2011-03-18 23:37:00 +01:00
|
|
|
// 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_impl.h"
|
|
|
|
|
|
|
|
#include <algorithm>
|
|
|
|
#include <set>
|
|
|
|
#include <string>
|
|
|
|
#include <stdint.h>
|
|
|
|
#include <stdio.h>
|
|
|
|
#include <vector>
|
2012-08-18 01:06:05 +02:00
|
|
|
#include <algorithm>
|
2011-03-18 23:37:00 +01:00
|
|
|
#include "db/builder.h"
|
|
|
|
#include "db/db_iter.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_set.h"
|
|
|
|
#include "db/write_batch_internal.h"
|
2011-03-30 20:35:40 +02:00
|
|
|
#include "leveldb/db.h"
|
|
|
|
#include "leveldb/env.h"
|
|
|
|
#include "leveldb/status.h"
|
|
|
|
#include "leveldb/table.h"
|
|
|
|
#include "leveldb/table_builder.h"
|
2011-03-18 23:37:00 +01:00
|
|
|
#include "port/port.h"
|
|
|
|
#include "table/block.h"
|
|
|
|
#include "table/merger.h"
|
|
|
|
#include "table/two_level_iterator.h"
|
|
|
|
#include "util/coding.h"
|
|
|
|
#include "util/logging.h"
|
|
|
|
#include "util/mutexlock.h"
|
2012-08-21 09:33:21 +02:00
|
|
|
#include "util/build_version.h"
|
2012-10-26 23:55:02 +02:00
|
|
|
#include "util/auto_split_logger.h"
|
2011-03-18 23:37:00 +01:00
|
|
|
|
|
|
|
namespace leveldb {
|
|
|
|
|
2012-09-18 22:18:32 +02:00
|
|
|
void dumpLeveldbBuildVersion(Logger * log);
|
|
|
|
|
2012-10-26 23:55:02 +02:00
|
|
|
static Status NewLogger(const std::string& dbname,
|
|
|
|
const std::string& db_log_dir,
|
|
|
|
Env* env,
|
|
|
|
size_t max_log_file_size,
|
|
|
|
Logger** logger) {
|
|
|
|
std::string db_absolute_path;
|
|
|
|
env->GetAbsolutePath(dbname, &db_absolute_path);
|
|
|
|
|
|
|
|
if (max_log_file_size > 0) { // need to auto split the log file?
|
|
|
|
AutoSplitLogger<Logger>* auto_split_logger =
|
|
|
|
new AutoSplitLogger<Logger>(env, dbname, db_log_dir, max_log_file_size);
|
|
|
|
Status s = auto_split_logger->GetStatus();
|
|
|
|
if (!s.ok()) {
|
|
|
|
delete auto_split_logger;
|
|
|
|
} else {
|
|
|
|
*logger = auto_split_logger;
|
|
|
|
}
|
|
|
|
return s;
|
|
|
|
} else {
|
|
|
|
// Open a log file in the same directory as the db
|
|
|
|
env->CreateDir(dbname); // In case it does not exist
|
|
|
|
std::string fname = InfoLogFileName(dbname, db_absolute_path, db_log_dir);
|
|
|
|
env->RenameFile(fname, OldInfoLogFileName(dbname, env->NowMicros(),
|
|
|
|
db_absolute_path, db_log_dir));
|
|
|
|
return env->NewLogger(fname, logger);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2012-03-09 01:23:21 +01:00
|
|
|
// Information kept for every waiting writer
|
|
|
|
struct DBImpl::Writer {
|
|
|
|
Status status;
|
|
|
|
WriteBatch* batch;
|
|
|
|
bool sync;
|
2012-07-05 22:39:28 +02:00
|
|
|
bool disableWAL;
|
2012-03-09 01:23:21 +01:00
|
|
|
bool done;
|
|
|
|
port::CondVar cv;
|
|
|
|
|
|
|
|
explicit Writer(port::Mutex* mu) : cv(mu) { }
|
|
|
|
};
|
|
|
|
|
2011-03-18 23:37:00 +01:00
|
|
|
struct DBImpl::CompactionState {
|
|
|
|
Compaction* const compaction;
|
|
|
|
|
|
|
|
// Sequence numbers < smallest_snapshot are not significant since we
|
|
|
|
// will never have to service a snapshot below smallest_snapshot.
|
|
|
|
// Therefore if we have seen a sequence number S <= smallest_snapshot,
|
|
|
|
// we can drop all entries for the same key with sequence numbers < S.
|
|
|
|
SequenceNumber smallest_snapshot;
|
|
|
|
|
|
|
|
// Files produced by compaction
|
|
|
|
struct Output {
|
|
|
|
uint64_t number;
|
|
|
|
uint64_t file_size;
|
|
|
|
InternalKey smallest, largest;
|
|
|
|
};
|
|
|
|
std::vector<Output> outputs;
|
|
|
|
|
|
|
|
// State kept for output being generated
|
|
|
|
WritableFile* outfile;
|
|
|
|
TableBuilder* builder;
|
|
|
|
|
|
|
|
uint64_t total_bytes;
|
|
|
|
|
|
|
|
Output* current_output() { return &outputs[outputs.size()-1]; }
|
|
|
|
|
|
|
|
explicit CompactionState(Compaction* c)
|
|
|
|
: compaction(c),
|
|
|
|
outfile(NULL),
|
|
|
|
builder(NULL),
|
|
|
|
total_bytes(0) {
|
|
|
|
}
|
|
|
|
};
|
|
|
|
|
2012-10-21 10:49:48 +02:00
|
|
|
struct DBImpl::DeletionState {
|
|
|
|
|
|
|
|
// the set of all live files that cannot be deleted
|
|
|
|
std::set<uint64_t> live;
|
|
|
|
|
|
|
|
// a list of all siles that exists in the db directory
|
|
|
|
std::vector<std::string> allfiles;
|
|
|
|
|
|
|
|
// the current filenumber, lognumber and prevlognumber
|
|
|
|
// that corresponds to the set of files in 'live'.
|
|
|
|
uint64_t filenumber, lognumber, prevlognumber;
|
|
|
|
|
|
|
|
// the list of all files to be evicted from the table cahce
|
|
|
|
std::vector<uint64_t> files_to_evict;
|
|
|
|
};
|
|
|
|
|
2011-03-18 23:37:00 +01:00
|
|
|
// Fix user-supplied options to be reasonable
|
|
|
|
template <class T,class V>
|
|
|
|
static void ClipToRange(T* ptr, V minvalue, V maxvalue) {
|
2011-04-21 00:48:11 +02:00
|
|
|
if (static_cast<V>(*ptr) > maxvalue) *ptr = maxvalue;
|
|
|
|
if (static_cast<V>(*ptr) < minvalue) *ptr = minvalue;
|
2011-03-18 23:37:00 +01:00
|
|
|
}
|
|
|
|
Options SanitizeOptions(const std::string& dbname,
|
|
|
|
const InternalKeyComparator* icmp,
|
2012-04-17 17:36:46 +02:00
|
|
|
const InternalFilterPolicy* ipolicy,
|
2011-03-18 23:37:00 +01:00
|
|
|
const Options& src) {
|
|
|
|
Options result = src;
|
|
|
|
result.comparator = icmp;
|
2012-04-17 17:36:46 +02:00
|
|
|
result.filter_policy = (src.filter_policy != NULL) ? ipolicy : NULL;
|
|
|
|
ClipToRange(&result.max_open_files, 20, 50000);
|
|
|
|
ClipToRange(&result.write_buffer_size, 64<<10, 1<<30);
|
|
|
|
ClipToRange(&result.block_size, 1<<10, 4<<20);
|
2011-03-18 23:37:00 +01:00
|
|
|
if (result.info_log == NULL) {
|
2012-10-26 23:55:02 +02:00
|
|
|
Status s = NewLogger(dbname, result.db_log_dir, src.env,
|
|
|
|
result.max_log_file_size, &result.info_log);
|
2011-03-18 23:37:00 +01:00
|
|
|
if (!s.ok()) {
|
|
|
|
// No place suitable for logging
|
2011-07-21 04:40:18 +02:00
|
|
|
result.info_log = NULL;
|
2011-03-18 23:37:00 +01:00
|
|
|
}
|
|
|
|
}
|
2011-04-12 21:38:58 +02:00
|
|
|
if (result.block_cache == NULL) {
|
|
|
|
result.block_cache = NewLRUCache(8 << 20);
|
|
|
|
}
|
2011-03-18 23:37:00 +01:00
|
|
|
return result;
|
|
|
|
}
|
|
|
|
|
|
|
|
DBImpl::DBImpl(const Options& options, const std::string& dbname)
|
|
|
|
: env_(options.env),
|
|
|
|
internal_comparator_(options.comparator),
|
2012-04-17 17:36:46 +02:00
|
|
|
internal_filter_policy_(options.filter_policy),
|
|
|
|
options_(SanitizeOptions(
|
|
|
|
dbname, &internal_comparator_, &internal_filter_policy_, options)),
|
2011-03-18 23:37:00 +01:00
|
|
|
owns_info_log_(options_.info_log != options.info_log),
|
2011-04-12 21:38:58 +02:00
|
|
|
owns_cache_(options_.block_cache != options.block_cache),
|
2011-03-18 23:37:00 +01:00
|
|
|
dbname_(dbname),
|
|
|
|
db_lock_(NULL),
|
|
|
|
shutting_down_(NULL),
|
|
|
|
bg_cv_(&mutex_),
|
|
|
|
mem_(new MemTable(internal_comparator_)),
|
2011-04-12 21:38:58 +02:00
|
|
|
imm_(NULL),
|
2011-03-18 23:37:00 +01:00
|
|
|
logfile_(NULL),
|
2011-06-22 04:36:45 +02:00
|
|
|
logfile_number_(0),
|
2011-03-18 23:37:00 +01:00
|
|
|
log_(NULL),
|
2012-03-09 01:23:21 +01:00
|
|
|
tmp_batch_(new WriteBatch),
|
2011-03-18 23:37:00 +01:00
|
|
|
bg_compaction_scheduled_(false),
|
2012-08-27 09:50:26 +02:00
|
|
|
bg_logstats_scheduled_(false),
|
2012-08-22 02:02:21 +02:00
|
|
|
manual_compaction_(NULL),
|
2012-09-15 02:11:35 +02:00
|
|
|
logger_(NULL),
|
2012-10-16 17:53:46 +02:00
|
|
|
disable_delete_obsolete_files_(false),
|
2012-10-23 19:34:09 +02:00
|
|
|
delete_obsolete_files_last_run_(0),
|
|
|
|
stall_level0_slowdown_(0),
|
|
|
|
stall_memtable_compaction_(0),
|
|
|
|
stall_level0_num_files_(0),
|
|
|
|
started_at_(options.env->NowMicros()) {
|
2011-05-21 04:17:43 +02:00
|
|
|
mem_->Ref();
|
2011-04-12 21:38:58 +02:00
|
|
|
has_imm_.Release_Store(NULL);
|
|
|
|
|
2012-09-06 02:44:13 +02:00
|
|
|
env_->GetAbsolutePath(dbname, &db_absolute_path_);
|
2012-06-23 04:30:03 +02:00
|
|
|
stats_ = new CompactionStats[options.num_levels];
|
2011-03-18 23:37:00 +01:00
|
|
|
// Reserve ten files or so for other uses and give the rest to TableCache.
|
2012-08-22 20:43:53 +02:00
|
|
|
const int table_cache_size = options_.max_open_files - 10;
|
2011-03-18 23:37:00 +01:00
|
|
|
table_cache_ = new TableCache(dbname_, &options_, table_cache_size);
|
|
|
|
|
|
|
|
versions_ = new VersionSet(dbname_, &options_, table_cache_,
|
|
|
|
&internal_comparator_);
|
2012-08-15 00:20:36 +02:00
|
|
|
|
2012-09-18 22:18:32 +02:00
|
|
|
dumpLeveldbBuildVersion(options_.info_log);
|
2012-08-22 20:43:53 +02:00
|
|
|
options_.Dump(options_.info_log);
|
|
|
|
|
2012-08-15 00:20:36 +02:00
|
|
|
#ifdef USE_SCRIBE
|
|
|
|
logger_ = new ScribeLogger("localhost", 1456);
|
|
|
|
#endif
|
|
|
|
|
|
|
|
char name[100];
|
2012-09-12 18:54:22 +02:00
|
|
|
Status st = env_->GetHostName(name, 100L);
|
2012-08-15 00:20:36 +02:00
|
|
|
if(st.ok()) {
|
|
|
|
host_name_ = name;
|
|
|
|
} else {
|
|
|
|
Log(options_.info_log, "Can't get hostname, use localhost as host name.");
|
|
|
|
host_name_ = "localhost";
|
|
|
|
}
|
|
|
|
last_log_ts = 0;
|
2011-03-18 23:37:00 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
DBImpl::~DBImpl() {
|
|
|
|
// Wait for background work to finish
|
|
|
|
mutex_.Lock();
|
|
|
|
shutting_down_.Release_Store(this); // Any non-NULL value is ok
|
2012-08-27 09:50:26 +02:00
|
|
|
while (bg_compaction_scheduled_ || bg_logstats_scheduled_) {
|
2011-06-07 16:40:26 +02:00
|
|
|
bg_cv_.Wait();
|
2011-03-18 23:37:00 +01:00
|
|
|
}
|
|
|
|
mutex_.Unlock();
|
|
|
|
|
|
|
|
if (db_lock_ != NULL) {
|
|
|
|
env_->UnlockFile(db_lock_);
|
|
|
|
}
|
|
|
|
|
|
|
|
delete versions_;
|
2011-05-21 04:17:43 +02:00
|
|
|
if (mem_ != NULL) mem_->Unref();
|
|
|
|
if (imm_ != NULL) imm_->Unref();
|
2012-03-09 01:23:21 +01:00
|
|
|
delete tmp_batch_;
|
2011-03-18 23:37:00 +01:00
|
|
|
delete log_;
|
|
|
|
delete logfile_;
|
|
|
|
delete table_cache_;
|
2012-06-23 04:30:03 +02:00
|
|
|
delete[] stats_;
|
2011-03-18 23:37:00 +01:00
|
|
|
|
|
|
|
if (owns_info_log_) {
|
|
|
|
delete options_.info_log;
|
|
|
|
}
|
2011-04-12 21:38:58 +02:00
|
|
|
if (owns_cache_) {
|
|
|
|
delete options_.block_cache;
|
|
|
|
}
|
2012-08-27 09:50:26 +02:00
|
|
|
|
|
|
|
delete logger_;
|
2011-03-18 23:37:00 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
Status DBImpl::NewDB() {
|
2012-06-23 04:30:03 +02:00
|
|
|
VersionEdit new_db(NumberLevels());
|
2011-03-18 23:37:00 +01:00
|
|
|
new_db.SetComparatorName(user_comparator()->Name());
|
2011-04-12 21:38:58 +02:00
|
|
|
new_db.SetLogNumber(0);
|
2011-03-18 23:37:00 +01:00
|
|
|
new_db.SetNextFile(2);
|
|
|
|
new_db.SetLastSequence(0);
|
|
|
|
|
|
|
|
const std::string manifest = DescriptorFileName(dbname_, 1);
|
|
|
|
WritableFile* file;
|
|
|
|
Status s = env_->NewWritableFile(manifest, &file);
|
|
|
|
if (!s.ok()) {
|
|
|
|
return s;
|
|
|
|
}
|
|
|
|
{
|
|
|
|
log::Writer log(file);
|
|
|
|
std::string record;
|
|
|
|
new_db.EncodeTo(&record);
|
|
|
|
s = log.AddRecord(record);
|
|
|
|
if (s.ok()) {
|
|
|
|
s = file->Close();
|
|
|
|
}
|
|
|
|
}
|
|
|
|
delete file;
|
|
|
|
if (s.ok()) {
|
|
|
|
// Make "CURRENT" file that points to the new manifest file.
|
|
|
|
s = SetCurrentFile(env_, dbname_, 1);
|
|
|
|
} else {
|
|
|
|
env_->DeleteFile(manifest);
|
|
|
|
}
|
|
|
|
return s;
|
|
|
|
}
|
|
|
|
|
|
|
|
void DBImpl::MaybeIgnoreError(Status* s) const {
|
|
|
|
if (s->ok() || options_.paranoid_checks) {
|
|
|
|
// No change needed
|
|
|
|
} else {
|
2011-07-21 04:40:18 +02:00
|
|
|
Log(options_.info_log, "Ignoring error %s", s->ToString().c_str());
|
2011-03-18 23:37:00 +01:00
|
|
|
*s = Status::OK();
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2012-10-21 10:49:48 +02:00
|
|
|
// Returns the list of live files in 'live' and the list
|
|
|
|
// of all files in the filesystem in 'allfiles'.
|
|
|
|
void DBImpl::FindObsoleteFiles(DeletionState& deletion_state) {
|
|
|
|
mutex_.AssertHeld();
|
|
|
|
|
2012-09-15 02:11:35 +02:00
|
|
|
// if deletion is disabled, do nothing
|
|
|
|
if (disable_delete_obsolete_files_) {
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
2012-10-16 17:53:46 +02:00
|
|
|
// This method is costly when the number of files is large.
|
|
|
|
// Do not allow it to trigger more often than once in
|
|
|
|
// delete_obsolete_files_period_micros.
|
|
|
|
if (options_.delete_obsolete_files_period_micros != 0) {
|
|
|
|
const uint64_t now_micros = env_->NowMicros();
|
|
|
|
if (delete_obsolete_files_last_run_ +
|
|
|
|
options_.delete_obsolete_files_period_micros > now_micros) {
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
delete_obsolete_files_last_run_ = now_micros;
|
|
|
|
}
|
|
|
|
|
2011-03-18 23:37:00 +01:00
|
|
|
// Make a set of all of the live files
|
2012-10-21 10:49:48 +02:00
|
|
|
deletion_state.live = pending_outputs_;
|
|
|
|
versions_->AddLiveFiles(&deletion_state.live);
|
|
|
|
|
|
|
|
// set of all files in the directory
|
|
|
|
env_->GetChildren(dbname_, &deletion_state.allfiles); // Ignore errors
|
|
|
|
|
|
|
|
// store the current filenum, lognum, etc
|
|
|
|
deletion_state.filenumber = versions_->ManifestFileNumber();
|
|
|
|
deletion_state.lognumber = versions_->LogNumber();
|
|
|
|
deletion_state.prevlognumber = versions_->PrevLogNumber();
|
|
|
|
}
|
2011-03-18 23:37:00 +01:00
|
|
|
|
2012-10-21 10:49:48 +02:00
|
|
|
// Diffs the files listed in filenames and those that do not
|
|
|
|
// belong to live files are posibly removed. If the removed file
|
|
|
|
// is a sst file, then it returns the file number in files_to_evict.
|
|
|
|
// It is not necesary to hold the mutex when invoking this method.
|
|
|
|
void DBImpl::PurgeObsoleteFiles(DeletionState& state) {
|
2011-03-18 23:37:00 +01:00
|
|
|
uint64_t number;
|
|
|
|
FileType type;
|
2012-09-06 02:44:13 +02:00
|
|
|
std::vector<std::string> old_log_files;
|
2012-10-21 10:49:48 +02:00
|
|
|
for (size_t i = 0; i < state.allfiles.size(); i++) {
|
|
|
|
if (ParseFileName(state.allfiles[i], &number, &type)) {
|
2011-03-18 23:37:00 +01:00
|
|
|
bool keep = true;
|
|
|
|
switch (type) {
|
|
|
|
case kLogFile:
|
2012-10-21 10:49:48 +02:00
|
|
|
keep = ((number >= state.lognumber) ||
|
|
|
|
(number == state.prevlognumber));
|
2011-03-18 23:37:00 +01:00
|
|
|
break;
|
|
|
|
case kDescriptorFile:
|
|
|
|
// Keep my manifest file, and any newer incarnations'
|
|
|
|
// (in case there is a race that allows other incarnations)
|
2012-10-21 10:49:48 +02:00
|
|
|
keep = (number >= state.filenumber);
|
2011-03-18 23:37:00 +01:00
|
|
|
break;
|
|
|
|
case kTableFile:
|
2012-10-21 10:49:48 +02:00
|
|
|
keep = (state.live.find(number) != state.live.end());
|
2011-03-18 23:37:00 +01:00
|
|
|
break;
|
|
|
|
case kTempFile:
|
|
|
|
// Any temp files that are currently being written to must
|
|
|
|
// be recorded in pending_outputs_, which is inserted into "live"
|
2012-10-21 10:49:48 +02:00
|
|
|
keep = (state.live.find(number) != state.live.end());
|
2011-03-18 23:37:00 +01:00
|
|
|
break;
|
2012-08-18 01:06:05 +02:00
|
|
|
case kInfoLogFile:
|
|
|
|
keep = true;
|
|
|
|
if (number != 0) {
|
2012-10-21 10:49:48 +02:00
|
|
|
old_log_files.push_back(state.allfiles[i]);
|
2012-08-18 01:06:05 +02:00
|
|
|
}
|
|
|
|
break;
|
2011-03-18 23:37:00 +01:00
|
|
|
case kCurrentFile:
|
|
|
|
case kDBLockFile:
|
|
|
|
keep = true;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (!keep) {
|
|
|
|
if (type == kTableFile) {
|
2012-10-21 10:49:48 +02:00
|
|
|
// record the files to be evicted from the cache
|
|
|
|
state.files_to_evict.push_back(number);
|
2011-03-18 23:37:00 +01:00
|
|
|
}
|
2011-07-21 04:40:18 +02:00
|
|
|
Log(options_.info_log, "Delete type=%d #%lld\n",
|
2011-03-18 23:37:00 +01:00
|
|
|
int(type),
|
|
|
|
static_cast<unsigned long long>(number));
|
2012-10-21 10:49:48 +02:00
|
|
|
Status st = env_->DeleteFile(dbname_ + "/" + state.allfiles[i]);
|
2012-08-23 04:15:06 +02:00
|
|
|
if(!st.ok()) {
|
|
|
|
Log(options_.info_log, "Delete type=%d #%lld FAILED\n",
|
|
|
|
int(type),
|
|
|
|
static_cast<unsigned long long>(number));
|
|
|
|
}
|
2011-03-18 23:37:00 +01:00
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
2012-08-18 01:06:05 +02:00
|
|
|
|
|
|
|
// Delete old log files.
|
2012-09-06 02:44:13 +02:00
|
|
|
int old_log_file_count = old_log_files.size();
|
|
|
|
if (old_log_file_count >= KEEP_LOG_FILE_NUM &&
|
|
|
|
!options_.db_log_dir.empty()) {
|
|
|
|
std::sort(old_log_files.begin(), old_log_files.end());
|
2012-08-18 01:06:05 +02:00
|
|
|
for (int i = 0; i >= (old_log_file_count - KEEP_LOG_FILE_NUM); i++) {
|
2012-09-06 02:44:13 +02:00
|
|
|
std::string& to_delete = old_log_files.at(i);
|
|
|
|
Log(options_.info_log, "Delete type=%d %s\n",
|
|
|
|
int(kInfoLogFile), to_delete.c_str());
|
2012-08-18 01:06:05 +02:00
|
|
|
env_->DeleteFile(dbname_ + "/" + to_delete);
|
|
|
|
}
|
|
|
|
}
|
2011-03-18 23:37:00 +01:00
|
|
|
}
|
|
|
|
|
2012-10-21 10:49:48 +02:00
|
|
|
void DBImpl::EvictObsoleteFiles(DeletionState& state) {
|
|
|
|
mutex_.AssertHeld();
|
|
|
|
for (unsigned int i = 0; i < state.files_to_evict.size(); i++) {
|
|
|
|
table_cache_->Evict(state.files_to_evict[i]);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
void DBImpl::DeleteObsoleteFiles() {
|
|
|
|
mutex_.AssertHeld();
|
|
|
|
DeletionState deletion_state;
|
|
|
|
std::set<uint64_t> live;
|
|
|
|
std::vector<std::string> allfiles;
|
|
|
|
std::vector<uint64_t> files_to_evict;
|
|
|
|
uint64_t filenumber, lognumber, prevlognumber;
|
|
|
|
FindObsoleteFiles(deletion_state);
|
|
|
|
PurgeObsoleteFiles(deletion_state);
|
|
|
|
EvictObsoleteFiles(deletion_state);
|
|
|
|
}
|
|
|
|
|
2011-03-18 23:37:00 +01:00
|
|
|
Status DBImpl::Recover(VersionEdit* edit) {
|
|
|
|
mutex_.AssertHeld();
|
|
|
|
|
|
|
|
// Ignore error from CreateDir since the creation of the DB is
|
|
|
|
// committed only when the descriptor is created, and this directory
|
|
|
|
// may already exist from a previous failed creation attempt.
|
|
|
|
env_->CreateDir(dbname_);
|
|
|
|
assert(db_lock_ == NULL);
|
|
|
|
Status s = env_->LockFile(LockFileName(dbname_), &db_lock_);
|
|
|
|
if (!s.ok()) {
|
|
|
|
return s;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (!env_->FileExists(CurrentFileName(dbname_))) {
|
|
|
|
if (options_.create_if_missing) {
|
|
|
|
s = NewDB();
|
|
|
|
if (!s.ok()) {
|
|
|
|
return s;
|
|
|
|
}
|
|
|
|
} else {
|
|
|
|
return Status::InvalidArgument(
|
|
|
|
dbname_, "does not exist (create_if_missing is false)");
|
|
|
|
}
|
|
|
|
} else {
|
|
|
|
if (options_.error_if_exists) {
|
|
|
|
return Status::InvalidArgument(
|
|
|
|
dbname_, "exists (error_if_exists is true)");
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2011-04-12 21:38:58 +02:00
|
|
|
s = versions_->Recover();
|
2011-03-18 23:37:00 +01:00
|
|
|
if (s.ok()) {
|
|
|
|
SequenceNumber max_sequence(0);
|
2011-06-22 04:36:45 +02:00
|
|
|
|
|
|
|
// Recover from all newer log files than the ones named in the
|
|
|
|
// descriptor (new log files may have been added by the previous
|
|
|
|
// incarnation without registering them in the descriptor).
|
|
|
|
//
|
|
|
|
// Note that PrevLogNumber() is no longer used, but we pay
|
|
|
|
// attention to it in case we are recovering a database
|
|
|
|
// produced by an older version of leveldb.
|
|
|
|
const uint64_t min_log = versions_->LogNumber();
|
|
|
|
const uint64_t prev_log = versions_->PrevLogNumber();
|
|
|
|
std::vector<std::string> filenames;
|
|
|
|
s = env_->GetChildren(dbname_, &filenames);
|
|
|
|
if (!s.ok()) {
|
|
|
|
return s;
|
2011-04-12 21:38:58 +02:00
|
|
|
}
|
2011-06-22 04:36:45 +02:00
|
|
|
uint64_t number;
|
|
|
|
FileType type;
|
|
|
|
std::vector<uint64_t> logs;
|
|
|
|
for (size_t i = 0; i < filenames.size(); i++) {
|
|
|
|
if (ParseFileName(filenames[i], &number, &type)
|
|
|
|
&& type == kLogFile
|
|
|
|
&& ((number >= min_log) || (number == prev_log))) {
|
|
|
|
logs.push_back(number);
|
|
|
|
}
|
2011-03-18 23:37:00 +01:00
|
|
|
}
|
2011-06-22 04:36:45 +02:00
|
|
|
|
|
|
|
// Recover in the order in which the logs were generated
|
|
|
|
std::sort(logs.begin(), logs.end());
|
|
|
|
for (size_t i = 0; i < logs.size(); i++) {
|
|
|
|
s = RecoverLogFile(logs[i], edit, &max_sequence);
|
2011-09-01 21:08:02 +02:00
|
|
|
|
|
|
|
// The previous incarnation may not have written any MANIFEST
|
|
|
|
// records after allocating this log number. So we manually
|
|
|
|
// update the file number allocation counter in VersionSet.
|
|
|
|
versions_->MarkFileNumberUsed(logs[i]);
|
2011-06-22 04:36:45 +02:00
|
|
|
}
|
|
|
|
|
2011-03-18 23:37:00 +01:00
|
|
|
if (s.ok()) {
|
2011-04-12 21:38:58 +02:00
|
|
|
if (versions_->LastSequence() < max_sequence) {
|
|
|
|
versions_->SetLastSequence(max_sequence);
|
|
|
|
}
|
2011-03-18 23:37:00 +01:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
return s;
|
|
|
|
}
|
|
|
|
|
|
|
|
Status DBImpl::RecoverLogFile(uint64_t log_number,
|
|
|
|
VersionEdit* edit,
|
|
|
|
SequenceNumber* max_sequence) {
|
|
|
|
struct LogReporter : public log::Reader::Reporter {
|
|
|
|
Env* env;
|
2011-07-21 04:40:18 +02:00
|
|
|
Logger* info_log;
|
2011-03-18 23:37:00 +01:00
|
|
|
const char* fname;
|
|
|
|
Status* status; // NULL if options_.paranoid_checks==false
|
|
|
|
virtual void Corruption(size_t bytes, const Status& s) {
|
2011-07-21 04:40:18 +02:00
|
|
|
Log(info_log, "%s%s: dropping %d bytes; %s",
|
2011-03-18 23:37:00 +01:00
|
|
|
(this->status == NULL ? "(ignoring error) " : ""),
|
|
|
|
fname, static_cast<int>(bytes), s.ToString().c_str());
|
|
|
|
if (this->status != NULL && this->status->ok()) *this->status = s;
|
|
|
|
}
|
|
|
|
};
|
|
|
|
|
|
|
|
mutex_.AssertHeld();
|
|
|
|
|
|
|
|
// Open the log file
|
|
|
|
std::string fname = LogFileName(dbname_, log_number);
|
|
|
|
SequentialFile* file;
|
|
|
|
Status status = env_->NewSequentialFile(fname, &file);
|
|
|
|
if (!status.ok()) {
|
|
|
|
MaybeIgnoreError(&status);
|
|
|
|
return status;
|
|
|
|
}
|
|
|
|
|
|
|
|
// Create the log reader.
|
|
|
|
LogReporter reporter;
|
|
|
|
reporter.env = env_;
|
|
|
|
reporter.info_log = options_.info_log;
|
|
|
|
reporter.fname = fname.c_str();
|
|
|
|
reporter.status = (options_.paranoid_checks ? &status : NULL);
|
|
|
|
// We intentially make log::Reader do checksumming even if
|
|
|
|
// paranoid_checks==false so that corruptions cause entire commits
|
|
|
|
// to be skipped instead of propagating bad information (like overly
|
|
|
|
// large sequence numbers).
|
2011-05-21 04:17:43 +02:00
|
|
|
log::Reader reader(file, &reporter, true/*checksum*/,
|
|
|
|
0/*initial_offset*/);
|
2011-07-21 04:40:18 +02:00
|
|
|
Log(options_.info_log, "Recovering log #%llu",
|
2011-03-18 23:37:00 +01:00
|
|
|
(unsigned long long) log_number);
|
|
|
|
|
|
|
|
// Read all the records and add to a memtable
|
|
|
|
std::string scratch;
|
|
|
|
Slice record;
|
|
|
|
WriteBatch batch;
|
|
|
|
MemTable* mem = NULL;
|
|
|
|
while (reader.ReadRecord(&record, &scratch) &&
|
|
|
|
status.ok()) {
|
|
|
|
if (record.size() < 12) {
|
|
|
|
reporter.Corruption(
|
|
|
|
record.size(), Status::Corruption("log record too small"));
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
WriteBatchInternal::SetContents(&batch, record);
|
|
|
|
|
|
|
|
if (mem == NULL) {
|
|
|
|
mem = new MemTable(internal_comparator_);
|
2011-05-21 04:17:43 +02:00
|
|
|
mem->Ref();
|
2011-03-18 23:37:00 +01:00
|
|
|
}
|
|
|
|
status = WriteBatchInternal::InsertInto(&batch, mem);
|
|
|
|
MaybeIgnoreError(&status);
|
|
|
|
if (!status.ok()) {
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
const SequenceNumber last_seq =
|
|
|
|
WriteBatchInternal::Sequence(&batch) +
|
|
|
|
WriteBatchInternal::Count(&batch) - 1;
|
|
|
|
if (last_seq > *max_sequence) {
|
|
|
|
*max_sequence = last_seq;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (mem->ApproximateMemoryUsage() > options_.write_buffer_size) {
|
2011-06-22 04:36:45 +02:00
|
|
|
status = WriteLevel0Table(mem, edit, NULL);
|
2011-03-18 23:37:00 +01:00
|
|
|
if (!status.ok()) {
|
|
|
|
// Reflect errors immediately so that conditions like full
|
|
|
|
// file-systems cause the DB::Open() to fail.
|
|
|
|
break;
|
|
|
|
}
|
2011-05-21 04:17:43 +02:00
|
|
|
mem->Unref();
|
2011-03-18 23:37:00 +01:00
|
|
|
mem = NULL;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
if (status.ok() && mem != NULL) {
|
2011-06-22 04:36:45 +02:00
|
|
|
status = WriteLevel0Table(mem, edit, NULL);
|
2011-03-18 23:37:00 +01:00
|
|
|
// Reflect errors immediately so that conditions like full
|
|
|
|
// file-systems cause the DB::Open() to fail.
|
|
|
|
}
|
|
|
|
|
2011-05-21 04:17:43 +02:00
|
|
|
if (mem != NULL) mem->Unref();
|
2011-03-18 23:37:00 +01:00
|
|
|
delete file;
|
|
|
|
return status;
|
|
|
|
}
|
|
|
|
|
2011-06-22 04:36:45 +02:00
|
|
|
Status DBImpl::WriteLevel0Table(MemTable* mem, VersionEdit* edit,
|
|
|
|
Version* base) {
|
2011-03-18 23:37:00 +01:00
|
|
|
mutex_.AssertHeld();
|
2011-04-12 21:38:58 +02:00
|
|
|
const uint64_t start_micros = env_->NowMicros();
|
2011-03-18 23:37:00 +01:00
|
|
|
FileMetaData meta;
|
|
|
|
meta.number = versions_->NewFileNumber();
|
|
|
|
pending_outputs_.insert(meta.number);
|
|
|
|
Iterator* iter = mem->NewIterator();
|
2011-07-21 04:40:18 +02:00
|
|
|
Log(options_.info_log, "Level-0 table #%llu: started",
|
2011-03-18 23:37:00 +01:00
|
|
|
(unsigned long long) meta.number);
|
2011-04-12 21:38:58 +02:00
|
|
|
|
|
|
|
Status s;
|
|
|
|
{
|
|
|
|
mutex_.Unlock();
|
2011-06-22 04:36:45 +02:00
|
|
|
s = BuildTable(dbname_, env_, options_, table_cache_, iter, &meta);
|
2011-04-12 21:38:58 +02:00
|
|
|
mutex_.Lock();
|
|
|
|
}
|
|
|
|
|
2011-07-21 04:40:18 +02:00
|
|
|
Log(options_.info_log, "Level-0 table #%llu: %lld bytes %s",
|
2011-03-18 23:37:00 +01:00
|
|
|
(unsigned long long) meta.number,
|
|
|
|
(unsigned long long) meta.file_size,
|
|
|
|
s.ToString().c_str());
|
|
|
|
delete iter;
|
|
|
|
pending_outputs_.erase(meta.number);
|
2011-04-12 21:38:58 +02:00
|
|
|
|
2011-06-22 04:36:45 +02:00
|
|
|
|
|
|
|
// Note that if file_size is zero, the file has been deleted and
|
|
|
|
// should not be added to the manifest.
|
|
|
|
int level = 0;
|
|
|
|
if (s.ok() && meta.file_size > 0) {
|
2011-07-15 02:20:57 +02:00
|
|
|
const Slice min_user_key = meta.smallest.user_key();
|
|
|
|
const Slice max_user_key = meta.largest.user_key();
|
2011-10-06 01:30:28 +02:00
|
|
|
if (base != NULL) {
|
|
|
|
level = base->PickLevelForMemTableOutput(min_user_key, max_user_key);
|
2011-06-22 04:36:45 +02:00
|
|
|
}
|
|
|
|
edit->AddFile(level, meta.number, meta.file_size,
|
|
|
|
meta.smallest, meta.largest);
|
|
|
|
}
|
|
|
|
|
2011-04-12 21:38:58 +02:00
|
|
|
CompactionStats stats;
|
|
|
|
stats.micros = env_->NowMicros() - start_micros;
|
|
|
|
stats.bytes_written = meta.file_size;
|
2012-10-23 19:34:09 +02:00
|
|
|
stats.files_out_levelnp1 = 1;
|
2011-06-22 04:36:45 +02:00
|
|
|
stats_[level].Add(stats);
|
2011-03-18 23:37:00 +01:00
|
|
|
return s;
|
|
|
|
}
|
|
|
|
|
|
|
|
Status DBImpl::CompactMemTable() {
|
|
|
|
mutex_.AssertHeld();
|
2011-04-12 21:38:58 +02:00
|
|
|
assert(imm_ != NULL);
|
2011-03-18 23:37:00 +01:00
|
|
|
|
|
|
|
// Save the contents of the memtable as a new Table
|
2012-06-23 04:30:03 +02:00
|
|
|
VersionEdit edit(NumberLevels());
|
2011-06-22 04:36:45 +02:00
|
|
|
Version* base = versions_->current();
|
|
|
|
base->Ref();
|
|
|
|
Status s = WriteLevel0Table(imm_, &edit, base);
|
|
|
|
base->Unref();
|
|
|
|
|
|
|
|
if (s.ok() && shutting_down_.Acquire_Load()) {
|
|
|
|
s = Status::IOError("Deleting DB during memtable compaction");
|
|
|
|
}
|
2011-03-18 23:37:00 +01:00
|
|
|
|
2011-04-12 21:38:58 +02:00
|
|
|
// Replace immutable memtable with the generated Table
|
2011-03-18 23:37:00 +01:00
|
|
|
if (s.ok()) {
|
2011-04-12 21:38:58 +02:00
|
|
|
edit.SetPrevLogNumber(0);
|
2011-06-22 04:36:45 +02:00
|
|
|
edit.SetLogNumber(logfile_number_); // Earlier logs no longer needed
|
2011-09-01 21:08:02 +02:00
|
|
|
s = versions_->LogAndApply(&edit, &mutex_);
|
2011-03-18 23:37:00 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
if (s.ok()) {
|
|
|
|
// Commit to the new state
|
2011-05-21 04:17:43 +02:00
|
|
|
imm_->Unref();
|
2011-04-12 21:38:58 +02:00
|
|
|
imm_ = NULL;
|
|
|
|
has_imm_.Release_Store(NULL);
|
2012-08-15 00:20:36 +02:00
|
|
|
MaybeScheduleLogDBDeployStats();
|
2012-10-21 10:49:48 +02:00
|
|
|
// we could have deleted obsolete files here, but it is not
|
|
|
|
// absolutely necessary because it could be also done as part
|
|
|
|
// of other background compaction
|
2011-03-18 23:37:00 +01:00
|
|
|
}
|
2011-04-12 21:38:58 +02:00
|
|
|
|
2011-03-18 23:37:00 +01:00
|
|
|
return s;
|
|
|
|
}
|
|
|
|
|
2011-10-06 01:30:28 +02:00
|
|
|
void DBImpl::CompactRange(const Slice* begin, const Slice* end) {
|
|
|
|
int max_level_with_files = 1;
|
|
|
|
{
|
|
|
|
MutexLock l(&mutex_);
|
|
|
|
Version* base = versions_->current();
|
2012-06-23 04:30:03 +02:00
|
|
|
for (int level = 1; level < NumberLevels(); level++) {
|
2011-10-06 01:30:28 +02:00
|
|
|
if (base->OverlapInLevel(level, begin, end)) {
|
|
|
|
max_level_with_files = level;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
TEST_CompactMemTable(); // TODO(sanjay): Skip if memtable does not overlap
|
|
|
|
for (int level = 0; level < max_level_with_files; level++) {
|
|
|
|
TEST_CompactRange(level, begin, end);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2012-06-23 04:30:03 +02:00
|
|
|
int DBImpl::NumberLevels() {
|
2012-08-15 00:20:36 +02:00
|
|
|
return options_.num_levels;
|
2012-06-23 04:30:03 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
int DBImpl::MaxMemCompactionLevel() {
|
2012-08-15 00:20:36 +02:00
|
|
|
return options_.max_mem_compaction_level;
|
2012-06-23 04:30:03 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
int DBImpl::Level0StopWriteTrigger() {
|
2012-08-15 00:20:36 +02:00
|
|
|
return options_.level0_stop_writes_trigger;
|
2012-06-23 04:30:03 +02:00
|
|
|
}
|
|
|
|
|
2012-07-06 20:42:09 +02:00
|
|
|
Status DBImpl::Flush(const FlushOptions& options) {
|
|
|
|
Status status = FlushMemTable(options);
|
|
|
|
return status;
|
|
|
|
}
|
|
|
|
|
2011-10-06 01:30:28 +02:00
|
|
|
void DBImpl::TEST_CompactRange(int level, const Slice* begin,const Slice* end) {
|
2011-06-22 04:36:45 +02:00
|
|
|
assert(level >= 0);
|
|
|
|
|
2011-10-06 01:30:28 +02:00
|
|
|
InternalKey begin_storage, end_storage;
|
|
|
|
|
2011-06-07 16:40:26 +02:00
|
|
|
ManualCompaction manual;
|
|
|
|
manual.level = level;
|
2011-10-06 01:30:28 +02:00
|
|
|
manual.done = false;
|
|
|
|
if (begin == NULL) {
|
|
|
|
manual.begin = NULL;
|
|
|
|
} else {
|
|
|
|
begin_storage = InternalKey(*begin, kMaxSequenceNumber, kValueTypeForSeek);
|
|
|
|
manual.begin = &begin_storage;
|
|
|
|
}
|
|
|
|
if (end == NULL) {
|
|
|
|
manual.end = NULL;
|
|
|
|
} else {
|
|
|
|
end_storage = InternalKey(*end, 0, static_cast<ValueType>(0));
|
|
|
|
manual.end = &end_storage;
|
|
|
|
}
|
|
|
|
|
|
|
|
MutexLock l(&mutex_);
|
|
|
|
while (!manual.done) {
|
|
|
|
while (manual_compaction_ != NULL) {
|
|
|
|
bg_cv_.Wait();
|
|
|
|
}
|
|
|
|
manual_compaction_ = &manual;
|
|
|
|
MaybeScheduleCompaction();
|
|
|
|
while (manual_compaction_ == &manual) {
|
|
|
|
bg_cv_.Wait();
|
|
|
|
}
|
2011-06-07 16:40:26 +02:00
|
|
|
}
|
2011-03-18 23:37:00 +01:00
|
|
|
}
|
|
|
|
|
2012-07-06 20:42:09 +02:00
|
|
|
Status DBImpl::FlushMemTable(const FlushOptions& options) {
|
2012-03-09 01:23:21 +01:00
|
|
|
// NULL batch means just wait for earlier writes to be done
|
|
|
|
Status s = Write(WriteOptions(), NULL);
|
2012-07-06 20:42:09 +02:00
|
|
|
if (s.ok() && options.wait) {
|
2011-04-12 21:38:58 +02:00
|
|
|
// Wait until the compaction completes
|
2012-07-06 20:42:09 +02:00
|
|
|
s = WaitForCompactMemTable();
|
2011-04-12 21:38:58 +02:00
|
|
|
}
|
|
|
|
return s;
|
2011-03-18 23:37:00 +01:00
|
|
|
}
|
|
|
|
|
2012-07-06 20:42:09 +02:00
|
|
|
Status DBImpl::WaitForCompactMemTable() {
|
2012-08-15 00:20:36 +02:00
|
|
|
Status s;
|
|
|
|
// Wait until the compaction completes
|
|
|
|
MutexLock l(&mutex_);
|
|
|
|
while (imm_ != NULL && bg_error_.ok()) {
|
|
|
|
bg_cv_.Wait();
|
|
|
|
}
|
|
|
|
if (imm_ != NULL) {
|
|
|
|
s = bg_error_;
|
|
|
|
}
|
|
|
|
return s;
|
2012-07-06 20:42:09 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
Status DBImpl::TEST_CompactMemTable() {
|
|
|
|
return FlushMemTable(FlushOptions());
|
|
|
|
}
|
|
|
|
|
2012-06-23 04:30:03 +02:00
|
|
|
Status DBImpl::TEST_WaitForCompactMemTable() {
|
2012-08-15 00:20:36 +02:00
|
|
|
return WaitForCompactMemTable();
|
2012-06-23 04:30:03 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
Status DBImpl::TEST_WaitForCompact() {
|
2012-08-15 00:20:36 +02:00
|
|
|
// Wait until the compaction completes
|
|
|
|
MutexLock l(&mutex_);
|
|
|
|
while (bg_compaction_scheduled_ && bg_error_.ok()) {
|
|
|
|
bg_cv_.Wait();
|
|
|
|
}
|
|
|
|
return bg_error_;
|
2012-06-23 04:30:03 +02:00
|
|
|
}
|
|
|
|
|
2011-03-18 23:37:00 +01:00
|
|
|
void DBImpl::MaybeScheduleCompaction() {
|
|
|
|
mutex_.AssertHeld();
|
|
|
|
if (bg_compaction_scheduled_) {
|
|
|
|
// Already scheduled
|
|
|
|
} else if (shutting_down_.Acquire_Load()) {
|
|
|
|
// DB is being deleted; no more background compactions
|
2011-06-07 16:40:26 +02:00
|
|
|
} else if (imm_ == NULL &&
|
|
|
|
manual_compaction_ == NULL &&
|
|
|
|
!versions_->NeedsCompaction()) {
|
2011-03-18 23:37:00 +01:00
|
|
|
// No work to be done
|
|
|
|
} else {
|
|
|
|
bg_compaction_scheduled_ = true;
|
|
|
|
env_->Schedule(&DBImpl::BGWork, this);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
void DBImpl::BGWork(void* db) {
|
|
|
|
reinterpret_cast<DBImpl*>(db)->BackgroundCall();
|
|
|
|
}
|
|
|
|
|
|
|
|
void DBImpl::BackgroundCall() {
|
2012-10-21 10:49:48 +02:00
|
|
|
DeletionState deletion_state;
|
2011-03-18 23:37:00 +01:00
|
|
|
MutexLock l(&mutex_);
|
|
|
|
assert(bg_compaction_scheduled_);
|
2011-06-07 16:40:26 +02:00
|
|
|
if (!shutting_down_.Acquire_Load()) {
|
2012-10-21 10:49:48 +02:00
|
|
|
Status s = BackgroundCompaction(deletion_state);
|
2012-08-23 01:57:51 +02:00
|
|
|
if (!s.ok()) {
|
|
|
|
// Wait a little bit before retrying background compaction in
|
|
|
|
// case this is an environmental problem and we do not want to
|
|
|
|
// chew up resources for failed compactions for the duration of
|
|
|
|
// the problem.
|
|
|
|
bg_cv_.SignalAll(); // In case a waiter can proceed despite the error
|
|
|
|
Log(options_.info_log, "Waiting after background compaction error: %s",
|
|
|
|
s.ToString().c_str());
|
|
|
|
mutex_.Unlock();
|
|
|
|
env_->SleepForMicroseconds(1000000);
|
|
|
|
mutex_.Lock();
|
|
|
|
}
|
2011-03-18 23:37:00 +01:00
|
|
|
}
|
2012-08-23 01:57:51 +02:00
|
|
|
|
2012-10-21 10:49:48 +02:00
|
|
|
// delete unnecessary files if any, this is done outside the mutex
|
|
|
|
if (!deletion_state.live.empty()) {
|
|
|
|
mutex_.Unlock();
|
|
|
|
PurgeObsoleteFiles(deletion_state);
|
|
|
|
mutex_.Lock();
|
|
|
|
EvictObsoleteFiles(deletion_state);
|
|
|
|
}
|
|
|
|
|
2011-03-18 23:37:00 +01:00
|
|
|
bg_compaction_scheduled_ = false;
|
|
|
|
|
2012-08-15 00:20:36 +02:00
|
|
|
MaybeScheduleLogDBDeployStats();
|
|
|
|
|
2011-03-18 23:37:00 +01:00
|
|
|
// Previous compaction may have produced too many files in a level,
|
|
|
|
// so reschedule another compaction if needed.
|
|
|
|
MaybeScheduleCompaction();
|
2011-06-07 16:40:26 +02:00
|
|
|
bg_cv_.SignalAll();
|
2011-03-18 23:37:00 +01:00
|
|
|
}
|
|
|
|
|
2012-10-21 10:49:48 +02:00
|
|
|
Status DBImpl::BackgroundCompaction(DeletionState& deletion_state) {
|
2011-03-18 23:37:00 +01:00
|
|
|
mutex_.AssertHeld();
|
2011-04-12 21:38:58 +02:00
|
|
|
|
|
|
|
if (imm_ != NULL) {
|
2012-08-23 01:57:51 +02:00
|
|
|
return CompactMemTable();
|
2011-04-12 21:38:58 +02:00
|
|
|
}
|
|
|
|
|
2011-06-07 16:40:26 +02:00
|
|
|
Compaction* c;
|
|
|
|
bool is_manual = (manual_compaction_ != NULL);
|
2011-10-06 01:30:28 +02:00
|
|
|
InternalKey manual_end;
|
2011-06-07 16:40:26 +02:00
|
|
|
if (is_manual) {
|
2011-10-06 01:30:28 +02:00
|
|
|
ManualCompaction* m = manual_compaction_;
|
|
|
|
c = versions_->CompactRange(m->level, m->begin, m->end);
|
|
|
|
m->done = (c == NULL);
|
|
|
|
if (c != NULL) {
|
|
|
|
manual_end = c->input(0, c->num_input_files(0) - 1)->largest;
|
|
|
|
}
|
|
|
|
Log(options_.info_log,
|
|
|
|
"Manual compaction at level-%d from %s .. %s; will stop at %s\n",
|
2011-06-07 16:40:26 +02:00
|
|
|
m->level,
|
2011-10-06 01:30:28 +02:00
|
|
|
(m->begin ? m->begin->DebugString().c_str() : "(begin)"),
|
|
|
|
(m->end ? m->end->DebugString().c_str() : "(end)"),
|
|
|
|
(m->done ? "(end)" : manual_end.DebugString().c_str()));
|
2011-06-07 16:40:26 +02:00
|
|
|
} else {
|
|
|
|
c = versions_->PickCompaction();
|
2011-03-18 23:37:00 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
Status status;
|
2011-06-07 16:40:26 +02:00
|
|
|
if (c == NULL) {
|
|
|
|
// Nothing to do
|
|
|
|
} else if (!is_manual && c->IsTrivialMove()) {
|
2011-03-18 23:37:00 +01:00
|
|
|
// Move file to next level
|
2011-03-22 19:32:49 +01:00
|
|
|
assert(c->num_input_files(0) == 1);
|
2011-03-18 23:37:00 +01:00
|
|
|
FileMetaData* f = c->input(0, 0);
|
|
|
|
c->edit()->DeleteFile(c->level(), f->number);
|
|
|
|
c->edit()->AddFile(c->level() + 1, f->number, f->file_size,
|
|
|
|
f->smallest, f->largest);
|
2011-09-01 21:08:02 +02:00
|
|
|
status = versions_->LogAndApply(c->edit(), &mutex_);
|
2011-06-07 16:40:26 +02:00
|
|
|
VersionSet::LevelSummaryStorage tmp;
|
2011-07-21 04:40:18 +02:00
|
|
|
Log(options_.info_log, "Moved #%lld to level-%d %lld bytes %s: %s\n",
|
2011-03-18 23:37:00 +01:00
|
|
|
static_cast<unsigned long long>(f->number),
|
|
|
|
c->level() + 1,
|
|
|
|
static_cast<unsigned long long>(f->file_size),
|
2011-06-07 16:40:26 +02:00
|
|
|
status.ToString().c_str(),
|
|
|
|
versions_->LevelSummary(&tmp));
|
2011-03-18 23:37:00 +01:00
|
|
|
} else {
|
|
|
|
CompactionState* compact = new CompactionState(c);
|
|
|
|
status = DoCompactionWork(compact);
|
|
|
|
CleanupCompaction(compact);
|
2012-01-25 23:56:52 +01:00
|
|
|
c->ReleaseInputs();
|
2012-10-21 10:49:48 +02:00
|
|
|
FindObsoleteFiles(deletion_state);
|
2011-03-18 23:37:00 +01:00
|
|
|
}
|
|
|
|
delete c;
|
|
|
|
|
|
|
|
if (status.ok()) {
|
|
|
|
// Done
|
|
|
|
} else if (shutting_down_.Acquire_Load()) {
|
|
|
|
// Ignore compaction errors found during shutting down
|
|
|
|
} else {
|
2011-07-21 04:40:18 +02:00
|
|
|
Log(options_.info_log,
|
2011-03-18 23:37:00 +01:00
|
|
|
"Compaction error: %s", status.ToString().c_str());
|
|
|
|
if (options_.paranoid_checks && bg_error_.ok()) {
|
|
|
|
bg_error_ = status;
|
|
|
|
}
|
|
|
|
}
|
2011-06-07 16:40:26 +02:00
|
|
|
|
|
|
|
if (is_manual) {
|
2011-10-06 01:30:28 +02:00
|
|
|
ManualCompaction* m = manual_compaction_;
|
2012-01-25 23:56:52 +01:00
|
|
|
if (!status.ok()) {
|
|
|
|
m->done = true;
|
|
|
|
}
|
2011-10-06 01:30:28 +02:00
|
|
|
if (!m->done) {
|
|
|
|
// We only compacted part of the requested range. Update *m
|
|
|
|
// to the range that is left to be compacted.
|
|
|
|
m->tmp_storage = manual_end;
|
|
|
|
m->begin = &m->tmp_storage;
|
|
|
|
}
|
2011-06-07 16:40:26 +02:00
|
|
|
manual_compaction_ = NULL;
|
|
|
|
}
|
2012-08-23 01:57:51 +02:00
|
|
|
return status;
|
2011-03-18 23:37:00 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
void DBImpl::CleanupCompaction(CompactionState* compact) {
|
|
|
|
mutex_.AssertHeld();
|
|
|
|
if (compact->builder != NULL) {
|
|
|
|
// May happen if we get a shutdown call in the middle of compaction
|
|
|
|
compact->builder->Abandon();
|
|
|
|
delete compact->builder;
|
|
|
|
} else {
|
|
|
|
assert(compact->outfile == NULL);
|
|
|
|
}
|
|
|
|
delete compact->outfile;
|
2011-04-21 00:48:11 +02:00
|
|
|
for (size_t i = 0; i < compact->outputs.size(); i++) {
|
2011-03-18 23:37:00 +01:00
|
|
|
const CompactionState::Output& out = compact->outputs[i];
|
|
|
|
pending_outputs_.erase(out.number);
|
|
|
|
}
|
|
|
|
delete compact;
|
|
|
|
}
|
|
|
|
|
|
|
|
Status DBImpl::OpenCompactionOutputFile(CompactionState* compact) {
|
|
|
|
assert(compact != NULL);
|
|
|
|
assert(compact->builder == NULL);
|
|
|
|
uint64_t file_number;
|
|
|
|
{
|
|
|
|
mutex_.Lock();
|
|
|
|
file_number = versions_->NewFileNumber();
|
|
|
|
pending_outputs_.insert(file_number);
|
|
|
|
CompactionState::Output out;
|
|
|
|
out.number = file_number;
|
|
|
|
out.smallest.Clear();
|
|
|
|
out.largest.Clear();
|
|
|
|
compact->outputs.push_back(out);
|
|
|
|
mutex_.Unlock();
|
|
|
|
}
|
|
|
|
|
|
|
|
// Make the output file
|
|
|
|
std::string fname = TableFileName(dbname_, file_number);
|
|
|
|
Status s = env_->NewWritableFile(fname, &compact->outfile);
|
|
|
|
if (s.ok()) {
|
|
|
|
compact->builder = new TableBuilder(options_, compact->outfile);
|
|
|
|
}
|
|
|
|
return s;
|
|
|
|
}
|
|
|
|
|
|
|
|
Status DBImpl::FinishCompactionOutputFile(CompactionState* compact,
|
|
|
|
Iterator* input) {
|
|
|
|
assert(compact != NULL);
|
|
|
|
assert(compact->outfile != NULL);
|
|
|
|
assert(compact->builder != NULL);
|
|
|
|
|
|
|
|
const uint64_t output_number = compact->current_output()->number;
|
|
|
|
assert(output_number != 0);
|
|
|
|
|
|
|
|
// Check for iterator errors
|
|
|
|
Status s = input->status();
|
|
|
|
const uint64_t current_entries = compact->builder->NumEntries();
|
|
|
|
if (s.ok()) {
|
|
|
|
s = compact->builder->Finish();
|
|
|
|
} else {
|
|
|
|
compact->builder->Abandon();
|
|
|
|
}
|
|
|
|
const uint64_t current_bytes = compact->builder->FileSize();
|
|
|
|
compact->current_output()->file_size = current_bytes;
|
|
|
|
compact->total_bytes += current_bytes;
|
|
|
|
delete compact->builder;
|
|
|
|
compact->builder = NULL;
|
|
|
|
|
|
|
|
// Finish and check for file errors
|
2012-08-04 00:20:58 +02:00
|
|
|
if (s.ok() && !options_.disableDataSync) {
|
2012-08-27 21:10:26 +02:00
|
|
|
if (options_.use_fsync) {
|
|
|
|
s = compact->outfile->Fsync();
|
|
|
|
} else {
|
|
|
|
s = compact->outfile->Sync();
|
|
|
|
}
|
2011-03-18 23:37:00 +01:00
|
|
|
}
|
|
|
|
if (s.ok()) {
|
|
|
|
s = compact->outfile->Close();
|
|
|
|
}
|
|
|
|
delete compact->outfile;
|
|
|
|
compact->outfile = NULL;
|
|
|
|
|
|
|
|
if (s.ok() && current_entries > 0) {
|
|
|
|
// Verify that the table is usable
|
2011-03-28 22:43:44 +02:00
|
|
|
Iterator* iter = table_cache_->NewIterator(ReadOptions(),
|
|
|
|
output_number,
|
|
|
|
current_bytes);
|
2011-03-18 23:37:00 +01:00
|
|
|
s = iter->status();
|
|
|
|
delete iter;
|
|
|
|
if (s.ok()) {
|
2011-07-21 04:40:18 +02:00
|
|
|
Log(options_.info_log,
|
2011-03-18 23:37:00 +01:00
|
|
|
"Generated table #%llu: %lld keys, %lld bytes",
|
|
|
|
(unsigned long long) output_number,
|
|
|
|
(unsigned long long) current_entries,
|
|
|
|
(unsigned long long) current_bytes);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
return s;
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
Status DBImpl::InstallCompactionResults(CompactionState* compact) {
|
|
|
|
mutex_.AssertHeld();
|
2011-07-21 04:40:18 +02:00
|
|
|
Log(options_.info_log, "Compacted %d@%d + %d@%d files => %lld bytes",
|
2011-03-18 23:37:00 +01:00
|
|
|
compact->compaction->num_input_files(0),
|
|
|
|
compact->compaction->level(),
|
|
|
|
compact->compaction->num_input_files(1),
|
|
|
|
compact->compaction->level() + 1,
|
|
|
|
static_cast<long long>(compact->total_bytes));
|
|
|
|
|
|
|
|
// Add compaction outputs
|
|
|
|
compact->compaction->AddInputDeletions(compact->compaction->edit());
|
|
|
|
const int level = compact->compaction->level();
|
2011-04-21 00:48:11 +02:00
|
|
|
for (size_t i = 0; i < compact->outputs.size(); i++) {
|
2011-03-18 23:37:00 +01:00
|
|
|
const CompactionState::Output& out = compact->outputs[i];
|
|
|
|
compact->compaction->edit()->AddFile(
|
|
|
|
level + 1,
|
|
|
|
out.number, out.file_size, out.smallest, out.largest);
|
|
|
|
}
|
2012-01-25 23:56:52 +01:00
|
|
|
return versions_->LogAndApply(compact->compaction->edit(), &mutex_);
|
2011-03-18 23:37:00 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
Status DBImpl::DoCompactionWork(CompactionState* compact) {
|
2011-04-12 21:38:58 +02:00
|
|
|
const uint64_t start_micros = env_->NowMicros();
|
|
|
|
int64_t imm_micros = 0; // Micros spent doing imm_ compactions
|
|
|
|
|
2011-07-21 04:40:18 +02:00
|
|
|
Log(options_.info_log, "Compacting %d@%d + %d@%d files",
|
2011-03-18 23:37:00 +01:00
|
|
|
compact->compaction->num_input_files(0),
|
|
|
|
compact->compaction->level(),
|
|
|
|
compact->compaction->num_input_files(1),
|
|
|
|
compact->compaction->level() + 1);
|
2012-09-21 19:47:08 +02:00
|
|
|
char scratch[256];
|
|
|
|
compact->compaction->Summary(scratch, sizeof(scratch));
|
2012-08-18 03:10:09 +02:00
|
|
|
Log(options_.info_log, "Compaction start summary: %s\n", scratch);
|
2011-03-18 23:37:00 +01:00
|
|
|
|
|
|
|
assert(versions_->NumLevelFiles(compact->compaction->level()) > 0);
|
|
|
|
assert(compact->builder == NULL);
|
|
|
|
assert(compact->outfile == NULL);
|
|
|
|
if (snapshots_.empty()) {
|
2011-04-12 21:38:58 +02:00
|
|
|
compact->smallest_snapshot = versions_->LastSequence();
|
2011-03-18 23:37:00 +01:00
|
|
|
} else {
|
|
|
|
compact->smallest_snapshot = snapshots_.oldest()->number_;
|
|
|
|
}
|
|
|
|
|
|
|
|
// Release mutex while we're actually doing the compaction work
|
|
|
|
mutex_.Unlock();
|
|
|
|
|
|
|
|
Iterator* input = versions_->MakeInputIterator(compact->compaction);
|
|
|
|
input->SeekToFirst();
|
|
|
|
Status status;
|
|
|
|
ParsedInternalKey ikey;
|
|
|
|
std::string current_user_key;
|
|
|
|
bool has_current_user_key = false;
|
|
|
|
SequenceNumber last_sequence_for_key = kMaxSequenceNumber;
|
|
|
|
for (; input->Valid() && !shutting_down_.Acquire_Load(); ) {
|
2011-04-12 21:38:58 +02:00
|
|
|
// Prioritize immutable compaction work
|
|
|
|
if (has_imm_.NoBarrier_Load() != NULL) {
|
|
|
|
const uint64_t imm_start = env_->NowMicros();
|
|
|
|
mutex_.Lock();
|
|
|
|
if (imm_ != NULL) {
|
|
|
|
CompactMemTable();
|
2011-06-07 16:40:26 +02:00
|
|
|
bg_cv_.SignalAll(); // Wakeup MakeRoomForWrite() if necessary
|
2011-04-12 21:38:58 +02:00
|
|
|
}
|
|
|
|
mutex_.Unlock();
|
|
|
|
imm_micros += (env_->NowMicros() - imm_start);
|
|
|
|
}
|
|
|
|
|
2011-03-18 23:37:00 +01:00
|
|
|
Slice key = input->key();
|
2011-05-21 04:17:43 +02:00
|
|
|
if (compact->compaction->ShouldStopBefore(key) &&
|
2011-03-22 19:32:49 +01:00
|
|
|
compact->builder != NULL) {
|
|
|
|
status = FinishCompactionOutputFile(compact, input);
|
|
|
|
if (!status.ok()) {
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
// Handle key/value, add to state, etc.
|
2011-03-18 23:37:00 +01:00
|
|
|
bool drop = false;
|
|
|
|
if (!ParseInternalKey(key, &ikey)) {
|
|
|
|
// Do not hide error keys
|
|
|
|
current_user_key.clear();
|
|
|
|
has_current_user_key = false;
|
|
|
|
last_sequence_for_key = kMaxSequenceNumber;
|
|
|
|
} else {
|
|
|
|
if (!has_current_user_key ||
|
|
|
|
user_comparator()->Compare(ikey.user_key,
|
|
|
|
Slice(current_user_key)) != 0) {
|
|
|
|
// First occurrence of this user key
|
|
|
|
current_user_key.assign(ikey.user_key.data(), ikey.user_key.size());
|
|
|
|
has_current_user_key = true;
|
|
|
|
last_sequence_for_key = kMaxSequenceNumber;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (last_sequence_for_key <= compact->smallest_snapshot) {
|
|
|
|
// Hidden by an newer entry for same user key
|
|
|
|
drop = true; // (A)
|
|
|
|
} else if (ikey.type == kTypeDeletion &&
|
|
|
|
ikey.sequence <= compact->smallest_snapshot &&
|
|
|
|
compact->compaction->IsBaseLevelForKey(ikey.user_key)) {
|
|
|
|
// For this user key:
|
|
|
|
// (1) there is no data in higher levels
|
|
|
|
// (2) data in lower levels will have larger sequence numbers
|
|
|
|
// (3) data in layers that are being compacted here and have
|
|
|
|
// smaller sequence numbers will be dropped in the next
|
|
|
|
// few iterations of this loop (by rule (A) above).
|
|
|
|
// Therefore this deletion marker is obsolete and can be dropped.
|
|
|
|
drop = true;
|
|
|
|
}
|
|
|
|
|
|
|
|
last_sequence_for_key = ikey.sequence;
|
|
|
|
}
|
|
|
|
#if 0
|
2011-07-21 04:40:18 +02:00
|
|
|
Log(options_.info_log,
|
2011-03-18 23:37:00 +01:00
|
|
|
" Compact: %s, seq %d, type: %d %d, drop: %d, is_base: %d, "
|
|
|
|
"%d smallest_snapshot: %d",
|
|
|
|
ikey.user_key.ToString().c_str(),
|
2011-04-21 00:48:11 +02:00
|
|
|
(int)ikey.sequence, ikey.type, kTypeValue, drop,
|
2011-03-18 23:37:00 +01:00
|
|
|
compact->compaction->IsBaseLevelForKey(ikey.user_key),
|
|
|
|
(int)last_sequence_for_key, (int)compact->smallest_snapshot);
|
|
|
|
#endif
|
|
|
|
|
|
|
|
if (!drop) {
|
|
|
|
// Open output file if necessary
|
|
|
|
if (compact->builder == NULL) {
|
|
|
|
status = OpenCompactionOutputFile(compact);
|
|
|
|
if (!status.ok()) {
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
if (compact->builder->NumEntries() == 0) {
|
|
|
|
compact->current_output()->smallest.DecodeFrom(key);
|
|
|
|
}
|
|
|
|
compact->current_output()->largest.DecodeFrom(key);
|
2011-04-21 00:48:11 +02:00
|
|
|
compact->builder->Add(key, input->value());
|
2011-03-18 23:37:00 +01:00
|
|
|
|
|
|
|
// Close output file if it is big enough
|
|
|
|
if (compact->builder->FileSize() >=
|
|
|
|
compact->compaction->MaxOutputFileSize()) {
|
|
|
|
status = FinishCompactionOutputFile(compact, input);
|
|
|
|
if (!status.ok()) {
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
input->Next();
|
|
|
|
}
|
|
|
|
|
|
|
|
if (status.ok() && shutting_down_.Acquire_Load()) {
|
|
|
|
status = Status::IOError("Deleting DB during compaction");
|
|
|
|
}
|
|
|
|
if (status.ok() && compact->builder != NULL) {
|
|
|
|
status = FinishCompactionOutputFile(compact, input);
|
|
|
|
}
|
|
|
|
if (status.ok()) {
|
|
|
|
status = input->status();
|
|
|
|
}
|
|
|
|
delete input;
|
|
|
|
input = NULL;
|
|
|
|
|
2011-04-12 21:38:58 +02:00
|
|
|
CompactionStats stats;
|
|
|
|
stats.micros = env_->NowMicros() - start_micros - imm_micros;
|
2012-10-23 19:34:09 +02:00
|
|
|
|
|
|
|
stats.files_in_leveln = compact->compaction->num_input_files(0);
|
|
|
|
stats.files_in_levelnp1 = compact->compaction->num_input_files(1);
|
|
|
|
stats.files_out_levelnp1 = compact->outputs.size();
|
|
|
|
|
|
|
|
for (int i = 0; i < compact->compaction->num_input_files(0); i++)
|
|
|
|
stats.bytes_readn += compact->compaction->input(0, i)->file_size;
|
|
|
|
|
|
|
|
for (int i = 0; i < compact->compaction->num_input_files(1); i++)
|
|
|
|
stats.bytes_readnp1 += compact->compaction->input(1, i)->file_size;
|
|
|
|
|
2011-04-21 00:48:11 +02:00
|
|
|
for (size_t i = 0; i < compact->outputs.size(); i++) {
|
2011-04-12 21:38:58 +02:00
|
|
|
stats.bytes_written += compact->outputs[i].file_size;
|
|
|
|
}
|
|
|
|
|
2011-03-18 23:37:00 +01:00
|
|
|
mutex_.Lock();
|
2011-04-12 21:38:58 +02:00
|
|
|
stats_[compact->compaction->level() + 1].Add(stats);
|
2011-03-18 23:37:00 +01:00
|
|
|
|
|
|
|
if (status.ok()) {
|
|
|
|
status = InstallCompactionResults(compact);
|
|
|
|
}
|
2011-05-21 04:17:43 +02:00
|
|
|
VersionSet::LevelSummaryStorage tmp;
|
2011-07-21 04:40:18 +02:00
|
|
|
Log(options_.info_log,
|
2012-10-23 19:34:09 +02:00
|
|
|
"compacted to: %s, %.1f MB/sec, level %d, files in(%d, %d) out(%d) "
|
|
|
|
"MB in(%.1f, %.1f) out(%.1f), amplify(%.1f)\n",
|
|
|
|
versions_->LevelSummary(&tmp),
|
|
|
|
(stats.bytes_readn + stats.bytes_readnp1 + stats.bytes_written) /
|
|
|
|
(double) stats.micros,
|
|
|
|
compact->compaction->level() + 1,
|
|
|
|
stats.files_in_leveln, stats.files_in_levelnp1, stats.files_out_levelnp1,
|
|
|
|
stats.bytes_readn / 1048576.0,
|
|
|
|
stats.bytes_readnp1 / 1048576.0,
|
|
|
|
stats.bytes_written / 1048576.0,
|
|
|
|
(stats.bytes_written + stats.bytes_readnp1) /
|
|
|
|
(double) stats.bytes_readn);
|
|
|
|
|
2011-03-18 23:37:00 +01:00
|
|
|
return status;
|
|
|
|
}
|
|
|
|
|
2011-05-28 02:53:58 +02:00
|
|
|
namespace {
|
|
|
|
struct IterState {
|
|
|
|
port::Mutex* mu;
|
|
|
|
Version* version;
|
|
|
|
MemTable* mem;
|
|
|
|
MemTable* imm;
|
|
|
|
};
|
|
|
|
|
|
|
|
static void CleanupIteratorState(void* arg1, void* arg2) {
|
|
|
|
IterState* state = reinterpret_cast<IterState*>(arg1);
|
|
|
|
state->mu->Lock();
|
|
|
|
state->mem->Unref();
|
|
|
|
if (state->imm != NULL) state->imm->Unref();
|
|
|
|
state->version->Unref();
|
|
|
|
state->mu->Unlock();
|
|
|
|
delete state;
|
|
|
|
}
|
2011-10-31 18:22:06 +01:00
|
|
|
} // namespace
|
2011-05-28 02:53:58 +02:00
|
|
|
|
2011-03-18 23:37:00 +01:00
|
|
|
Iterator* DBImpl::NewInternalIterator(const ReadOptions& options,
|
|
|
|
SequenceNumber* latest_snapshot) {
|
2011-05-28 02:53:58 +02:00
|
|
|
IterState* cleanup = new IterState;
|
2011-03-18 23:37:00 +01:00
|
|
|
mutex_.Lock();
|
2011-04-12 21:38:58 +02:00
|
|
|
*latest_snapshot = versions_->LastSequence();
|
2011-03-18 23:37:00 +01:00
|
|
|
|
|
|
|
// Collect together all needed child iterators
|
|
|
|
std::vector<Iterator*> list;
|
|
|
|
list.push_back(mem_->NewIterator());
|
2011-05-28 02:53:58 +02:00
|
|
|
mem_->Ref();
|
2011-04-12 21:38:58 +02:00
|
|
|
if (imm_ != NULL) {
|
|
|
|
list.push_back(imm_->NewIterator());
|
2011-05-28 02:53:58 +02:00
|
|
|
imm_->Ref();
|
2011-04-12 21:38:58 +02:00
|
|
|
}
|
2011-03-18 23:37:00 +01:00
|
|
|
versions_->current()->AddIterators(options, &list);
|
|
|
|
Iterator* internal_iter =
|
|
|
|
NewMergingIterator(&internal_comparator_, &list[0], list.size());
|
|
|
|
versions_->current()->Ref();
|
2011-05-28 02:53:58 +02:00
|
|
|
|
|
|
|
cleanup->mu = &mutex_;
|
|
|
|
cleanup->mem = mem_;
|
|
|
|
cleanup->imm = imm_;
|
|
|
|
cleanup->version = versions_->current();
|
|
|
|
internal_iter->RegisterCleanup(CleanupIteratorState, cleanup, NULL);
|
2011-03-18 23:37:00 +01:00
|
|
|
|
|
|
|
mutex_.Unlock();
|
|
|
|
return internal_iter;
|
|
|
|
}
|
|
|
|
|
|
|
|
Iterator* DBImpl::TEST_NewInternalIterator() {
|
|
|
|
SequenceNumber ignored;
|
|
|
|
return NewInternalIterator(ReadOptions(), &ignored);
|
|
|
|
}
|
|
|
|
|
2011-03-23 00:24:02 +01:00
|
|
|
int64_t DBImpl::TEST_MaxNextLevelOverlappingBytes() {
|
2011-03-22 19:32:49 +01:00
|
|
|
MutexLock l(&mutex_);
|
|
|
|
return versions_->MaxNextLevelOverlappingBytes();
|
|
|
|
}
|
|
|
|
|
2011-03-18 23:37:00 +01:00
|
|
|
Status DBImpl::Get(const ReadOptions& options,
|
|
|
|
const Slice& key,
|
|
|
|
std::string* value) {
|
2011-06-22 04:36:45 +02:00
|
|
|
Status s;
|
|
|
|
MutexLock l(&mutex_);
|
|
|
|
SequenceNumber snapshot;
|
|
|
|
if (options.snapshot != NULL) {
|
|
|
|
snapshot = reinterpret_cast<const SnapshotImpl*>(options.snapshot)->number_;
|
|
|
|
} else {
|
|
|
|
snapshot = versions_->LastSequence();
|
2011-03-18 23:37:00 +01:00
|
|
|
}
|
2011-06-22 04:36:45 +02:00
|
|
|
|
2011-08-22 23:08:51 +02:00
|
|
|
MemTable* mem = mem_;
|
|
|
|
MemTable* imm = imm_;
|
2011-06-22 04:36:45 +02:00
|
|
|
Version* current = versions_->current();
|
2011-08-22 23:08:51 +02:00
|
|
|
mem->Ref();
|
|
|
|
if (imm != NULL) imm->Ref();
|
2011-06-22 04:36:45 +02:00
|
|
|
current->Ref();
|
2011-08-22 23:08:51 +02:00
|
|
|
|
|
|
|
bool have_stat_update = false;
|
2011-06-22 04:36:45 +02:00
|
|
|
Version::GetStats stats;
|
2011-08-22 23:08:51 +02:00
|
|
|
|
|
|
|
// Unlock while reading from files and memtables
|
|
|
|
{
|
2011-06-22 04:36:45 +02:00
|
|
|
mutex_.Unlock();
|
2011-08-22 23:08:51 +02:00
|
|
|
// First look in the memtable, then in the immutable memtable (if any).
|
|
|
|
LookupKey lkey(key, snapshot);
|
2011-09-01 21:08:02 +02:00
|
|
|
if (mem->Get(lkey, value, &s)) {
|
2011-08-22 23:08:51 +02:00
|
|
|
// Done
|
2011-09-01 21:08:02 +02:00
|
|
|
} else if (imm != NULL && imm->Get(lkey, value, &s)) {
|
2011-08-22 23:08:51 +02:00
|
|
|
// Done
|
|
|
|
} else {
|
|
|
|
s = current->Get(options, lkey, value, &stats);
|
|
|
|
have_stat_update = true;
|
|
|
|
}
|
2011-06-22 04:36:45 +02:00
|
|
|
mutex_.Lock();
|
|
|
|
}
|
2011-08-22 23:08:51 +02:00
|
|
|
|
2012-09-25 18:15:32 +02:00
|
|
|
if (!options_.disable_seek_compaction &&
|
|
|
|
have_stat_update && current->UpdateStats(stats)) {
|
2011-06-22 04:36:45 +02:00
|
|
|
MaybeScheduleCompaction();
|
|
|
|
}
|
2011-08-22 23:08:51 +02:00
|
|
|
mem->Unref();
|
|
|
|
if (imm != NULL) imm->Unref();
|
2011-06-22 04:36:45 +02:00
|
|
|
current->Unref();
|
|
|
|
return s;
|
2011-03-18 23:37:00 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
Iterator* DBImpl::NewIterator(const ReadOptions& options) {
|
|
|
|
SequenceNumber latest_snapshot;
|
|
|
|
Iterator* internal_iter = NewInternalIterator(options, &latest_snapshot);
|
2011-05-21 04:17:43 +02:00
|
|
|
return NewDBIterator(
|
|
|
|
&dbname_, env_, user_comparator(), internal_iter,
|
|
|
|
(options.snapshot != NULL
|
|
|
|
? reinterpret_cast<const SnapshotImpl*>(options.snapshot)->number_
|
|
|
|
: latest_snapshot));
|
2011-03-18 23:37:00 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
const Snapshot* DBImpl::GetSnapshot() {
|
|
|
|
MutexLock l(&mutex_);
|
2011-04-12 21:38:58 +02:00
|
|
|
return snapshots_.New(versions_->LastSequence());
|
2011-03-18 23:37:00 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
void DBImpl::ReleaseSnapshot(const Snapshot* s) {
|
|
|
|
MutexLock l(&mutex_);
|
2011-05-21 04:17:43 +02:00
|
|
|
snapshots_.Delete(reinterpret_cast<const SnapshotImpl*>(s));
|
2011-03-18 23:37:00 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
// Convenience methods
|
|
|
|
Status DBImpl::Put(const WriteOptions& o, const Slice& key, const Slice& val) {
|
|
|
|
return DB::Put(o, key, val);
|
|
|
|
}
|
|
|
|
|
|
|
|
Status DBImpl::Delete(const WriteOptions& options, const Slice& key) {
|
|
|
|
return DB::Delete(options, key);
|
|
|
|
}
|
|
|
|
|
2012-03-09 01:23:21 +01:00
|
|
|
Status DBImpl::Write(const WriteOptions& options, WriteBatch* my_batch) {
|
|
|
|
Writer w(&mutex_);
|
|
|
|
w.batch = my_batch;
|
|
|
|
w.sync = options.sync;
|
2012-07-05 22:39:28 +02:00
|
|
|
w.disableWAL = options.disableWAL;
|
2012-03-09 01:23:21 +01:00
|
|
|
w.done = false;
|
2011-09-01 21:08:02 +02:00
|
|
|
|
2011-04-21 00:48:11 +02:00
|
|
|
MutexLock l(&mutex_);
|
2012-03-09 01:23:21 +01:00
|
|
|
writers_.push_back(&w);
|
|
|
|
while (!w.done && &w != writers_.front()) {
|
|
|
|
w.cv.Wait();
|
|
|
|
}
|
|
|
|
if (w.done) {
|
|
|
|
return w.status;
|
|
|
|
}
|
|
|
|
|
|
|
|
// May temporarily unlock and wait.
|
|
|
|
Status status = MakeRoomForWrite(my_batch == NULL);
|
2011-04-21 00:48:11 +02:00
|
|
|
uint64_t last_sequence = versions_->LastSequence();
|
2012-03-09 01:23:21 +01:00
|
|
|
Writer* last_writer = &w;
|
|
|
|
if (status.ok() && my_batch != NULL) { // NULL batch is for compactions
|
|
|
|
WriteBatch* updates = BuildBatchGroup(&last_writer);
|
2011-04-21 00:48:11 +02:00
|
|
|
WriteBatchInternal::SetSequence(updates, last_sequence + 1);
|
|
|
|
last_sequence += WriteBatchInternal::Count(updates);
|
|
|
|
|
2012-03-09 01:23:21 +01:00
|
|
|
// Add to log and apply to memtable. We can release the lock
|
|
|
|
// during this phase since &w is currently responsible for logging
|
|
|
|
// and protects against concurrent loggers and concurrent writes
|
|
|
|
// into mem_.
|
2011-09-01 21:08:02 +02:00
|
|
|
{
|
|
|
|
mutex_.Unlock();
|
2012-07-05 22:39:28 +02:00
|
|
|
if (!options.disableWAL) {
|
|
|
|
status = log_->AddRecord(WriteBatchInternal::Contents(updates));
|
|
|
|
if (status.ok() && options.sync) {
|
2012-08-27 21:10:26 +02:00
|
|
|
if (options_.use_fsync) {
|
|
|
|
status = logfile_->Fsync();
|
|
|
|
} else {
|
|
|
|
status = logfile_->Sync();
|
|
|
|
}
|
2012-07-05 22:39:28 +02:00
|
|
|
}
|
2011-09-01 21:08:02 +02:00
|
|
|
}
|
|
|
|
if (status.ok()) {
|
|
|
|
status = WriteBatchInternal::InsertInto(updates, mem_);
|
|
|
|
}
|
|
|
|
mutex_.Lock();
|
2011-03-18 23:37:00 +01:00
|
|
|
}
|
2012-03-09 01:23:21 +01:00
|
|
|
if (updates == tmp_batch_) tmp_batch_->Clear();
|
2011-09-01 21:08:02 +02:00
|
|
|
|
|
|
|
versions_->SetLastSequence(last_sequence);
|
2011-03-18 23:37:00 +01:00
|
|
|
}
|
2012-03-09 01:23:21 +01:00
|
|
|
|
|
|
|
while (true) {
|
|
|
|
Writer* ready = writers_.front();
|
|
|
|
writers_.pop_front();
|
|
|
|
if (ready != &w) {
|
|
|
|
ready->status = status;
|
|
|
|
ready->done = true;
|
|
|
|
ready->cv.Signal();
|
|
|
|
}
|
|
|
|
if (ready == last_writer) break;
|
|
|
|
}
|
|
|
|
|
|
|
|
// Notify new head of write queue
|
|
|
|
if (!writers_.empty()) {
|
|
|
|
writers_.front()->cv.Signal();
|
|
|
|
}
|
|
|
|
|
2011-03-18 23:37:00 +01:00
|
|
|
return status;
|
|
|
|
}
|
|
|
|
|
2012-03-09 01:23:21 +01:00
|
|
|
// REQUIRES: Writer list must be non-empty
|
|
|
|
// REQUIRES: First writer must have a non-NULL batch
|
|
|
|
WriteBatch* DBImpl::BuildBatchGroup(Writer** last_writer) {
|
|
|
|
assert(!writers_.empty());
|
|
|
|
Writer* first = writers_.front();
|
|
|
|
WriteBatch* result = first->batch;
|
|
|
|
assert(result != NULL);
|
|
|
|
|
|
|
|
size_t size = WriteBatchInternal::ByteSize(first->batch);
|
|
|
|
|
|
|
|
// Allow the group to grow up to a maximum size, but if the
|
|
|
|
// original write is small, limit the growth so we do not slow
|
|
|
|
// down the small write too much.
|
|
|
|
size_t max_size = 1 << 20;
|
|
|
|
if (size <= (128<<10)) {
|
|
|
|
max_size = size + (128<<10);
|
|
|
|
}
|
|
|
|
|
|
|
|
*last_writer = first;
|
|
|
|
std::deque<Writer*>::iterator iter = writers_.begin();
|
|
|
|
++iter; // Advance past "first"
|
|
|
|
for (; iter != writers_.end(); ++iter) {
|
|
|
|
Writer* w = *iter;
|
|
|
|
if (w->sync && !first->sync) {
|
|
|
|
// Do not include a sync write into a batch handled by a non-sync write.
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
2012-07-05 22:39:28 +02:00
|
|
|
if (!w->disableWAL && first->disableWAL) {
|
|
|
|
// Do not include a write that needs WAL into a batch that has
|
|
|
|
// WAL disabled.
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
2012-03-09 01:23:21 +01:00
|
|
|
if (w->batch != NULL) {
|
|
|
|
size += WriteBatchInternal::ByteSize(w->batch);
|
|
|
|
if (size > max_size) {
|
|
|
|
// Do not make batch too big
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
// Append to *reuslt
|
|
|
|
if (result == first->batch) {
|
|
|
|
// Switch to temporary batch instead of disturbing caller's batch
|
|
|
|
result = tmp_batch_;
|
|
|
|
assert(WriteBatchInternal::Count(result) == 0);
|
|
|
|
WriteBatchInternal::Append(result, first->batch);
|
|
|
|
}
|
|
|
|
WriteBatchInternal::Append(result, w->batch);
|
|
|
|
}
|
|
|
|
*last_writer = w;
|
|
|
|
}
|
|
|
|
return result;
|
|
|
|
}
|
|
|
|
|
2011-09-01 21:08:02 +02:00
|
|
|
// REQUIRES: mutex_ is held
|
2012-03-09 01:23:21 +01:00
|
|
|
// REQUIRES: this thread is currently at the front of the writer queue
|
2011-04-12 21:38:58 +02:00
|
|
|
Status DBImpl::MakeRoomForWrite(bool force) {
|
|
|
|
mutex_.AssertHeld();
|
2012-03-09 01:23:21 +01:00
|
|
|
assert(!writers_.empty());
|
2011-05-21 04:17:43 +02:00
|
|
|
bool allow_delay = !force;
|
2011-04-12 21:38:58 +02:00
|
|
|
Status s;
|
2012-06-23 04:30:03 +02:00
|
|
|
|
2011-04-12 21:38:58 +02:00
|
|
|
while (true) {
|
|
|
|
if (!bg_error_.ok()) {
|
|
|
|
// Yield previous error
|
|
|
|
s = bg_error_;
|
|
|
|
break;
|
2011-05-21 04:17:43 +02:00
|
|
|
} else if (
|
|
|
|
allow_delay &&
|
2012-06-23 04:30:03 +02:00
|
|
|
versions_->NumLevelFiles(0) >=
|
|
|
|
options_.level0_slowdown_writes_trigger) {
|
2011-05-21 04:17:43 +02:00
|
|
|
// We are getting close to hitting a hard limit on the number of
|
|
|
|
// L0 files. Rather than delaying a single write by several
|
|
|
|
// seconds when we hit the hard limit, start delaying each
|
|
|
|
// individual write by 1ms to reduce latency variance. Also,
|
|
|
|
// this delay hands over some CPU to the compaction thread in
|
|
|
|
// case it is sharing the same core as the writer.
|
|
|
|
mutex_.Unlock();
|
|
|
|
env_->SleepForMicroseconds(1000);
|
2012-10-23 19:34:09 +02:00
|
|
|
stall_level0_slowdown_ += 1000;
|
2011-05-21 04:17:43 +02:00
|
|
|
allow_delay = false; // Do not delay a single write more than once
|
2012-05-31 02:04:35 +02:00
|
|
|
Log(options_.info_log, "delaying write...\n");
|
2011-05-21 04:17:43 +02:00
|
|
|
mutex_.Lock();
|
2011-04-12 21:38:58 +02:00
|
|
|
} else if (!force &&
|
|
|
|
(mem_->ApproximateMemoryUsage() <= options_.write_buffer_size)) {
|
|
|
|
// There is room in current memtable
|
|
|
|
break;
|
|
|
|
} else if (imm_ != NULL) {
|
|
|
|
// We have filled up the current memtable, but the previous
|
|
|
|
// one is still being compacted, so we wait.
|
2012-08-23 01:57:51 +02:00
|
|
|
Log(options_.info_log, "wait for memtable compaction...\n");
|
2012-10-23 19:34:09 +02:00
|
|
|
uint64_t t1 = env_->NowMicros();
|
2011-06-07 16:40:26 +02:00
|
|
|
bg_cv_.Wait();
|
2012-10-23 19:34:09 +02:00
|
|
|
stall_memtable_compaction_ += env_->NowMicros() - t1;
|
2012-06-23 04:30:03 +02:00
|
|
|
} else if (versions_->NumLevelFiles(0) >=
|
|
|
|
options_.level0_stop_writes_trigger) {
|
2011-05-21 04:17:43 +02:00
|
|
|
// There are too many level-0 files.
|
2011-07-21 04:40:18 +02:00
|
|
|
Log(options_.info_log, "waiting...\n");
|
2012-10-23 19:34:09 +02:00
|
|
|
uint64_t t1 = env_->NowMicros();
|
2011-06-07 16:40:26 +02:00
|
|
|
bg_cv_.Wait();
|
2012-10-23 19:34:09 +02:00
|
|
|
stall_level0_num_files_ += env_->NowMicros() - t1;
|
2011-04-12 21:38:58 +02:00
|
|
|
} else {
|
|
|
|
// Attempt to switch to a new memtable and trigger compaction of old
|
|
|
|
assert(versions_->PrevLogNumber() == 0);
|
|
|
|
uint64_t new_log_number = versions_->NewFileNumber();
|
|
|
|
WritableFile* lfile = NULL;
|
|
|
|
s = env_->NewWritableFile(LogFileName(dbname_, new_log_number), &lfile);
|
|
|
|
if (!s.ok()) {
|
2012-08-27 08:45:35 +02:00
|
|
|
// Avoid chewing through file number space in a tight loop.
|
|
|
|
versions_->ReuseFileNumber(new_log_number);
|
2011-04-12 21:38:58 +02:00
|
|
|
break;
|
|
|
|
}
|
|
|
|
delete log_;
|
|
|
|
delete logfile_;
|
|
|
|
logfile_ = lfile;
|
2011-06-22 04:36:45 +02:00
|
|
|
logfile_number_ = new_log_number;
|
2011-04-12 21:38:58 +02:00
|
|
|
log_ = new log::Writer(lfile);
|
|
|
|
imm_ = mem_;
|
|
|
|
has_imm_.Release_Store(imm_);
|
|
|
|
mem_ = new MemTable(internal_comparator_);
|
2011-05-21 04:17:43 +02:00
|
|
|
mem_->Ref();
|
2011-04-12 21:38:58 +02:00
|
|
|
force = false; // Do not force another compaction if have room
|
|
|
|
MaybeScheduleCompaction();
|
|
|
|
}
|
|
|
|
}
|
|
|
|
return s;
|
|
|
|
}
|
|
|
|
|
|
|
|
bool DBImpl::GetProperty(const Slice& property, std::string* value) {
|
|
|
|
value->clear();
|
|
|
|
|
2011-03-18 23:37:00 +01:00
|
|
|
MutexLock l(&mutex_);
|
|
|
|
Slice in = property;
|
|
|
|
Slice prefix("leveldb.");
|
|
|
|
if (!in.starts_with(prefix)) return false;
|
|
|
|
in.remove_prefix(prefix.size());
|
|
|
|
|
|
|
|
if (in.starts_with("num-files-at-level")) {
|
|
|
|
in.remove_prefix(strlen("num-files-at-level"));
|
|
|
|
uint64_t level;
|
|
|
|
bool ok = ConsumeDecimalNumber(&in, &level) && in.empty();
|
2012-08-29 21:29:43 +02:00
|
|
|
if (!ok || (int)level >= NumberLevels()) {
|
2011-03-18 23:37:00 +01:00
|
|
|
return false;
|
|
|
|
} else {
|
2011-04-12 21:38:58 +02:00
|
|
|
char buf[100];
|
2011-04-21 00:48:11 +02:00
|
|
|
snprintf(buf, sizeof(buf), "%d",
|
|
|
|
versions_->NumLevelFiles(static_cast<int>(level)));
|
2011-04-12 21:38:58 +02:00
|
|
|
*value = buf;
|
2011-03-18 23:37:00 +01:00
|
|
|
return true;
|
|
|
|
}
|
2011-04-12 21:38:58 +02:00
|
|
|
} else if (in == "stats") {
|
2012-10-23 19:34:09 +02:00
|
|
|
char buf[1000];
|
|
|
|
uint64_t total_bytes = 0;
|
|
|
|
uint64_t micros_up = env_->NowMicros() - started_at_;
|
|
|
|
double seconds_up = micros_up / 1000000.0;
|
|
|
|
|
|
|
|
// Pardon the long line but I think it is easier to read this way.
|
2011-04-12 21:38:58 +02:00
|
|
|
snprintf(buf, sizeof(buf),
|
|
|
|
" Compactions\n"
|
2012-10-23 19:34:09 +02:00
|
|
|
"Level Files Size(MB) Time(sec) Read(MB) Write(MB) Rn(MB) Rnp1(MB) Wnew(MB) Amplify Read(MB/s) Write(MB/s) Rn Rnp1 Wnp1 NewW Count\n"
|
|
|
|
"------------------------------------------------------------------------------------------------------------------------------------------------------------\n"
|
2011-04-12 21:38:58 +02:00
|
|
|
);
|
|
|
|
value->append(buf);
|
2012-06-23 04:30:03 +02:00
|
|
|
for (int level = 0; level < NumberLevels(); level++) {
|
2011-04-12 21:38:58 +02:00
|
|
|
int files = versions_->NumLevelFiles(level);
|
|
|
|
if (stats_[level].micros > 0 || files > 0) {
|
2012-10-23 19:34:09 +02:00
|
|
|
int64_t bytes_read = stats_[level].bytes_readn +
|
|
|
|
stats_[level].bytes_readnp1;
|
|
|
|
int64_t bytes_new = stats_[level].bytes_written -
|
|
|
|
stats_[level].bytes_readnp1;
|
|
|
|
double amplify = (stats_[level].bytes_readn == 0)
|
|
|
|
? 0.0
|
|
|
|
: (stats_[level].bytes_written + stats_[level].bytes_readnp1) /
|
|
|
|
(double) stats_[level].bytes_readn;
|
|
|
|
|
|
|
|
total_bytes += bytes_read + stats_[level].bytes_written;
|
2011-04-12 21:38:58 +02:00
|
|
|
snprintf(
|
|
|
|
buf, sizeof(buf),
|
2012-10-23 19:34:09 +02:00
|
|
|
"%3d %8d %8.0f %9.0f %9.0f %9.0f %9.0f %9.0f %9.0f %7.1f %9.1f %11.1f %8d %8d %8d %8d %8d\n",
|
2011-04-12 21:38:58 +02:00
|
|
|
level,
|
|
|
|
files,
|
|
|
|
versions_->NumLevelBytes(level) / 1048576.0,
|
|
|
|
stats_[level].micros / 1e6,
|
2012-10-23 19:34:09 +02:00
|
|
|
bytes_read / 1048576.0,
|
|
|
|
stats_[level].bytes_written / 1048576.0,
|
|
|
|
stats_[level].bytes_readn / 1048576.0,
|
|
|
|
stats_[level].bytes_readnp1 / 1048576.0,
|
|
|
|
bytes_new / 1048576.0,
|
|
|
|
amplify,
|
2012-10-26 22:00:42 +02:00
|
|
|
(bytes_read / 1048576.0) / (stats_[level].micros / 1000000.0),
|
|
|
|
(stats_[level].bytes_written / 1048576.0) /
|
|
|
|
(stats_[level].micros / 1000000.0),
|
2012-10-23 19:34:09 +02:00
|
|
|
stats_[level].files_in_leveln,
|
|
|
|
stats_[level].files_in_levelnp1,
|
|
|
|
stats_[level].files_out_levelnp1,
|
|
|
|
stats_[level].files_out_levelnp1 - stats_[level].files_in_levelnp1,
|
|
|
|
stats_[level].count);
|
2011-04-12 21:38:58 +02:00
|
|
|
value->append(buf);
|
|
|
|
}
|
|
|
|
}
|
2012-10-23 19:34:09 +02:00
|
|
|
|
|
|
|
snprintf(buf, sizeof(buf),
|
|
|
|
"Amplification: %.1f rate, %.2f GB in, %.2f GB out\n",
|
|
|
|
(double) total_bytes / stats_[0].bytes_written,
|
|
|
|
stats_[0].bytes_written / (1048576.0 * 1024),
|
|
|
|
total_bytes / (1048576.0 * 1024));
|
|
|
|
value->append(buf);
|
|
|
|
|
|
|
|
snprintf(buf, sizeof(buf), "Uptime(secs): %.1f\n", seconds_up);
|
|
|
|
value->append(buf);
|
|
|
|
|
|
|
|
snprintf(buf, sizeof(buf),
|
|
|
|
"Stalls(secs): %.3f level0_slowdown, %.3f level0_numfiles, "
|
|
|
|
"%.3f memtable_compaction\n",
|
|
|
|
stall_level0_slowdown_ / 1000000.0,
|
|
|
|
stall_level0_num_files_ / 1000000.0,
|
|
|
|
stall_memtable_compaction_ / 1000000.0);
|
|
|
|
value->append(buf);
|
|
|
|
|
2011-04-12 21:38:58 +02:00
|
|
|
return true;
|
2011-10-06 01:30:28 +02:00
|
|
|
} else if (in == "sstables") {
|
|
|
|
*value = versions_->current()->DebugString();
|
|
|
|
return true;
|
2011-03-18 23:37:00 +01:00
|
|
|
}
|
2011-04-12 21:38:58 +02:00
|
|
|
|
2011-03-18 23:37:00 +01:00
|
|
|
return false;
|
|
|
|
}
|
|
|
|
|
|
|
|
void DBImpl::GetApproximateSizes(
|
|
|
|
const Range* range, int n,
|
|
|
|
uint64_t* sizes) {
|
|
|
|
// TODO(opt): better implementation
|
|
|
|
Version* v;
|
|
|
|
{
|
|
|
|
MutexLock l(&mutex_);
|
|
|
|
versions_->current()->Ref();
|
|
|
|
v = versions_->current();
|
|
|
|
}
|
|
|
|
|
|
|
|
for (int i = 0; i < n; i++) {
|
|
|
|
// Convert user_key into a corresponding internal key.
|
|
|
|
InternalKey k1(range[i].start, kMaxSequenceNumber, kValueTypeForSeek);
|
|
|
|
InternalKey k2(range[i].limit, kMaxSequenceNumber, kValueTypeForSeek);
|
|
|
|
uint64_t start = versions_->ApproximateOffsetOf(v, k1);
|
|
|
|
uint64_t limit = versions_->ApproximateOffsetOf(v, k2);
|
|
|
|
sizes[i] = (limit >= start ? limit - start : 0);
|
|
|
|
}
|
|
|
|
|
|
|
|
{
|
|
|
|
MutexLock l(&mutex_);
|
|
|
|
v->Unref();
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
// Default implementations of convenience methods that subclasses of DB
|
|
|
|
// can call if they wish
|
|
|
|
Status DB::Put(const WriteOptions& opt, const Slice& key, const Slice& value) {
|
|
|
|
WriteBatch batch;
|
|
|
|
batch.Put(key, value);
|
|
|
|
return Write(opt, &batch);
|
|
|
|
}
|
|
|
|
|
|
|
|
Status DB::Delete(const WriteOptions& opt, const Slice& key) {
|
|
|
|
WriteBatch batch;
|
|
|
|
batch.Delete(key);
|
|
|
|
return Write(opt, &batch);
|
|
|
|
}
|
|
|
|
|
|
|
|
DB::~DB() { }
|
|
|
|
|
|
|
|
Status DB::Open(const Options& options, const std::string& dbname,
|
|
|
|
DB** dbptr) {
|
|
|
|
*dbptr = NULL;
|
|
|
|
|
|
|
|
DBImpl* impl = new DBImpl(options, dbname);
|
|
|
|
impl->mutex_.Lock();
|
2012-06-23 04:30:03 +02:00
|
|
|
VersionEdit edit(impl->NumberLevels());
|
2011-03-18 23:37:00 +01:00
|
|
|
Status s = impl->Recover(&edit); // Handles create_if_missing, error_if_exists
|
|
|
|
if (s.ok()) {
|
2011-04-12 21:38:58 +02:00
|
|
|
uint64_t new_log_number = impl->versions_->NewFileNumber();
|
2011-03-18 23:37:00 +01:00
|
|
|
WritableFile* lfile;
|
2011-04-12 21:38:58 +02:00
|
|
|
s = options.env->NewWritableFile(LogFileName(dbname, new_log_number),
|
2011-03-18 23:37:00 +01:00
|
|
|
&lfile);
|
|
|
|
if (s.ok()) {
|
2011-04-12 21:38:58 +02:00
|
|
|
edit.SetLogNumber(new_log_number);
|
2011-03-18 23:37:00 +01:00
|
|
|
impl->logfile_ = lfile;
|
2011-06-22 04:36:45 +02:00
|
|
|
impl->logfile_number_ = new_log_number;
|
2011-03-18 23:37:00 +01:00
|
|
|
impl->log_ = new log::Writer(lfile);
|
2011-09-01 21:08:02 +02:00
|
|
|
s = impl->versions_->LogAndApply(&edit, &impl->mutex_);
|
2011-03-18 23:37:00 +01:00
|
|
|
}
|
|
|
|
if (s.ok()) {
|
|
|
|
impl->DeleteObsoleteFiles();
|
2011-05-21 04:17:43 +02:00
|
|
|
impl->MaybeScheduleCompaction();
|
2012-08-15 00:20:36 +02:00
|
|
|
impl->MaybeScheduleLogDBDeployStats();
|
2011-03-18 23:37:00 +01:00
|
|
|
}
|
|
|
|
}
|
|
|
|
impl->mutex_.Unlock();
|
|
|
|
if (s.ok()) {
|
|
|
|
*dbptr = impl;
|
|
|
|
} else {
|
|
|
|
delete impl;
|
|
|
|
}
|
|
|
|
return s;
|
|
|
|
}
|
|
|
|
|
2011-05-21 04:17:43 +02:00
|
|
|
Snapshot::~Snapshot() {
|
|
|
|
}
|
|
|
|
|
2011-03-18 23:37:00 +01:00
|
|
|
Status DestroyDB(const std::string& dbname, const Options& options) {
|
|
|
|
Env* env = options.env;
|
|
|
|
std::vector<std::string> filenames;
|
|
|
|
// Ignore error in case directory does not exist
|
|
|
|
env->GetChildren(dbname, &filenames);
|
|
|
|
if (filenames.empty()) {
|
|
|
|
return Status::OK();
|
|
|
|
}
|
|
|
|
|
|
|
|
FileLock* lock;
|
2011-07-15 02:20:57 +02:00
|
|
|
const std::string lockname = LockFileName(dbname);
|
|
|
|
Status result = env->LockFile(lockname, &lock);
|
2011-03-18 23:37:00 +01:00
|
|
|
if (result.ok()) {
|
|
|
|
uint64_t number;
|
|
|
|
FileType type;
|
2011-04-21 00:48:11 +02:00
|
|
|
for (size_t i = 0; i < filenames.size(); i++) {
|
2011-07-15 02:20:57 +02:00
|
|
|
if (ParseFileName(filenames[i], &number, &type) &&
|
2012-03-09 16:51:04 +01:00
|
|
|
type != kDBLockFile) { // Lock file will be deleted at end
|
2011-03-18 23:37:00 +01:00
|
|
|
Status del = env->DeleteFile(dbname + "/" + filenames[i]);
|
|
|
|
if (result.ok() && !del.ok()) {
|
|
|
|
result = del;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
env->UnlockFile(lock); // Ignore error since state is already gone
|
2011-07-15 02:20:57 +02:00
|
|
|
env->DeleteFile(lockname);
|
2011-03-18 23:37:00 +01:00
|
|
|
env->DeleteDir(dbname); // Ignore error in case dir contains other files
|
|
|
|
}
|
|
|
|
return result;
|
|
|
|
}
|
|
|
|
|
2012-08-21 09:33:21 +02:00
|
|
|
//
|
|
|
|
// A global method that can dump out the build version
|
2012-09-18 22:18:32 +02:00
|
|
|
void dumpLeveldbBuildVersion(Logger * log) {
|
|
|
|
Log(log, "Git sha %s", leveldb_build_git_sha);
|
|
|
|
Log(log, "Git datetime %s", leveldb_build_git_datetime);
|
|
|
|
Log(log, "Compile time %s %s", leveldb_build_compile_time, leveldb_build_compile_date);
|
2012-08-21 09:33:21 +02:00
|
|
|
}
|
|
|
|
|
2011-10-31 18:22:06 +01:00
|
|
|
} // namespace leveldb
|