leveldb深度剖析-初始化流程

本篇來介紹一下leveldb初始化流程，初始化流程中涉及了很多內容，例如:文件組織方式，存儲內容序列化和反序列化等等。在瞭解這些內容後，是有助於日後分析存儲流程和壓縮流程的。

一、創建/Open數據庫

leveldb屬於輕量級kv數據庫，可以很方便寫一個hello程序並結合gdb單步調試,來輔助我們閱讀源碼，起到事半功倍的效果。

1.1、版本管理

leveldb採用分層思想對數據進行管理，那麼就需要一個對象來管理每層數據元信息，這個對象就是VersionSet，在源碼中和Version有關的對象一共有三個分別爲：

對象	作用
VersionSet	Version集合，所有的Version都掛在VersionSet對象下面，一個db只有一個VersionSet
Version	一個db可能存在多個Version，Version之間採用鏈表方式管理，鏈表最後節點爲當前最新Version信息
VersionEdit	該對象用於生成最新的Version。VersionEdit + 當前使用的Version 生成 新的Version存放到鏈表尾部，只在需要生成新的Version時纔會創建

1.2、Open函數

調用接口Open，用於創建數據庫(如果數據不存在)，下面是Open函數流程圖：

Status DB::Open(const Options& options, const std::string& dbname,
                DB** dbptr) {
  *dbptr = NULL;

  DBImpl* impl = new DBImpl(options, dbname);//初始化DBImpl對象
  impl->mutex_.Lock();
  VersionEdit edit;
  // Recover handles create_if_missing, error_if_exists 恢復數據
  bool save_manifest = false;
  Status s = impl->Recover(&edit, &save_manifest);//恢復數據
  
  if (s.ok() && impl->mem_ == NULL) {//創建memTable對象
    // Create new log and a corresponding memtable.
    uint64_t new_log_number = impl->versions_->NewFileNumber();
    // 創建*.log文件
    WritableFile* lfile;
    s = options.env->NewWritableFile(LogFileName(dbname, new_log_number),
                                     &lfile);
    if (s.ok()) {
      edit.SetLogNumber(new_log_number);
      impl->logfile_ = lfile;
      impl->logfile_number_ = new_log_number;
      impl->log_ = new log::Writer(lfile);
      impl->mem_ = new MemTable(impl->internal_comparator_);
      impl->mem_->Ref();
    }
  }
  
  if (s.ok() && save_manifest) {//save_manifest爲true表示需要重新創建一個MANIFEST文件
    edit.SetPrevLogNumber(0);  // No older logs needed after recovery.
    edit.SetLogNumber(impl->logfile_number_);
    s = impl->versions_->LogAndApply(&edit, &impl->mutex_);//由於log相關數據變化 所以要寫回Version信息
  }
  
  if (s.ok()) {
    impl->DeleteObsoleteFiles();//刪除舊版本文件
    impl->MaybeScheduleCompaction();//啓動壓縮調度流程 單獨啓動一個線程
  }
  
  impl->mutex_.Unlock();
  
  if (s.ok()) {
    assert(impl->mem_ != NULL);
    *dbptr = impl;
  } else {
    delete impl;
  }
  return s;
}

1.2.1、Recover數據恢復

DBImpl的構造函數實現並不複雜，不再展開說明。這裏詳細說明一下Recover的實現，Recover只做了兩件主要事情：恢復Version信息和恢復MemTable(讀取.log文件)，具體流程圖如下：

/**
 * 數據庫環境恢復
 * @param edit 
 * @param save_manifest 
 */
Status DBImpl::Recover(VersionEdit* edit, bool *save_manifest) {
  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);
  // 創建文件鎖 支持多進程併發訪問同一個數據庫 env_指向env_posix.cc
  Status s = env_->LockFile(LockFileName(dbname_), &db_lock_);
  if (!s.ok()) {
    return s;
  }
  // 判斷CURRENT文件是否存在
  if (!env_->FileExists(CurrentFileName(dbname_))) {
    if (options_.create_if_missing) {
      s = NewDB(); //創建數據庫，主要包含CURRENT、MANIFEST文件，不包含*.log文件
      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)");
    }
  }
  /**
   * 以上是數據文件校驗 如果不存在則創建新的數據庫文件  以下內容是從數據庫文件 
   * 中恢復數據,如版本信息
   */
  s = versions_->Recover(save_manifest);//version_set.cc 
  if (!s.ok()) {
    return s;
  }
  SequenceNumber max_sequence(0);

  // 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();
  //獲取目錄下所有目錄或文件名字
  std::vector<std::string> filenames;
  s = env_->GetChildren(dbname_, &filenames);
  if (!s.ok()) {
    return s;
  }
  std::set<uint64_t> expected;
  versions_->AddLiveFiles(&expected);
  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)) {
      expected.erase(number);
      if (type == kLogFile && ((number >= min_log) || (number == prev_log)))
        logs.push_back(number);
    }
  }
  if (!expected.empty()) {//不空表示有文件丟失 出錯
    char buf[50];
    snprintf(buf, sizeof(buf), "%d missing files; e.g.",
             static_cast<int>(expected.size()));
    return Status::Corruption(buf, TableFileName(dbname_, *(expected.begin())));
  }

  // Recover in the order in which the logs were generated
  // 按從小到大順序讀取.log文件 生成MemTable結構
  std::sort(logs.begin(), logs.end());
  for (size_t i = 0; i < logs.size(); i++) {
    s = RecoverLogFile(logs[i], (i == logs.size() - 1), save_manifest, edit,
                       &max_sequence);
    if (!s.ok()) {
      return s;
    }

    // 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]);
  }

  if (versions_->LastSequence() < max_sequence) {
    versions_->SetLastSequence(max_sequence);//保存最大記錄號
  }

  return Status::OK();
}

這裏需要提示一下：.log文件相當於數據庫操作日誌，.log中每條記錄都與MemTable中記錄一一對應.遍歷完所有.log文件MemTable就生成了。

這裏簡單說明一下數據插入流程:leveldb首先將一條記錄插入到.log文件，然後在插入到MemTable中。當leveldb數據庫重啓時，需要讀取每個.log文件以便恢復MemTable，這個恢復流程就是在RecoverLoFile中實現的。

1.2.2、VersionEdit恢復

在之前的博客介紹了MANIFEST存儲的內容爲VersionEdit，該對象包含了很多內容，例如:log文件編號，壓縮點，待刪文件序號以及保存的最小key和最大key。該函數主要是針對MANIFEST文件的解析

/**
 * 恢復數據 從MANIFEST文件中讀取出VersionEdit來初始化VersionSet
 * @param save_manifest  輸出參數 
 *          true  -  需要創建新的manifest文件
 *          false -  不需要創建新的manifest文件
 * @return 返回操作狀態
 */
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 file  
  std::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);

  //讀取MANIFEST文件
  std::string dscname = dbname_ + "/" + current;
  SequentialFile* file;
  s = env_->NewSequentialFile(dscname, &file);//env_posix.cc
  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;
  // current_ 保存的當前最新的Version信息 後面會調用Builder saveto方法與current_
  // 指定的Version信息進行合併
  Builder builder(this, current_); //Builder類定義在version_set.cc

  {
    LogReporter reporter;
    reporter.status = &s;
    //創建讀取執行器 讀取MANIFEST文件
    log::Reader reader(file, &reporter, true/*checksum*/, 0/*initial_offset*/);
    Slice record;
    std::string scratch;//在ReadRecord內部使用
    // 讀取MAINIFEST文件 逐條讀取反序列化VersionEdit對象
    while (reader.ReadRecord(&record, &scratch) && s.ok()) {//log_reader.cc
      VersionEdit edit;
      s = edit.DecodeFrom(record);//解碼 到VersionEdit
      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());
        }
      }
      //對於過期的VersionEdit對象中最有用的內容就是: 文件信息以及壓縮點,所以需要對其進行整合
      if (s.ok()) {
        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()) {
    //創建新的Version 與舊的Version進行合併，並保存到新的Version中
    Version* v = new Version(this);
    builder.SaveTo(v);
    
    // Install recovered version
    Finalize(v);//預計算壓縮層次以及分數
    AppendVersion(v);// 將version插入到Version雙向鏈表尾部
    // 設置VersionSet中的序號信息
    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;
}

二、總結

本篇了主要介紹leveldb初始化流程，只有瞭解數據庫組織方式，才有助於我們後續分析存儲流程，如果對於leveldb存儲劉結構還不清楚的,可參考本篇《leveldb深度剖析-存儲結構(1)》，下一篇將介紹存儲流程。