Version* v = new Version(this);{Builder builder(this, current_);builder.Apply(edit);builder.SaveTo(v);}Finalize(v);

import leveldb
ret ＝ leveldb.RepairDB('/data/mon.iecvq/store.db')

private:friend class VersionSet;typedef std::set< std::pair<int, uint64_t> > DeletedFileSet;std::string comparator_;uint64_t log_number_;uint64_t prev_log_number_;uint64_t next_file_number_;SequenceNumber last_sequence_;bool has_comparator_;bool has_log_number_;bool has_prev_log_number_;bool has_next_file_number_;bool has_last_sequence_;std::vector< std::pair<int, InternalKey> > compact_pointers_;DeletedFileSet deleted_files_;std::vector< std::pair<int, FileMetaData> > new_files_;

enum Tag {kComparator           = 1,kLogNumber            = 2,kNextFileNumber       = 3,kLastSequence         = 4,kCompactPointer       = 5,kDeletedFile          = 6,kNewFile              = 7,// 8 was used for large value refskPrevLogNumber        = 9
};

      std::string record;/*先调用VersionEdit的 EncodeTo方法，序列化成字符串*/edit->EncodeTo(&record);/*注意，descriptor_log_是log文件*/s = descriptor_log_->AddRecord(record);if (s.ok()) {s = descriptor_file_->Sync();}if (!s.ok()) {Log(options_->info_log, "MANIFEST write: %s\n", s.ToString().c_str());}

void VersionEdit::EncodeTo(std::string* dst) const {/*将比较器的标识和名称放入序列化字符串中*/if (has_comparator_) {PutVarint32(dst, kComparator);PutLengthPrefixedSlice(dst, comparator_);}/*将日志文件编号的标识和名称放入序列化字符串中*/if (has_log_number_) {PutVarint32(dst, kLogNumber);PutVarint64(dst, log_number_);}/*将前一个日志的标识和名称放入序列化字符串中*/if (has_prev_log_number_) {PutVarint32(dst, kPrevLogNumber);PutVarint64(dst, prev_log_number_);}/* 将下一个文件的标识和名称放入序列化字符串中 */if (has_next_file_number_) {PutVarint32(dst, kNextFileNumber);PutVarint64(dst, next_file_number_);}/*上一个序列号的标识和名称放入序列化字符串中*/if (has_last_sequence_) {PutVarint32(dst, kLastSequence);PutVarint64(dst, last_sequence_);}/*将每个压缩点的标识，层次和InternalKey放入序列化字符串*/for (size_t i = 0; i < compact_pointers_.size(); i++) {PutVarint32(dst, kCompactPointer);PutVarint32(dst, compact_pointers_[i].first);  // levelPutLengthPrefixedSlice(dst, compact_pointers_[i].second.Encode());}/*删除文件部分*/for (DeletedFileSet::const_iterator iter = deleted_files_.begin();iter != deleted_files_.end();++iter) {PutVarint32(dst, kDeletedFile);PutVarint32(dst, iter->first);   // levelPutVarint64(dst, iter->second);  // file number}/*新增文件部分*/for (size_t i = 0; i < new_files_.size(); i++) {const FileMetaData& f = new_files_[i].second;PutVarint32(dst, kNewFile);PutVarint32(dst, new_files_[i].first);  // levelPutVarint64(dst, f.number);PutVarint64(dst, f.file_size);PutLengthPrefixedSlice(dst, f.smallest.Encode());PutLengthPrefixedSlice(dst, f.largest.Encode());}
}

Status VersionSet::Recover(bool *save_manifest) {struct LogReporter : public log::Reader::Reporter {Status* status;virtual void Corruption(size_t bytes, const Status& s) {if (this->status->ok()) *this->status = s;}};// Read "CURRENT" file, which contains a pointer to the current manifest filestd::string current;Status s = ReadFileToString(env_, CurrentFileName(dbname_), &current);if (!s.ok()) {return s;}if (current.empty() || current[current.size()-1] != '\n') {return Status::Corruption("CURRENT file does not end with newline");}current.resize(current.size() - 1);/*CURRENT文件记录着MANIFEST的文件名字，名字为MANIFEST-number */std::string dscname = dbname_ + "/" + current;SequentialFile* file;s = env_->NewSequentialFile(dscname, &file);if (!s.ok()) {return s;}bool have_log_number = false;bool have_prev_log_number = false;bool have_next_file = false;bool have_last_sequence = false;uint64_t next_file = 0;uint64_t last_sequence = 0;uint64_t log_number = 0;uint64_t prev_log_number = 0;Builder builder(this, current_);{LogReporter reporter;reporter.status = &s;log::Reader reader(file, &reporter, true/*checksum*/, 0/*initial_offset*/);Slice record;std::string scratch;/*文件中一条记录一条记录的读出，并DecodeFrom，生成VersionEdit */while (reader.ReadRecord(&record, &scratch) && s.ok()) {VersionEdit edit;s = edit.DecodeFrom(record);if (s.ok()) {if (edit.has_comparator_ &&edit.comparator_ != icmp_.user_comparator()->Name()) {s = Status::InvalidArgument(edit.comparator_ + " does not match existing comparator ",icmp_.user_comparator()->Name());}}if (s.ok()) {/*按照次序，讲Verison的变化量层层回放，最重会得到最终版本的Version*/builder.Apply(&edit);}if (edit.has_log_number_) {log_number = edit.log_number_;have_log_number = true;}if (edit.has_prev_log_number_) {prev_log_number = edit.prev_log_number_;have_prev_log_number = true;}if (edit.has_next_file_number_) {next_file = edit.next_file_number_;have_next_file = true;}if (edit.has_last_sequence_) {last_sequence = edit.last_sequence_;have_last_sequence = true;}}}delete file;file = NULL;if (s.ok()) {if (!have_next_file) {s = Status::Corruption("no meta-nextfile entry in descriptor");} else if (!have_log_number) {s = Status::Corruption("no meta-lognumber entry in descriptor");} else if (!have_last_sequence) {s = Status::Corruption("no last-sequence-number entry in descriptor");}if (!have_prev_log_number) {prev_log_number = 0;}MarkFileNumberUsed(prev_log_number);MarkFileNumberUsed(log_number);}if (s.ok()) {Version* v = new Version(this);/*通过回放所有的VersionEdit，得到最终版本的Version，存入v*/builder.SaveTo(v);// Install recovered versionFinalize(v);/* AppendVersion将版本v放入VersionSet集合，同时设置curret_等于v */AppendVersion(v);manifest_file_number_ = next_file;next_file_number_ = next_file + 1;last_sequence_ = last_sequence;log_number_ = log_number;prev_log_number_ = prev_log_number;// See if we can reuse the existing MANIFEST file.if (ReuseManifest(dscname, current)) {// No need to save new manifest} else {*save_manifest = true;}}return s;
}

Status VersionEdit::DecodeFrom(const Slice& src) {Clear();Slice input = src;const char* msg = NULL;uint32_t tag;// Temporary storage for parsingint level;uint64_t number;FileMetaData f;Slice str;InternalKey key;while (msg == NULL && GetVarint32(&input, &tag)) {switch (tag) {case kComparator:if (GetLengthPrefixedSlice(&input, &str)) {comparator_ = str.ToString();has_comparator_ = true;} else {msg = "comparator name";}break;case kLogNumber:if (GetVarint64(&input, &log_number_)) {has_log_number_ = true;} else {msg = "log number";}break;case kPrevLogNumber:if (GetVarint64(&input, &prev_log_number_)) {has_prev_log_number_ = true;} else {msg = "previous log number";}break;case kNextFileNumber:if (GetVarint64(&input, &next_file_number_)) {has_next_file_number_ = true;} else {msg = "next file number";}break;case kLastSequence:if (GetVarint64(&input, &last_sequence_)) {has_last_sequence_ = true;} else {msg = "last sequence number";}break;case kCompactPointer:if (GetLevel(&input, &level) &&GetInternalKey(&input, &key)) {compact_pointers_.push_back(std::make_pair(level, key));} else {msg = "compaction pointer";}break;case kDeletedFile:if (GetLevel(&input, &level) &&GetVarint64(&input, &number)) {deleted_files_.insert(std::make_pair(level, number));} else {msg = "deleted file";}break;case kNewFile:if (GetLevel(&input, &level) &&GetVarint64(&input, &f.number) &&GetVarint64(&input, &f.file_size) &&GetInternalKey(&input, &f.smallest) &&GetInternalKey(&input, &f.largest)) {new_files_.push_back(std::make_pair(level, f));} else {msg = "new-file entry";}break;default:msg = "unknown tag";break;}}if (msg == NULL && !input.empty()) {msg = "invalid tag";}Status result;if (msg != NULL) {result = Status::Corruption("VersionEdit", msg);}return result;
}

// Unlock during expensive MANIFEST log write{mu->Unlock();// Write new record to MANIFEST logif (s.ok()) {std::string record;/*将当前VersionEdit Encode，作为记录，写入MANIFEST*/edit->EncodeTo(&record); s = descriptor_log_->AddRecord(record);if (s.ok()) {/*调用Sync持久化*/s = descriptor_file_->Sync();}if (!s.ok()) {Log(options_->info_log, "MANIFEST write: %s\n", s.ToString().c_str());}}// If we just created a new descriptor file, install it by writing a// new CURRENT file that points to it.if (s.ok() && !new_manifest_file.empty()) {s = SetCurrentFile(env_, dbname_, manifest_file_number_);}mu->Lock();}// Install the new versionif (s.ok()) {/* 新的版本已经生成(通过current_和VersionEdit)， VersionEdit也已经写入MANIFEST，* 此时可以将v设置成current_, 同时将最新的version v链入VersionSet的双向链表*/AppendVersion(v);log_number_ = edit->log_number_;prev_log_number_ = edit->prev_log_number_;} else {delete v;if (!new_manifest_file.empty()) {delete descriptor_log_;delete descriptor_file_;descriptor_log_ = NULL;descriptor_file_ = NULL;env_->DeleteFile(new_manifest_file);}}return s;

void VersionSet::AppendVersion(Version* v) {// Make "v" currentassert(v->refs_ == 0);assert(v != current_);if (current_ != NULL) {current_->Unref();}current_ = v;v->Ref();// Append to linked listv->prev_ = dummy_versions_.prev_;v->next_ = &dummy_versions_;v->prev_->next_ = v;v->next_->prev_ = v;
}

import leveldb
ret ＝ leveldb.RepairDB('/data/mon.iecvq/store.db')

sst_0
sst_1
...
sst_6

BUG #5175 leveldb: LOG and MANIFEST file grow without bound (LOG being _text_ log !)Descriptionleveldb has two files that seem to grow without bound and are only cleared on db open.The first is the LOG file which is a textual debug log that leveldb creates. It's opened when the db is opened and is never closed or cycled (and leveldb doesn't have any method to make it cycle other than close/reopen).To make matter worse, if this file is on XFS, it's subject to XFS pre-allocation (so each time the file size reaches the currently on-disk allocated size, the allocated size doubles, and so you get used disk space of 2M/4M/.../256M/512M giving big jumps in used disk space as reported by 'du -sh')The second file is the MANIFEST file which grows a little at each compaction and is also only trimmed on db open. For long running (i.e. weeks/month), those can actually grow quite a bit(especially LOG). Note that the same issue exists on OSD as well but they seem to receive a whole lot less updates than mon so it shows less.

bool VersionSet::ReuseManifest(const std::string& dscname,const std::string& dscbase) {if (!options_->reuse_logs) {return false;}FileType manifest_type;uint64_t manifest_number;uint64_t manifest_size;/*如果老的MANIFEST文件太大了，就不在延用，return false*延用还是不延用的关键在如下语句：descriptor_log_ = new log::Writer(descriptor_file_, manifest_size);* 如果dscriptor_log_ 为NULL，当情况有变，发生了版本的跃升，有VersionEdit需要写入的MANIFEST的时候，* 会首先判断descriptor_log_是否为NULL，如果为NULL，表示不要在延用老的MANIFEST了，要另起炉灶* 所谓另起炉灶，即起一个空的MANIFEST，先要记下版本的Snapshot，然后将VersionEdit追加写入*/if (!ParseFileName(dscbase, &manifest_number, &manifest_type) ||manifest_type != kDescriptorFile ||!env_->GetFileSize(dscname, &manifest_size).ok() ||// Make new compacted MANIFEST if old one is too bigmanifest_size >= TargetFileSize(options_)) {return false;}assert(descriptor_file_ == NULL);assert(descriptor_log_ == NULL);Status r = env_->NewAppendableFile(dscname, &descriptor_file_);if (!r.ok()) {Log(options_->info_log, "Reuse MANIFEST: %s\n", r.ToString().c_str());assert(descriptor_file_ == NULL);return false;}Log(options_->info_log, "Reusing MANIFEST %s\n", dscname.c_str());descriptor_log_ = new log::Writer(descriptor_file_, manifest_size);manifest_file_number_ = manifest_number;return true;
}

  descriptor_log_ = new log::Writer(descriptor_file_, manifest_size);manifest_file_number_ = manifest_number;

/* descriptor_log_ == NULL 对应的是不延用老的MANIFEST文件 */
if (descriptor_log_ == NULL) {// No reason to unlock *mu here since we only hit this path in the// first call to LogAndApply (when opening the database).assert(descriptor_file_ == NULL);new_manifest_file = DescriptorFileName(dbname_, manifest_file_number_);edit->SetNextFile(next_file_number_);s = env_->NewWritableFile(new_manifest_file, &descriptor_file_);if (s.ok()) {descriptor_log_ = new log::Writer(descriptor_file_);/*当前的版本情况打个快照，作为新MANIFEST的新起点8/s = WriteSnapshot(descriptor_log_);}}// Unlock during expensive MANIFEST log write{mu->Unlock();// Write new record to MANIFEST logif (s.ok()) {std::string record;edit->EncodeTo(&record);s = descriptor_log_->AddRecord(record);if (s.ok()) {s = descriptor_file_->Sync();}if (!s.ok()) {Log(options_->info_log, "MANIFEST write: %s\n", s.ToString().c_str());}}// If we just created a new descriptor file, install it by writing a// new CURRENT file that points to it.if (s.ok() && !new_manifest_file.empty()) {s = SetCurrentFile(env_, dbname_, manifest_file_number_);}mu->Lock();}