分布式（五）-------集群环境数据一致性-------zookeeper

 

QuorumPeerConfig.parseProperties（）方法
if (type == LearnerType.OBSERVER){              //观察者
                    observers.put(Long.valueOf(sid), new QuorumServer(sid, hostname, port, electionPort, type));
                }        //port为通信，commit，ack端口      electionPort为选举端口

 

在SyncRequestProcess中持久化事物（打快照），累计多次事物，打一个快照

 

QuorumPeer中四种状态，不停地循环读取

  while (running) {
                switch (getPeerState()) {
                case LOOKING:      //正在进行领导者选举
                case Observing      //观察
                case Following       //跟随
                case Leading        //领导者

zk集群中的事物id（Zxid）：组成leader周期epoch编号+值（+1）             learnerHandler类

long zxid = qp.getZxid();
            long newEpoch = leader.getEpochToPropose(this.getSid(), lastAcceptedEpoch);//sid为follower的sid时myid
，learner的epoch，这个方法是每个线程共用的
   

 

PrepRequestProcessor：checkAcl，构造txn

SyncRequestProcessor：持久化txn,快照

FinalRequestProcessor：更新内存，返回response

Leader：LeaderZooKeeperServer.setupRequestProcessors()方法，请求处理链

firstProcessor=PrepRequestProcessor.next=ProposalRequestProcessor.next=commitProcessor.next=toBeAppliedProcessor.next=FinalRequestProcessor

SyncRequestProcessor.next=AckRequestProcessor

       流程：request--->commitProcessor阻塞---->发起提议-->持久化

Follower：FollowerZooKeeperServer.setupRequestProcessors()方法，请求处理链同理

firstProcessor=FollowerRequestProcessor.next=commitProcessor.next=FinalRequestProcessor

SyncRequestProcessor.next=SendAckRequestProcessor

Observer:同理