Kafka源码分析

本文主要针对于Kafka的源码进行分析，版本为kafka-0.8.2.1。 由于时间有限，可能更新比较慢...

Kafka.scala

// 读取配置文件
val props = Utils.loadProps(args(0))
val serverConfig = new KafkaConfig(props)
KafkaMetricsReporter.startReporters(serverConfig.props)

val kafkaServerStartable = new KafkaServerStartable(serverConfig)

// 注册一个关闭钩子，当JVM关闭时调用KafkaServerStartable.shutdown
Runtime.getRuntime().addShutdownHook(new Thread() {
  override def run() = kafkaServerStartable.shutdown
})  

// 运行并等待结束
kafkaServerStartable.startup
kafkaServerStartable.awaitShutdown

　　

Server

实际调用类为KafkaServer

def startup() {
  kafkaScheduler.startup()

  // 初始化Zookeeper内相关路径
  zkClient = initZk()

  // 日志管理器
  logManager = createLogManager(zkClient, brokerState)
  logManager.startup()

  socketServer = new SocketServer(...)
  socketServer.startup()

  // 启动副本管理器
  replicaManager = new ReplicaManager(config, time, zkClient, kafkaScheduler, logManager, isShuttingDown)

  // 创建偏移量管理器
  offsetManager = createOffsetManager()

  // 实例化调度器
  kafkaController = new KafkaController(config, zkClient, brokerState)

  // 请求处理器
  apis = new KafkaApis(...)

  // 网络请求处理
  requestHandlerPool = new KafkaRequestHandlerPool(config.brokerId, socketServer.requestChannel, apis, config.numIoThreads)
  brokerState.newState(RunningAsBroker)

  Mx4jLoader.maybeLoad()
  replicaManager.startup()
  kafkaController.startup()

  // Topic配置管理器
  topicConfigManager = new TopicConfigManager(zkClient, logManager)
  topicConfigManager.startup()

  // Broker的心跳检查
  kafkaHealthcheck = new KafkaHealthcheck(...)
  kafkaHealthcheck.startup()

  registerStats()
  startupComplete.set(true)
  info("started")
}

在KafkaServer的startup中看到主要进行几个主要服务的初始化和启动。

private def initZk(): ZkClient =
{
  info("Connecting to zookeeper on " + config.zkConnect)

  // Kafka在Zookeeper中的工作根目录
  val chroot = {
    if (config.zkConnect.indexOf("/") > 0)
      config.zkConnect.substring(config.zkConnect.indexOf("/"))
    else
      ""
  }
  // 创建工作根目录
  if (chroot.length > 1) {
    val zkConnForChrootCreation = config.zkConnect.substring(0, config.zkConnect.indexOf("/"))
    val zkClientForChrootCreation = new ZkClient(...)
    ZkUtils.makeSurePersistentPathExists(zkClientForChrootCreation, chroot)
    info("Created zookeeper path " + chroot)
    zkClientForChrootCreation.close()
  }

  // 实例化ZkClient
  val zkClient = new ZkClient(config.zkConnect, config.zkSessionTimeoutMs, config.zkConnectionTimeoutMs, ZKStringSerializer)
  // 在Zookeeper中创建必要持久路径
  ZkUtils.setupCommonPaths(zkClient)
  zkClient
}

　　

KafkaScheduler实际为对线程池ScheduledThreadPoolExecutor的封装，这里不做过多的分析。

KafkaHealthcheck(...)
{
  val brokerIdPath = ZkUtils.BrokerIdsPath + "/" + brokerId
  val sessionExpireListener = new SessionExpireListener

  def startup()
  {
    // 注册一个Zookeeper事件(状态)监听器
    zkClient.subscribeStateChanges(sessionExpireListener)
    // 在Zookeeper的/brokers/ids/id目录创建临时节点并写入节点信息
    register()
  }
}

IZkStateListener 定义了两种事件：一种是连接状态的改变，例如由未连接改变成连接上，连接上改为过期等；

另一种创建一个新的session（连接）， 通常是由于session失效然后新的session被建立时触发。

class SessionExpireListener() extends IZkStateListener
{
  @throws(classOf[Exception])
  def handleStateChanged(state: KeeperState) {}

  @throws(classOf[Exception])
  def handleNewSession() = register()
}

　　

ReplicaManager

def startup()
{
  scheduler.schedule("isr-expiration", maybeShrinkIsr, period = config.replicaLagTimeMaxMs, unit = TimeUnit.MILLISECONDS)
}
// 定时调用maybeShrinkIsr
private def maybeShrinkIsr(): Unit =
{
  trace("Evaluating ISR list of partitions to see which replicas can be removed from the ISR")
  allPartitions.values.foreach(partition => partition.maybeShrinkIsr(config.replicaLagTimeMaxMs, config.replicaLagMaxMessages))
}

　　

这里调用了cluster.Partition中的maybeShrinkIsr来将卡住的或者低效的副本从ISR中去除并更新HighWatermark。

def maybeShrinkIsr(replicaMaxLagTimeMs: Long,  replicaMaxLagMessages: Long)
{
  inWriteLock(leaderIsrUpdateLock) {
    leaderReplicaIfLocal() match {
      case Some(leaderReplica) =>
      // 找出卡住和低效的Replica并从ISR中去除
        val outOfSyncReplicas = getOutOfSyncReplicas(leaderReplica, replicaMaxLagTimeMs, replicaMaxLagMessages)
        if(outOfSyncReplicas.size > 0) {
          val newInSyncReplicas = inSyncReplicas -- outOfSyncReplicas
          assert(newInSyncReplicas.size > 0)
          // 更新ZK中的ISR
          updateIsr(newInSyncReplicas)
          // 计算HW并更新
          maybeIncrementLeaderHW(leaderReplica)
          replicaManager.isrShrinkRate.mark()
        }
  ...
}

def getOutOfSyncReplicas(leaderReplica: Replica, keepInSyncTimeMs: Long, keepInSyncMessages: Long): Set[Replica] =
{
  // Leader的最后写入偏移量
  val leaderLogEndOffset = leaderReplica.logEndOffset
  // ISR中排除LeaderReplica的其他集合
  val candidateReplicas = inSyncReplicas - leaderReplica
  // 卡住的Replica集合
  val stuckReplicas = candidateReplicas.filter(r => (time.milliseconds - r.logEndOffsetUpdateTimeMs) > keepInSyncTimeMs)
  // 低效的Replica
  // 条件1 Replicas的offset > 0
  // 条件2 Leader的offset - Replicas的offset > 阀值
  val slowReplicas = candidateReplicas.filter(r =>
    r.logEndOffset.messageOffset >= 0 &&
    leaderLogEndOffset.messageOffset - r.logEndOffset.messageOffset > keepInSyncMessages)
  // 返回卡住的和低效的Replicas
  stuckReplicas ++ slowReplicas
}

　　

Cluster

Controller