场景： 3 节点集群 (b1, b2, b3)，分区 tp1 的 isr[1, 2, 3]，leader 是 1，现在希望把 tp1 的 leader 切换为 3，怎么操作？

1. 通过 zk 客户端，修改 zk 上 tp1 的 isr 列表为 [3, 2, 1]

2. 执行命令行 kafka-preferred-replica-election.sh，把需要修改 leader 的分区信息写入到 zk 的 /admin/preferred_replica_election 节点

// kafka.admin.PreferredReplicaLeaderElectionCommand#writePreferredReplicaElectionData
  def writePreferredReplicaElectionData(zkClient: KafkaZkClient,
                                        partitionsUndergoingPreferredReplicaElection: Set[TopicPartition]) {
    try {
      zkClient.createPreferredReplicaElection(partitionsUndergoingPreferredReplicaElection.toSet)
      println("Created preferred replica election path with %s".format(partitionsUndergoingPreferredReplicaElection.mkString(",")))
    } catch {
      case _: NodeExistsException =>
        throw new AdminOperationException("Preferred replica leader election currently in progress for " +
          "%s. Aborting operation".format(zkClient.getPreferredReplicaElection.mkString(",")))
      case e2: Throwable => throw new AdminOperationException(e2.toString)
    }
  }

3. KafkaController 监听 /admin/preferred_replica_election 节点，重新选举 leader，取 isr 中第一个副本为 leader，发送 LeaderAndIsrRequest 请求给其他 broker，通知他们转变副本角色

// kafka.controller.KafkaController.PreferredReplicaLeaderElection
  case object PreferredReplicaLeaderElection extends ControllerEvent {
    override def state: ControllerState = ControllerState.ManualLeaderBalance

    override def process(): Unit = {
      if (!isActive) return

      // We need to register the watcher if the path doesn't exist in order to detect future preferred replica
      // leader elections and we get the `path exists` check for free
      if (zkClient.registerZNodeChangeHandlerAndCheckExistence(preferredReplicaElectionHandler)) {
        val partitions = zkClient.getPreferredReplicaElection
        val partitionsForTopicsToBeDeleted = partitions.filter(p => topicDeletionManager.isTopicQueuedUpForDeletion(p.topic))
        if (partitionsForTopicsToBeDeleted.nonEmpty) {
          error(s"Skipping preferred replica election for partitions $partitionsForTopicsToBeDeleted since the " +
            "respective topics are being deleted")
        }
        onPreferredReplicaElection(partitions -- partitionsForTopicsToBeDeleted)
      }
    }
  }
  
// kafka.controller.KafkaController#onPreferredReplicaElection
  private def onPreferredReplicaElection(partitions: Set[TopicPartition], isTriggeredByAutoRebalance: Boolean = false) {
    info(s"Starting preferred replica leader election for partitions ${partitions.mkString(",")}")
    try {
      partitionStateMachine.handleStateChanges(partitions.toSeq, OnlinePartition, Option(PreferredReplicaPartitionLeaderElectionStrategy))
    } catch {
      case e: Throwable => error(s"Error completing preferred replica leader election for partitions ${partitions.mkString(",")}", e)
    } finally {
      removePartitionsFromPreferredReplicaElection(partitions, isTriggeredByAutoRebalance)
    }
  }