西柚的大数据从踩坑到放弃-zookeeper：五、ZK案例分析-客户端动态获取集群机器上下线信息

ZK案例分析

任意一个接入zk集群的客户端，如何知道zk集群内znode节点的上下线？

1.首先模拟集群机器变动，先来看看代码

package zkmanage;

import org.apache.zookeeper.*;

import java.io.IOException;

/**
 * 模拟zk集群上下线1台新机器，
 * 传参hostname代表上线机器信息，
 * 停线程代表下线机器
 *
 */
public class DistributeServer {
    private String connectStr = "ubuntu001:2181,ubuntu002:2181,ubuntu003:2181";
    private int sessionTimeout = 2000;
    private ZooKeeper zkClient;

    public static void main(String[] args) throws Exception {

        //1.连接集群
        DistributeServer server = new DistributeServer();
        server.getConnect();

        //2.注册znode
        server.regist(args[0]);
        
        //3.业务代码
        server.business();

    }

    private void business() throws InterruptedException {
        Thread.sleep(Long.MAX_VALUE);
    }

    private void regist(String hostname) throws KeeperException, InterruptedException {
        zkClient.create("/leafznodes/leafznode", hostname.getBytes(), ZooDefs.Ids.OPEN_ACL_UNSAFE, CreateMode.EPHEMERAL_SEQUENTIAL);
        System.out.println(hostname + " is online");
    }

    private void getConnect() throws IOException {

        zkClient = new ZooKeeper(connectStr, sessionTimeout, new Watcher() {
            public void process(WatchedEvent watchedEvent) {

            }
        });
    }
}

这里可以看到是创建了非持久型节点，在创建节点这一过程的client和server断开后znode销毁，所以说停这个程序就相当于下线节点了

2.客户端（这里指的是接入zk想了解集群信息的机器，可以理解为运维人员）想获得所有节点信息（这个信息在上面用的hosts指代），模拟这一情景

package zkmanage;

import org.apache.zookeeper.KeeperException;
import org.apache.zookeeper.WatchedEvent;
import org.apache.zookeeper.Watcher;
import org.apache.zookeeper.ZooKeeper;
import org.omg.PortableServer.THREAD_POLICY_ID;

import java.io.IOException;
import java.util.ArrayList;
import java.util.List;

public class DistributeClient {
    private String connectStr = "ubuntu001:2181,ubuntu002:2181,ubuntu003:2181";
    private int sessionTimeout = 2000;
    private ZooKeeper zkClient;

    public static void main(String[] args) throws Exception {
        //模拟客户端感知znode节点变化
        //1.连接集群
        DistributeClient client = new DistributeClient();

        client.getConnect();
        //2.注册监听
        client.getChildren();

        //3.业务代码
        client.business();
    }

    private void business() throws InterruptedException {
        Thread.sleep(Long.MAX_VALUE);
    }

    private void getChildren() throws KeeperException, InterruptedException {
        ArrayList<String> hosts = new ArrayList<String>();
        List<String> children = zkClient.getChildren("/leafznodes", true);
        for (String child: children
             ) {
            byte[] data = zkClient.getData("/leafznodes/" + child, false, null);
            hosts.add(new String(data));
        }

        System.out.println(hosts);
    }

    private void getConnect() throws IOException {
        zkClient = new ZooKeeper(connectStr, sessionTimeout, new Watcher() {
            public void process(WatchedEvent watchedEvent) {
                try {
                    getChildren();
                } catch (KeeperException e) {
                    e.printStackTrace();
                } catch (InterruptedException e) {
                    e.printStackTrace();
                }
            }
        });
    }
}

3.运行起来后，无论在shell里或者是DistributeServer类里，在/leafznodes路径下创建或删除节点时，在DistributeClient控制台上均能打印当前全部节点信息。这是因为getChildren里面注册了监听，而监听执行的process方法又调用了getchildren，又导致注册了一个针对/leafznodes路径的监听（禁止套娃）。确实死循环了，不过这程序也就调用一下api，也不一直在转，没什么好死机的。这么做的目的是能让程序一直运行一直监听，不然监听一次失效，前面说过这个事的。

运行效果类似下面

[]
[]
[node]
[node, node1]
[node2222, node, node1]
[node2222, node444, node, node1]
[node2222, node444, node5555, node, node1]
[node2222, node5555, node, node1]
[node2222, node, node1]