Zookeeper断网重连事件回调源码分析

最新推荐文章于 2024-02-04 00:19:03 发布

原创

最新推荐文章于 2024-02-04 00:19:03 发布 · 1.6k 阅读

0 ·

CC 4.0 BY-SA版权

文章标签：

#zookeeper #源码分析

文章详细分析了使用Curator作为zk客户端时，断网后重连如何触发ConnectionStateListener的RECONNECTED事件，涉及源码追踪和事件处理机制。

“不积跬步，无以至千里。”

背景

确定使用Curator作为zk客户端的情况下，断网[发生SUSPENDED | LOST事件]重连后每次都会回调org.apache.curator.framework.state.ConnectionStateListener#stateChanged方法，且事件类型为org.apache.curator.framework.state.ConnectionState#RECONNECTED
部署zookeeper的版本为最新稳定版3.8.3，curator-recipes相关依赖的版本为5.5.0

源码分析过程

首先需要构建一个CuratorFramework对象，并基于这个CuratorFramework对象创建一个用于实现Leader选举功能的LeaderSelector，并将它启动

public static final String leaderSelectorPath = "/source-code-analyse/reconnect";
public static void main(String[] args) throws InterruptedException {
     
     
     CuratorFramework curatorFramework = newCuratorFramework();
     LeaderSelectorListener leaderSelectorListener = new LeaderSelectorListener() {
     
     
         @Override
         public void stateChanged(CuratorFramework curatorFramework, ConnectionState connectionState) {
     
     
             System.out.println("Thread " + Thread.currentThread().getName() + " Connection state changed : " + connectionState);
         }

         @Override
         public void takeLeadership(CuratorFramework curatorFramework) throws Exception {
     
     
             System.out.println("Thread " + Thread.currentThread().getName() + " get the leader.");
             TimeUnit.SECONDS.sleep(20);
         }
     };
     LeaderSelector leaderSelector = new LeaderSelector(curatorFramework, leaderSelectorPath, leaderSelectorListener);
     leaderSelector.start();
     TimeUnit.SECONDS.sleep(100);
     leaderSelector.close();
     curatorFramework.close();
     System.out.println("Test completed.");
 }

  public static CuratorFramework newCuratorFramework() {
     
     
     CuratorFramework curatorFramework = CuratorFrameworkFactory.builder()
             .connectString("192.168.0.104:2181")
             .sessionTimeoutMs(30000)
             .connectionTimeoutMs(6000)
             .retryPolicy(new ExponentialBackoffRetry(1000, 3))
             .threadFactory(ThreadUtils.newThreadFactory("ReconnectionTestThread")).build();
     curatorFramework.start();
     return curatorFramework;
 }

由于LeaderSelector的功能实现需要基于CuratorFramework，于是应该先看看CuratorFramework的start方法，直接看实现类CuratorFrameworkImpl

@Override
public void start()
{
     
     
  log.info("Starting");
  if ( !state.compareAndSet(CuratorFrameworkState.LATENT, CuratorFrameworkState.STARTED) )
  {
     
     
    throw new IllegalStateException("Cannot be started more than once");
  }
  try
  {
     
     
    connectionStateManager.start();
    //省略代码

发现CuratorFrameworkImpl内部维护了一个与连接状态管理器，start方法中会启动它

在ConnectionStateManager的start方法中，会向线程池提交一个任务，去调用processEvents方法

public void start()
{
     
     
  Preconditions.checkState(state.compareAndSet(State.LATENT, State.STARTED), "Cannot be started more than once");
  service.submit
    (
    new Callable<Object>()
    {
     
     
      @Override
      public Object call() throws Exception
      {
     
     
        processEvents();
        return null;
      }
    }
  );
}

processEvents方法里面核心的内容就是，从eventQueue的一个阻塞队列中不断调用poll方法获取ConnectionState对象，因为处在一个while循环中，只要当前连接状态正常，就会一直去poll

private void processEvents()
 {
     
     
     while ( state.get() == State.STARTED )
     {
     
     
         try
         {
     
     
             int useSessionTimeoutMs = getUseSessionTimeoutMs();
             long elapsedMs = startOfSuspendedEpoch ==