PartitioinLeaderSelector分析

版权声明:本文为博主原创文章,未经博主允许不得转载。 https://blog.csdn.net/zhanglh046/article/details/72822066

PartitionLeaderSelector主要负责分区leader副本的选举。

1 NoOpLeaderSelector

def selectLeader(topicAndPartition: TopicAndPartition, currentLeaderAndIsr: LeaderAndIsr): (LeaderAndIsr, Seq[Int]) = {
    warn("I should never have been asked to perform leader election,returning the current LeaderAndIsr and replica assignment.")
    (currentLeaderAndIsr, controllerContext.partitionReplicaAssignment(topicAndPartition))
  }

2 OfflinePartitionLeaderSelector

从该分区的AR副本集中过滤出可用的ISR列表

如果 ISR至少有一个可用副本,则从ISR列表中选举出一个副本作为leader

如果 ISR中没有存活副本,而且如果AR中没有存活的副本抛出异常

如果ISR中没有存活副本,但是AR副本集中有副本,那么就把AR第一个副本作为leader

def selectLeader(topicAndPartition: TopicAndPartition, currentLeaderAndIsr: LeaderAndIsr): (LeaderAndIsr, Seq[Int]) = {
  controllerContext.partitionReplicaAssignment.get(topicAndPartition) match {
    case Some(assignedReplicas) =>
      // 从该分区的AR副本集中过滤出可用的副本
      val liveAssignedReplicas = assignedReplicas.filter(r => controllerContext.liveBrokerIds.contains(r))
      // 从传递进来的ISR列表过滤出可用的副本
      val liveBrokersInIsr = currentLeaderAndIsr.isr.filter(r => controllerContext.liveBrokerIds.contains(r))
      val currentLeaderEpoch = currentLeaderAndIsr.leaderEpoch
      val currentLeaderIsrZkPathVersion = currentLeaderAndIsr.zkVersion
      val newLeaderAndIsr =
        // 如果 ISR中没有存活副本
        if (liveBrokersInIsr.isEmpty) {
          if (!LogConfig.fromProps(config.originals, AdminUtils.fetchEntityConfig(controllerContext.zkUtils,
            ConfigType.Topic, topicAndPartition.topic)).uncleanLeaderElectionEnable) {
            throw new NoReplicaOnlineException(("No broker in ISR for partition " +
              "%s is alive. Live brokers are: [%s],".format(topicAndPartition, controllerContext.liveBrokerIds)) +
              " ISR brokers are: [%s]".format(currentLeaderAndIsr.isr.mkString(",")))
          }
          debug("No broker in ISR is alive for %s. Pick the leader from the alive assigned replicas: %s"
            .format(topicAndPartition, liveAssignedReplicas.mkString(",")))
          // 而且如果AR中没有存活的副本
          if (liveAssignedReplicas.isEmpty) {
            throw new NoReplicaOnlineException(("No replica for partition " +
              "%s is alive. Live brokers are: [%s],".format(topicAndPartition, controllerContext.liveBrokerIds)) +
              " Assigned replicas are: [%s]".format(assignedReplicas))
          } else {// 如果AR中有存活的副本
            ControllerStats.uncleanLeaderElectionRate.mark()
            // AR的存活的副本中取出第一个副本作为leader,这样的话就是有数据丢失的风险
            val newLeader = liveAssignedReplicas.head
            warn("No broker in ISR is alive for %s. Elect leader %d from live brokers %s. There's potential data loss."
              .format(topicAndPartition, newLeader, liveAssignedReplicas.mkString(",")))
            new LeaderAndIsr(newLeader, currentLeaderEpoch + 1, List(newLeader), currentLeaderIsrZkPathVersion + 1)
          }
        } else {// 如果 ISR有存活副本
          // AR存活的副本列表过滤出ISR存活的副本列表
          val liveReplicasInIsr = liveAssignedReplicas.filter(r => liveBrokersInIsr.contains(r))
          // ISR中存活的副本列表第一个作为新的leader
          val newLeader = liveReplicasInIsr.head
          debug("Some broker in ISR is alive for %s. Select %d from ISR %s to be the leader."
            .format(topicAndPartition, newLeader, liveBrokersInIsr.mkString(",")))
          new LeaderAndIsr(newLeader, currentLeaderEpoch + 1, liveBrokersInIsr.toList, currentLeaderIsrZkPathVersion + 1)
        }
      info("Selected new leader and ISR %s for offline partition %s".format(newLeaderAndIsr.toString(), topicAndPartition))
      (newLeaderAndIsr, liveAssignedReplicas)
    case None =>
      throw new NoReplicaOnlineException("Partition %s doesn't have replicas assigned to it".format(topicAndPartition))
  }
}

 

3 ReassignedPartitionLeaderSelector

获取正在重新分配的副本,然后从可用ISR列表中选举出一个leader,然后将当前的ISR更新为新ISR,然后将重新分配非副本集合作为接受LeaderAndIsr请求的副本集

def selectLeader(topicAndPartition: TopicAndPartition, currentLeaderAndIsr: LeaderAndIsr): (LeaderAndIsr, Seq[Int]) = {
  // ControllerContext获取该分区正在重新分配的副本
  val reassignedInSyncReplicas = controllerContext.partitionsBeingReassigned(topicAndPartition).newReplicas
  val currentLeaderEpoch = currentLeaderAndIsr.leaderEpoch
  val currentLeaderIsrZkPathVersion = currentLeaderAndIsr.zkVersion
  // 过滤出当前可用的副本且ISR列表包含该副本
  val aliveReassignedInSyncReplicas = reassignedInSyncReplicas.filter(r => controllerContext.liveBrokerIds.contains(r) &&
                                                                           currentLeaderAndIsr.isr.contains(r))
  val newLeaderOpt = aliveReassignedInSyncReplicas.headOption
  newLeaderOpt match {
    case Some(newLeader) => (new LeaderAndIsr(newLeader, currentLeaderEpoch + 1, currentLeaderAndIsr.isr,
      currentLeaderIsrZkPathVersion + 1), reassignedInSyncReplicas)
    case None =>
      reassignedInSyncReplicas.size match {
        case 0 =>
          throw new NoReplicaOnlineException("List of reassigned replicas for partition " +
            " %s is empty. Current leader and ISR: [%s]".format(topicAndPartition, currentLeaderAndIsr))
        case _ =>
          throw new NoReplicaOnlineException("None of the reassigned replicas for partition " +
            "%s are in-sync with the leader. Current leader and ISR: [%s]".format(topicAndPartition, currentLeaderAndIsr))
      }
  }
}

 

4 PreferredReplicaPartitionLeaderSelector

先取出当前分区分配的所有副本(AR),可能包括不可用,然后取出第一个副本,如果第一个副本就是leader抛出异常,否则如果该副本所在broker是存活的且它在ISR列表中,就把他作为leader,否则抛出异常

def selectLeader(topicAndPartition: TopicAndPartition, currentLeaderAndIsr: LeaderAndIsr): (LeaderAndIsr, Seq[Int]) = {
  // 获取该分区的AR副本集
  val assignedReplicas = controllerContext.partitionReplicaAssignment(topicAndPartition)
  // 获取副本集第一个副本
  val preferredReplica = assignedReplicas.head
  // 检查该副本是不是leader,如果是leader抛出异常
  val currentLeader = controllerContext.partitionLeadershipInfo(topicAndPartition).leaderAndIsr.leader
  if (currentLeader == preferredReplica) {
    throw new LeaderElectionNotNeededException("Preferred replica %d is already the current leader for partition %s"
                                                 .format(preferredReplica, topicAndPartition))
  } else {
    info("Current leader %d for partition %s is not the preferred replica.".format(currentLeader, topicAndPartition) +
      " Trigerring preferred replica leader election")
    // 在检查该副本是不是可用的且它在ISR列表中
    if (controllerContext.liveBrokerIds.contains(preferredReplica) && currentLeaderAndIsr.isr.contains(preferredReplica)) {
      (new LeaderAndIsr(preferredReplica, currentLeaderAndIsr.leaderEpoch + 1, currentLeaderAndIsr.isr,
        currentLeaderAndIsr.zkVersion + 1), assignedReplicas)
    } else {
      throw new StateChangeFailedException("Preferred replica %d for partition ".format(preferredReplica) +
        "%s is either not alive or not in the isr. Current leader and ISR: [%s]".format(topicAndPartition, currentLeaderAndIsr))
    }
  }
}

 

5 ControlledShutdownLeaderSelector

将ISR中处于关闭状态的副本从集合中去除掉,返回一个新新的ISR集合,然后选取第一个副本作为leader,然后令当前AR作为接收LeaderAndIsr请求的副本

 

def selectLeader(topicAndPartition: TopicAndPartition, currentLeaderAndIsr: LeaderAndIsr): (LeaderAndIsr, Seq[Int]) = {
  val currentLeaderEpoch = currentLeaderAndIsr.leaderEpoch
  val currentLeaderIsrZkPathVersion = currentLeaderAndIsr.zkVersion
  // 获取leader
  val currentLeader = currentLeaderAndIsr.leader
  // 获取该分区的AR副本集
  val assignedReplicas = controllerContext.partitionReplicaAssignment(topicAndPartition)
  // 获取可用或者正处于关闭的broker
  val liveOrShuttingDownBrokerIds = controllerContext.liveOrShuttingDownBrokerIds

  val liveAssignedReplicas = assignedReplicas.filter(r => liveOrShuttingDownBrokerIds.contains(r))
  // 过滤掉正处于关闭的副本
  val newIsr = currentLeaderAndIsr.isr.filter(brokerId => !controllerContext.shuttingDownBrokerIds.contains(brokerId))
  liveAssignedReplicas.find(newIsr.contains) match {
    case Some(newLeader) =>
      debug("Partition %s : current leader = %d, new leader = %d".format(topicAndPartition, currentLeader, newLeader))
      (LeaderAndIsr(newLeader, currentLeaderEpoch + 1, newIsr, currentLeaderIsrZkPathVersion + 1), liveAssignedReplicas)
    case None =>
      throw new StateChangeFailedException(("No other replicas in ISR %s for %s besides" +
        " shutting down brokers %s").format(currentLeaderAndIsr.isr.mkString(","), topicAndPartition, controllerContext.shuttingDownBrokerIds.mkString(",")))
  }
}

 

 

阅读更多
想对作者说点什么?

博主推荐

换一批

没有更多推荐了,返回首页