Kafka原理-分区leader选举
0.说明
kafka源码版本为1.0
1.分区状态
kafka源码定义了4种状态
NewPartition： 表示正在创建新的分区， 是一个中间状态,只是在Controller的内存中存了状态信息
OnlinePartition： 表示在线状态, 只有在线的分区才能提供服务.
OfflinePartition： 表示下线状态, 分区可能因为Broker宕机或者删除Topic等原因流转到这个状态, 下线后不能提供服务
NonExistentPartition： 表示分区不存在
2.选举源码分析
源码入口PartitionStateMachine#electLeaderForPartition
注释说明leader选举发生在OfflinePartition,OnlinePartition->OnlinePartition状态变更的时候
/** * Invoked on the OfflinePartition,OnlinePartition->OnlinePartition state change. * It invokes the leader election API to elect a leader for the input offline partition * @param topic The topic of the offline partition * @param partition The offline partition * @param leaderSelector Specific leader selector (e.g., offline/reassigned/etc.) */ def electLeaderForPartition(topic: String, partition: Int, leaderSelector: PartitionLeaderSelector) { val topicAndPartition = TopicAndPartition(topic, partition) val stateChangeLog = stateChangeLogger.withControllerEpoch(controller.epoch) // handle leader election for the partitions whose leader is no longer alive stateChangeLog.trace(s"Started leader election for partition $topicAndPartition") try { var zookeeperPathUpdateSucceeded: Boolean = false var newLeaderAndIsr: LeaderAndIsr = null var replicasForThisPartition: Seq[Int] = Seq.empty[Int] while(!zookeeperPathUpdateSucceeded) {
// 01：从zk种获取分区元数据 val currentLeaderIsrAndEpoch = getLeaderIsrAndEpochOrThrowException(topic, partition) val currentLeaderAndIsr = currentLeaderIsrAndEpoch.leaderAndIsr val controllerEpoch = currentLeaderIsrAndEpoch.controllerEpoch
// 02：这里表示其他controller成为新的首领，旧的请求抛异常 if (controllerEpoch > controller.epoch) { val failMsg = s"Aborted leader election for partition $topicAndPartition since the LeaderAndIsr path was " + s"already written by another controller. This probably means that the current controller $controllerId went " + s"through a soft failure and another controller was elected with epoch $controllerEpoch." stateChangeLog.error(failMsg) throw new StateChangeFailedException(stateChangeLog.messageWithPrefix(failMsg)) } // 03：选举的实现 val (leaderAndIsr, replicas) = leaderSelector.selectLeader(topicAndPartition, currentLeaderAndIsr) val (updateSucceeded, newVersion) = ReplicationUtils.updateLeaderAndIsr(zkUtils, topic, partition, leaderAndIsr, controller.epoch, currentLeaderAndIsr.zkVersion) newLeaderAndIsr = leaderAndIsr.withZkVersion(newVersion) zookeeperPathUpdateSucceeded = updateSucceeded replicasForThisPartition = replicas } val newLeaderIsrAndControllerEpoch = LeaderIsrAndControllerEpoch(newLeaderAndIsr, controller.epoch) // 04：更新ControllerContext的leader信息（内存中的缓存） controllerContext.partitionLeadershipInfo.put(TopicAndPartition(topic, partition), newLeaderIsrAndControllerEpoch) stateChangeLog.trace(s"Elected leader ${newLeaderAndIsr.leader} for Offline partition $topicAndPartition") val replicas = controllerContext.partitionReplicaAssignment(TopicAndPartition(topic, partition)) // 05：向broker添加LeaderAndIsr请求 brokerRequestBatch.addLeaderAndIsrRequestForBrokers(replicasForThisPartition, topic, partition, newLeaderIsrAndControllerEpoch, replicas) } catch { //省略一些异常处理 } debug(s"After leader election, leader cache for $topicAndPartition is updated to ${controllerContext.partitionLeadershipInfo(topicAndPartition)}") }
实际选举leaderSelector.selectLeader(topicAndPartition, currentLeaderAndIsr)策略来自PartitionLeaderSelector，具体有以下四种
（1）OfflinePartitionLeaderSelector
触发场景：
- Controller 重新加载
- 脚本执行脏选举
- Controller 监听到有Broker启动了
- Controller 监听 LeaderAndIsrResponseReceived请求：
- Controller 监听 UncleanLeaderElectionEnable请求：
unclean.leader.election.enable设置为 true时
选举规则
找AR中第一个在线且在isr中的副本，若找不到，且unclean.leader.election.enable为true，找AR中第一个在线的副本。
（2）ReassignedPartitionLeaderSelector
触发场景：
- 分区副本重分配：只有当之前的Leader副本在经过重分配之后不存在了，或者故障下线了才会触发
选举规则
找AR中第一个在线且在isr中的副本
（3）PreferredReplicaPartitionLeaderSelector
触发场景：
- 自动定时执行优先副本选举任务：
auto.leader.rebalance.enable=true （源码KafkaController#checkAndTriggerAutoLeaderRebalance) Controller 重新加载的时候：先执行OfflinePartitionLeaderSelector再执行PreferredReplicaPartitionLeaderSelector （源码KafkaController#onControllerFailover)- 手动执行优先副本选举脚本：
kafka-leader-election.sh并且选择的模式是PREFERRED (先写zk节点，后controller触发PreferredReplicaElectionListener)
选举规则
只有满足以下条件才会选举成功：是第一个副本 && 副本在线 && 副本在ISR列表中。
代码如下
class PreferredReplicaPartitionLeaderSelector(controllerContext: ControllerContext) extends PartitionLeaderSelector with Logging { logIdent = "[PreferredReplicaPartitionLeaderSelector]: " def selectLeader(topicAndPartition: TopicAndPartition, currentLeaderAndIsr: LeaderAndIsr): (LeaderAndIsr, Seq[Int]) = {
// 从内存中（AR）拿到第一个副本，就是首选副本 val assignedReplicas = controllerContext.partitionReplicaAssignment(topicAndPartition) val preferredReplica = assignedReplicas.head // 判断实际的leader是不是首选副本，若是就不需要选举了 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) + " Triggering preferred replica leader election") // 否则，首选副本在线并且在isr中，就选举其为新leader if (controllerContext.isReplicaOnline(preferredReplica, topicAndPartition) && currentLeaderAndIsr.isr.contains(preferredReplica)) { val newLeaderAndIsr = currentLeaderAndIsr.newLeader(preferredReplica) (newLeaderAndIsr, assignedReplicas) } else { throw new StateChangeFailedException(s"Preferred replica $preferredReplica for partition $topicAndPartition " + s"is either not alive or not in the isr. Current leader and ISR: [$currentLeaderAndIsr]") } } } }
（4）ControlledShutdownLeaderSelector
触发场景：
- Broker关机的时候：当Broker关机的过程中,会向Controller发起一个请求, 让它重新发起一次选举, 把在所有正在关机(也就是发起请求的那个Broker,或其它同时正在关机的Broker) 的Broker里面的副本给剔除掉。
选举规则
在AR中找到第一个满足条件的副本：副本在线 && 副本在ISR中 && 副本所在的Broker不在正在关闭的Broker集合中 。
