Zookeeper实战

这里的实战主要是用Java API对Zookeeper进行相应的操作，来实现我们的业务需求。首先简单地测试一下Zookeeper节点的增删改查操作，然后测试两个例子，分别是分布式应用HA以及分布式锁。

1. 增删改查的测试demo

首先，我们建立一个工程，导入相应的jar包依赖，如下：

然后，我们开始进行简单地测试：

package cn.darrenchan.bigdata.zk;

import java.util.List;

import org.apache.zookeeper.CreateMode;
import org.apache.zookeeper.KeeperException;
import org.apache.zookeeper.WatchedEvent;
import org.apache.zookeeper.Watcher;
import org.apache.zookeeper.ZooDefs.Ids;
import org.apache.zookeeper.ZooKeeper;
import org.apache.zookeeper.data.Stat;
import org.junit.Before;
import org.junit.Test;

public class SimpleZkClient {

    private static final String connectString = "weekend01:2181,weekend02:2181,weekend03:2181";
    private static final int sessionTimeout = 2000;//会话超时时间2s    

    private ZooKeeper zkClient = null;

    @Before
    public void init() throws Exception {
        zkClient = new ZooKeeper(connectString, sessionTimeout, new Watcher() {
            @Override
            public void process(WatchedEvent event) {
                // 收到事件通知后的回调函数（应该是我们自己的事件处理逻辑）
                System.out.println(event.getType() + "---" + event.getPath());
                try {
                    // 每次设成true，是因为每次只能监控一遍
                    zkClient.getChildren("/", true);
                } catch (Exception e) {
                }
            }
        });

    }

    /**
     * 数据的增删改查
     *
     * @throws InterruptedException
     * @throws KeeperException
     */

    // 创建数据节点到zk中
    @Test
    public void testCreate() throws KeeperException, InterruptedException {
        // 参数1：要创建的节点的路径 参数2：节点大数据 参数3：节点的权限 参数4：节点的类型
        String nodeCreated = zkClient.create("/eclipse", "hellozk".getBytes(), Ids.OPEN_ACL_UNSAFE, CreateMode.PERSISTENT);
        //上传的数据可以是任何类型，但都要转成byte[]
        System.out.println(nodeCreated);
    }

    //判断znode是否存在
    @Test
    public void testExist() throws Exception{
        Stat stat = zkClient.exists("/eclipse", false);
        System.out.println(stat==null?"not exist":"exist");
    }

    // 获取子节点
    @Test
    public void getChildren() throws Exception {
        List<String> children = zkClient.getChildren("/", true);
        for (String child : children) {
            System.out.println(child);
        }
        Thread.sleep(Long.MAX_VALUE);
    }

    //获取znode的数据
    @Test
    public void getData() throws Exception {

        byte[] data = zkClient.getData("/eclipse", false, null);
        System.out.println(new String(data));

    }

    //删除znode
    @Test
    public void deleteZnode() throws Exception {

        //参数2：指定要删除的版本，-1表示删除所有版本
        zkClient.delete("/eclipse", -1);

    }
    //删除znode
    @Test
    public void setData() throws Exception {

        zkClient.setData("/eclipse", "i miss you huiru".getBytes(), -1);

        byte[] data = zkClient.getData("/eclipse", false, null);
        System.out.println(new String(data));

    }

}

Zookeeper的监听器工作机制

监听器是一个接口，我们的代码中可以实现Wather这个接口，实现其中的process方法，方法中即我们自己的业务逻辑。

监听器的注册是在获取数据的操作中实现：

getData(path,watch?)监听的事件是：节点数据变化事件。

getChildren(path,watch?)监听的事件是：节点下的子节点增减变化事件。

2. 分布式应用HA

某分布式系统中，主节点可以有多台，可以动态上下线。

任意一台客户端都能实时感知到主节点服务器的上下线。

客户端实现：

package cn.darrenchan.bigdata.zkdist;

import java.util.ArrayList;
import java.util.List;

import org.apache.zookeeper.WatchedEvent;
import org.apache.zookeeper.Watcher;
import org.apache.zookeeper.Watcher.Event.EventType;
import org.apache.zookeeper.ZooKeeper;
import org.apache.zookeeper.data.Stat;

public class AppClient {
    private String groupNode = "sgroup";
    private ZooKeeper zk;
    private Stat stat = new Stat();
    private volatile List<String> serverList;

    /**
     * 连接zookeeper
     */
    public void connectZookeeper() throws Exception {
        zk = new ZooKeeper("weekend01:2181,weekend02:2181,weekend03:2181", 2000, new Watcher() {
            public void process(WatchedEvent event) {
                // 如果发生了"/sgroup"节点下的子节点变化事件, 更新server列表, 并重新注册监听
                if (event.getType() == EventType.NodeChildrenChanged && ("/" + groupNode).equals(event.getPath())) {
                    try {
                        updateServerList();
                    } catch (Exception e) {
                        e.printStackTrace();
                    }
                }
            }
        });

        updateServerList();
    }

    /**
     * 更新server列表
     */
    private void updateServerList() throws Exception {
        List<String> newServerList = new ArrayList<String>();

        // 获取并监听groupNode的子节点变化
        // watch参数为true, 表示监听子节点变化事件.
        // 每次都需要重新注册监听, 因为一次注册, 只能监听一次事件, 如果还想继续保持监听, 必须重新注册
        List<String> subList = zk.getChildren("/" + groupNode, true);
        for (String subNode : subList) {
            // 获取每个子节点下关联的server地址
            byte[] data = zk.getData("/" + groupNode + "/" + subNode, false, stat);
            newServerList.add(new String(data, "utf-8"));
        }

        // 替换server列表
        serverList = newServerList;

        System.out.println("server list updated: " + serverList);
    }

    /**
     * client的工作逻辑写在这个方法中 此处不做任何处理, 只让client sleep
     */
    public void handle() throws InterruptedException {
        Thread.sleep(Long.MAX_VALUE);
    }

    public static void main(String[] args) throws Exception {
        AppClient ac = new AppClient();
        ac.connectZookeeper();

        ac.handle();
    }
}

服务器实现：

package cn.darrenchan.bigdata.zkdist;

import org.apache.zookeeper.CreateMode;
import org.apache.zookeeper.WatchedEvent;
import org.apache.zookeeper.Watcher;
import org.apache.zookeeper.ZooDefs.Ids;
import org.apache.zookeeper.ZooKeeper;

public class AppServer {
    private String groupNode = "sgroup";
    private String subNode = "sub";

    /**
     * 连接zookeeper
     *
     * @param address
     *            server的地址
     */
    public void connectZookeeper(String address) throws Exception {
        ZooKeeper zk = new ZooKeeper("weekend01:2181,weekend02:2181,weekend03:2181", 2000, new Watcher() {
            public void process(WatchedEvent event) {
                // 不做处理
            }
        });
        // 在"/sgroup"下创建子节点
        // 子节点的类型设置为EPHEMERAL_SEQUENTIAL, 表明这是一个临时节点, 且在子节点的名称后面加上一串数字后缀
        // 将server的地址数据关联到新创建的子节点上
        String createdPath = zk.create("/" + groupNode + "/" + subNode, address.getBytes("utf-8"), Ids.OPEN_ACL_UNSAFE,
                CreateMode.EPHEMERAL_SEQUENTIAL);
        System.out.println("create: " + createdPath);
    }

    /**
     * server的工作逻辑写在这个方法中 此处不做任何处理, 只让server sleep
     */
    public void handle() throws InterruptedException {
        Thread.sleep(Long.MAX_VALUE);
    }

    public static void main(String[] args) throws Exception {
        // 在参数中指定server的地址
        if (args.length == 0) {
            System.err.println("The first argument must be server address");
            System.exit(1);
        }

        AppServer as = new AppServer();
        as.connectZookeeper(args[0]);
        as.handle();
    }
}

3. 分布式锁

开启10个线程，按照顺序执行业务，一个执行完进而执行下一个，直到全部执行完为止。

package cn.darrenchan.bigdata.zklock;

import java.util.Collections;
import java.util.List;
import java.util.concurrent.CountDownLatch;

import org.apache.zookeeper.CreateMode;
import org.apache.zookeeper.WatchedEvent;
import org.apache.zookeeper.Watcher;
import org.apache.zookeeper.Watcher.Event.EventType;
import org.apache.zookeeper.Watcher.Event.KeeperState;
import org.apache.zookeeper.ZooDefs.Ids;
import org.apache.zookeeper.ZooKeeper;
import org.apache.zookeeper.data.Stat;

public class DistributedClient {
    // 超时时间
    private static final int SESSION_TIMEOUT = 5000;
    // zookeeper server列表
    private String hosts = "weekend01:2181,weekend02:2181,weekend03:2181";
    private String groupNode = "locks";
    private String subNode = "sub";

    private ZooKeeper zk;
    // 当前client创建的子节点
    private String thisPath;
    // 当前client等待的子节点
    private String waitPath;

    private CountDownLatch latch = new CountDownLatch(1);

    /**
     * 连接zookeeper
     */
    public void connectZookeeper() throws Exception {
        zk = new ZooKeeper(hosts, SESSION_TIMEOUT, new Watcher() {
            public void process(WatchedEvent event) {
                try {
                    System.out.println(event.getState()+",,,"+event.getType()+",,,"+event.getPath());
                    // 连接建立时, 打开latch, 唤醒wait在该latch上的线程
                    if (event.getState() == KeeperState.SyncConnected) {
                        latch.countDown();
                    }

                    // 发生了waitPath的删除事件
                    if (event.getType() == EventType.NodeDeleted && event.getPath().equals(waitPath)) {
                        doSomething();
                    }
                } catch (Exception e) {
                    e.printStackTrace();
                }
            }
        });

        // 等待连接建立
        latch.await();

        // 创建子节点
        thisPath = zk.create("/" + groupNode + "/" + subNode, null, Ids.OPEN_ACL_UNSAFE,
                CreateMode.EPHEMERAL_SEQUENTIAL);

        // wait一小会, 让结果更清晰一些
        Thread.sleep(10);

        // 注意, 没有必要监听"/locks"的子节点的变化情况
        List<String> childrenNodes = zk.getChildren("/" + groupNode, false);

        // 列表中只有一个子节点, 那肯定就是thisPath, 说明client获得锁
        if (childrenNodes.size() == 1) {
            doSomething();
        } else {
            String thisNode = thisPath.substring(("/" + groupNode + "/").length());
            // 排序
            Collections.sort(childrenNodes);
            int index = childrenNodes.indexOf(thisNode);
            if (index == -1) {
                // never happened
            } else if (index == 0) {
                // inddx == 0, 说明thisNode在列表中最小, 当前client获得锁
                doSomething();
            } else {
                // 获得排名比thisPath前1位的节点
                this.waitPath = "/" + groupNode + "/" + childrenNodes.get(index - 1);
                // 在waitPath上注册监听器, 当waitPath被删除时, zookeeper会回调监听器的process方法
                zk.getData(waitPath, true, new Stat());
            }
        }
    }

    private void doSomething() throws Exception {
        try {
            System.out.println("gain lock: " + thisPath);
            Thread.sleep(2000);
            // do something
        } finally {
            System.out.println("finished: " + thisPath);
            // 将thisPath删除, 监听thisPath的client将获得通知
            // 相当于释放锁
            zk.delete(this.thisPath, -1);
        }
    }

    public static void main(String[] args) throws Exception {
        for (int i = 0; i < 10; i++) {
            new Thread() {
                public void run() {
                    try {
                        DistributedClient dl = new DistributedClient();
                        dl.connectZookeeper();
                    } catch (Exception e) {
                        e.printStackTrace();
                    }
                }
            }.start();
        }

        Thread.sleep(Long.MAX_VALUE);
    }
}

实现效果：

SyncConnected,,,None,,,null
SyncConnected,,,None,,,null
SyncConnected,,,None,,,null
SyncConnected,,,None,,,null
SyncConnected,,,None,,,null
SyncConnected,,,None,,,null
SyncConnected,,,None,,,null
SyncConnected,,,None,,,null
SyncConnected,,,None,,,null
SyncConnected,,,None,,,null
gain lock: /locks/sub0000000000
finished: /locks/sub0000000000
SyncConnected,,,NodeDeleted,,,/locks/sub0000000000
gain lock: /locks/sub0000000001
finished: /locks/sub0000000001
SyncConnected,,,NodeDeleted,,,/locks/sub0000000001
gain lock: /locks/sub0000000002
finished: /locks/sub0000000002
SyncConnected,,,NodeDeleted,,,/locks/sub0000000002
gain lock: /locks/sub0000000003
finished: /locks/sub0000000003
SyncConnected,,,NodeDeleted,,,/locks/sub0000000003
gain lock: /locks/sub0000000004
finished: /locks/sub0000000004
SyncConnected,,,NodeDeleted,,,/locks/sub0000000004
gain lock: /locks/sub0000000005
finished: /locks/sub0000000005
SyncConnected,,,NodeDeleted,,,/locks/sub0000000005
gain lock: /locks/sub0000000006
finished: /locks/sub0000000006
SyncConnected,,,NodeDeleted,,,/locks/sub0000000006
gain lock: /locks/sub0000000007
finished: /locks/sub0000000007
SyncConnected,,,NodeDeleted,,,/locks/sub0000000007
gain lock: /locks/sub0000000008
finished: /locks/sub0000000008
SyncConnected,,,NodeDeleted,,,/locks/sub0000000008
gain lock: /locks/sub0000000009
finished: /locks/sub0000000009