基于zookeeper自己实现高可用 java

前言

高可用HA(High Availability)是分布式系统架构设计中必须考虑的因素之一,成熟的架构大部分都自带高可用实现，只需配置即可，例如基于zookeeper的storm、spark，kafka，基于数据库的quartz（Terracotta不考虑）等等。很多组件都基于zookeeper，zookeeper也成为了后端人员必修功课之一。

基于zookeeper的高可用方案

原理：

各工作节点，在zookeeper同一目录下一同创建EPHEMERAL_SEQUENTIAL节点，抢到第一个编号0的节点，成为leader。

基础提要：
1、临时节点会随着zookeeper连接的关闭自动删除

实现：

watcher类

1、DefaultWatcher

package com.ai.xiajw.watch;

import org.apache.zookeeper.WatchedEvent;
import org.apache.zookeeper.Watcher;

import java.util.concurrent.CountDownLatch;

public class DefaultWatcher implements Watcher {

    private CountDownLatch countDownLatch;

    public DefaultWatcher(CountDownLatch countDownLatch){
        this.countDownLatch = countDownLatch;
    }

    @Override
    public void process(WatchedEvent watchedEvent) {
        Event.KeeperState keeperState = watchedEvent.getState();
        Event.EventType eventType = watchedEvent.getType();

        if(keeperState.equals(Event.KeeperState.SyncConnected) && eventType.equals(Event.EventType.None)){ // zookeeper连接成功
            countDownLatch.countDown();
            System.out.println("zookeeper 创建连接成功........");
        }
    }
}

作用：监控zookeeper连接状态，防止zookeeper连接还未成功建立就进行其他zookeeper操作。

2、LeaderNodeWatcher

package com.ai.xiajw.watch;

import org.apache.zookeeper.WatchedEvent;
import org.apache.zookeeper.Watcher;

import java.util.concurrent.atomic.AtomicInteger;

public class LeaderNodeWatcher implements Watcher {

    private AtomicInteger state;

    public LeaderNodeWatcher(AtomicInteger state){
        this.state = state;
    }

    @Override
    public void process(WatchedEvent watchedEvent) {
        Event.KeeperState keeperState = watchedEvent.getState();
        Event.EventType eventType = watchedEvent.getType();
        if(eventType.equals(Event.EventType.NodeDeleted)){
            System.out.println("leader node deleted!");
            state.getAndIncrement(); // 更改状态
        }
    }
}