests
for (size_t i = 0; i < manifests_.size(); i++) {
ArchiveFile(dbname_ + "/" + manifests_[i]);
}
// Install new manifest
status = env_->RenameFile(tmp, DescriptorFileName(dbname_, 1));
if (status.ok()) {
status = SetCurrentFile(env_, dbname_, 1);
} else {
env_->DeleteFile(tmp);
}
}
return status;
}
void ArchiveFile(const std::string& fname) {
// Move into another directory. E.g., for
// dir/foo
// rename to
// dir/lost/foo
const char* slash = strrchr(fname.c_str(), '/');
std::string new_dir;
if (slash != nullptr) {
new_dir.assign(fname.data(), slash - fname.data());
}
new_dir.append("/lost");
env_->CreateDir(new_dir); // Ignore error
std::string new_file = new_dir;
new_file.append("/");
new_file.append((slash == nullptr) ? fname.c_str() : slash + 1);
Status s = env_->RenameFile(fname, new_file);
Log(options_.info_log, "Archiving %s: %s\n",
fname.c_str(), s.ToString().c_str());
}
};
} // namespace
Status RepairDB(const std::string& dbname, const Options& options) {
Repairer repairer(dbname, options);
return repairer.Run();
}
} // namespace leveldb