一.前言

ZKFailoverController作为主备namenode的控制器，负责对namenode进行监控和主备切换，目前这个切换是依赖于zookeeper的，当然目前也支持手工切换主备namenode。

二.启动入口

启动入口其实就是一个main函数.也就是说ZKFailoverController就是一个jar程序.
实现类是DFSZKFailoverController

public static void main(String args[])
      throws Exception {
    
    StringUtils.startupShutdownMessage(DFSZKFailoverController.class,
        args, LOG);
    if (DFSUtil.parseHelpArgument(args, 
        ZKFailoverController.USAGE, System.out, true)) {
    
      System.exit(0);
    }
    
    GenericOptionsParser parser = new GenericOptionsParser(
        new HdfsConfiguration(), args);
    try {
    
      //构建DFSZKFailoverController
      DFSZKFailoverController zkfc = DFSZKFailoverController.create(  parser.getConfiguration());
      
      // zkfc.run 启动
      System.exit(zkfc.run(parser.getRemainingArgs()));
    } catch (Throwable t) {
    
      LOG.error("DFSZKFailOverController exiting due to earlier exception "
          + t);
      terminate(1, t);
    }
  }

三. doRun

doRun里面有四个方法会比较重要

3.1.initZK 初始化ZK,构建连接信息

  private void initZK() throws HadoopIllegalArgumentException, IOException,
      KeeperException {
    

    // 获取zk 集群信息
    zkQuorum = conf.get(ZK_QUORUM_KEY);

    // zk超时时间
    // ha.zookeeper.session-timeout.ms : 10ms
    int zkTimeout = conf.getInt(ZK_SESSION_TIMEOUT_KEY,
        ZK_SESSION_TIMEOUT_DEFAULT);
    // Parse ACLs from configuration.

    // zookeeper ACL认证
    // ha.zookeeper.acl : world:anyone:rwcda
    String zkAclConf = conf.get(ZK_ACL_KEY, ZK_ACL_DEFAULT);
    zkAclConf = ZKUtil.resolveConfIndirection(zkAclConf);

    List<ACL> zkAcls = ZKUtil.parseACLs(zkAclConf);
    if (zkAcls.isEmpty()) {
    
      zkAcls = Ids.CREATOR_ALL_ACL;
    }
    // 解析授权
    // Parse authentication from configuration.
    List<ZKAuthInfo> zkAuths = SecurityUtil.getZKAuthInfos(conf, ZK_AUTH_KEY);

    // Sanity check configuration.
    Preconditions.checkArgument(zkQuorum != null,
        "Missing required configuration '%s' for ZooKeeper quorum",
        ZK_QUORUM_KEY);
    Preconditions.checkArgument(zkTimeout > 0,
        "Invalid ZK session timeout %s", zkTimeout);

    // 最大重试次数 3
    // ha.failover-controller.active-standby-elector.zk.op.retries :  3
    int maxRetryNum = conf.getInt(
        CommonConfigurationKeys.HA_FC_ELECTOR_ZK_OP_RETRIES_KEY,
        CommonConfigurationKeys.HA_FC_ELECTOR_ZK_OP_RETRIES_DEFAULT);

    // getParentZnode : /hadoop-ha/ [namenode serviceId]
    // 构建ActiveStandbyElector
    elector = new ActiveStandbyElector(zkQuorum,
        zkTimeout, getParentZnode(), zkAcls, zkAuths,
        new ElectorCallbacks(), maxRetryNum);
  }

3.2.formatZK(force, interactive);

格式化zk,其实就是构建工作空间.
默认工作空间为: ** /hadoop-ha/[namenode cluster id]**

  private int formatZK(boolean force, boolean interactive)
      throws IOException, InterruptedException, KeeperException {
    
    if (elector.parentZNodeExists()) {
    
      if (!force && (!interactive || !confirmFormat())) {
    
        return ERR_CODE_FORMAT_DENIED;
      }
      
      try {
    
        elector.clearParentZNode();
      } catch (IOException e) {
    
        LOG.error("Unable to clear zk parent znode", e);
        return 1;
      }
    }
    
    elector.ensureParentZNode();
    return 0;
  }

3.3. initRPC() 初始化ZKFCRpcServer

  protected void initRPC() throws IOException {
    
    InetSocketAddress bindAddr = getRpcAddressToBindTo();
    rpcServer = new ZKFCRpcServer(conf, bindAddr, this, getPolicyProvider());
  }

3.4. initHM 开启健康检查 HealthMonitor

健康检查是一个定时线程,会定时检查所属namenode的健康状态,获取namonode的状态之后,如果健康的话,会尝试获取zk锁,如果拿到zk锁的话,该zkfc所在的namenode会自动进入active状态.否则依旧是standy状态.

  private void initHM() {
    
    healthMonitor = new HealthMonitor(conf, localTarget);
    healthMonitor.addCallback(new HealthCallbacks());
    healthMonitor.addServiceStateCallback(new ServiceStateCallBacks());
    healthMonitor.start();
  }

3.5. 启动ZKFCRpcServer

  startRPC();

  protected void startRPC() throws IOException {
    
    rpcServer.start();
  }

四. HealthMonitor

健康检查(HealthMonitor)是一个线程,会定时与namenode进行通讯,然后根据反馈的namenode状态进行不同的操作. 默认检查周期 1秒 ( ha.health-monitor.check-interval.ms : 1000 )

核心的方法为: recheckElectability

  /**
   * Check the current state of the service, and join the election
   * if it should be in the election.
   */
  private void recheckElectability() {
    
    // Maintain lock ordering of elector -> ZKFC
    synchronized (elector) {
    
      synchronized (this) {
    
        boolean healthy = lastHealthState == State.SERVICE_HEALTHY;
    
        long remainingDelay = delayJoiningUntilNanotime - System.nanoTime(); 
        if (remainingDelay > 0) {
    
          if (healthy) {
    
            LOG.info("Would have joined master election, but this node is " +
                "prohibited from doing so for " +
                TimeUnit.NANOSECONDS.toMillis(remainingDelay) + " more ms");
          }
          scheduleRecheck(remainingDelay);
          return;
        }
    
        switch (lastHealthState) {
    
        // namenode健康, 自动参与选举.尝试获取zk锁.
        case SERVICE_HEALTHY:
          elector.joinElection(targetToData(localTarget));
          if (quitElectionOnBadState) {
    
            quitElectionOnBadState = false;
          }
          break;
          
        case INITIALIZING:
          LOG.info("Ensuring that " + localTarget + " does not " +
              "participate in active master election");
          elector.quitElection(false);
          serviceState = HAServiceState.INITIALIZING;
          break;
    
        case SERVICE_UNHEALTHY:
        case SERVICE_NOT_RESPONDING:
          LOG.info("Quitting master election for " + localTarget +
              " and marking that fencing is necessary");
          elector.quitElection(true);
          serviceState = HAServiceState.INITIALIZING;
          break;
          
        case HEALTH_MONITOR_FAILED:
          fatalError("Health monitor failed!");
          break;
          
        default:
          throw new IllegalArgumentException("Unhandled state:"
                                               + lastHealthState);
        }
      }
    }
  }