大数据学习（23）—— ZooKeeper实战

本片介绍两方面内容，一方面是命令行操作，另一方面是Java调用API。

ZooKeeper集群环境的搭建在Hadoop集群搭建里已经讲过了，这里不再赘述，本篇内容基于zk3.5.8。

这里补充一点，除了前面提到的leader和follower之外，zk还有一种角色叫observer。可以这么理解，follower是候选人，leader是当前的领导者，observer是普通人。leader只能从follower中选出，observer不参与选举但可以提供读服务。

observer的配置，只需要在zoo.cfg中增加server.1=server01:2888:3888:observer就行，可以在不影响写性能的前提下扩展读并发。

命令行操作

首先启动zk集群，进入命令行。

[hadoop@server01 ~]$ zkServer.sh start
/bin/java
ZooKeeper JMX enabled by default
Using config: /usr/apache-zookeeper-3.5.8/bin/../conf/zoo.cfg
Starting zookeeper ... STARTED
[hadoop@server01 ~]$ zkServer.sh status
/bin/java
ZooKeeper JMX enabled by default
Using config: /usr/apache-zookeeper-3.5.8/bin/../conf/zoo.cfg
Client port found: 2181. Client address: localhost.
Mode: follower
[hadoop@server01 ~]$ zkCli.sh

官网说From the shell, type help to get a listing of commands that can be executed from the client，我试着输入help命令，命令列表确实输出了，但是注意看最后一行。

[zk: server01:2181(CONNECTED) 0] help
ZooKeeper -server host:port cmd args
    addauth scheme auth
    close
    config [-c] [-w] [-s]
    connect host:port
    create [-s] [-e] [-c] [-t ttl] path [data] [acl]
    delete [-v version] path
    deleteall path
    delquota [-n|-b] path
    get [-s] [-w] path
    getAcl [-s] path
    history
    listquota path
    ls [-s] [-w] [-R] path
    ls2 path [watch]
    printwatches on|off
    quit
    reconfig [-s] [-v version] [[-file path] | [-members serverID=host:port1:port2;port3[,...]*]] | [-add serverId=host:port1:port2;port3[,...]]* [-remove serverId[,...]*]
    redo cmdno
    removewatches path [-c|-d|-a] [-l]
    rmr path
    set [-s] [-v version] path data
    setAcl [-s] [-v version] [-R] path acl
    setquota -n|-b val path
    stat [-w] path
    sync path
Command not found: Command not found help

最后一行提示根本就没有help这个命令。官网这不是误导人嘛。

先试试ls命令，我这个zk已经有hadoop和yarn两个集群用来选主了，所有有它们的目录。

[zk: server01:2181(CONNECTED) 1] ls /
[hadoop-ha, yarn-leader-election, zookeeper]

创建一个目录，并给它赋值。再把值取出来。

[zk: server01:2181(CONNECTED) 2] create /game "this is a game directory"
Created /game
[zk: server01:2181(CONNECTED) 3] get /game
this is a game directory
[zk: server01:2181(CONNECTED) 4] ls /
[game, hadoop-ha, yarn-leader-election, zookeeper]

现在有没有感性认识了？这个目录树的非叶子可以存储数据，是一个键值对。

[zk: server01:2181(CONNECTED) 7] create /game/pal "XianJianXiLie"
Created /game/pal
[zk: server01:2181(CONNECTED) 8] get /game/pal
XianJianXiLie
[zk: server01:2181(CONNECTED) 9] stat /game/pal
cZxid = 0xd0000000a
ctime = Wed Mar 10 22:16:50 CST 2021
mZxid = 0xd0000000a
mtime = Wed Mar 10 22:16:50 CST 2021
pZxid = 0xd0000000a
cversion = 0
dataVersion = 0
aclVersion = 0
ephemeralOwner = 0x0
dataLength = 13
numChildren = 0

把/game/pal的值修改成李逍遥，注意观察，mZxid和dataVersion 加1了，数据长度变了，其他属性没变化。

[zk: server01:2181(CONNECTED) 10] set /game/pal "LiXiaoYao"
[zk: server01:2181(CONNECTED) 11] stat /game/pal
cZxid = 0xd0000000a
ctime = Wed Mar 10 22:16:50 CST 2021
mZxid = 0xd0000000b
mtime = Wed Mar 10 22:18:08 CST 2021
pZxid = 0xd0000000a
cversion = 0
dataVersion = 1
aclVersion = 0
ephemeralOwner = 0x0
dataLength = 9
numChildren = 0
[zk: server01:2181(CONNECTED) 12] stat /game
cZxid = 0xd00000009
ctime = Wed Mar 10 22:09:04 CST 2021
mZxid = 0xd00000009
mtime = Wed Mar 10 22:09:04 CST 2021
pZxid = 0xd0000000a
cversion = 1
dataVersion = 0
aclVersion = 0
ephemeralOwner = 0x0
dataLength = 24
numChildren = 1

现在把znode删掉，它提示不能删除非空目录。我们还有个deleteall命令，这个可以级联删除。

[zk: server01:2181(CONNECTED) 13] delete /game
Node not empty: /game
[zk: server01:2181(CONNECTED) 14] deleteall /game
[zk: server01:2181(CONNECTED) 15] ls /
[hadoop-ha, yarn-leader-election, zookeeper]

上面创建的是永久节点。还可以用-e参数创建临时节点，当客户端断开连接后，节点自动消失。用-s创建序列节点，zk自动在节点名后面补十位递增数字。

[zk: server01:2181(CONNECTED) 4] create -e /game/3DO
Created /game/3DO
[zk: server01:2181(CONNECTED) 5] ls /game
[3DO]
[zk: server01:2181(CONNECTED) 6] create -s /game/3DO
Created /game/3DO0000000001
[zk: server01:2181(CONNECTED) 7] ls /game
[3DO, 3DO0000000001]

上面的命令在server01的客户端上创建了一个临时节点/game/3DO和一个序列节点/game/3DO0000000001，从server02上再连一个客户端看看。

[zk: server02:2181(CONNECTED) 0] ls /game
[3DO, 3DO0000000001]

server02上这两个目录也存在。现在把server01客户端断开连接。

[zk: server01:2181(CONNECTED) 8] quit

WATCHER::

WatchedEvent state:Closed type:None path:null
2021-03-11 10:39:09,535 [myid:] - INFO  [main:ZooKeeper@1422] - Session: 0x10000138b4a0001 closed
2021-03-11 10:39:09,537 [myid:] - INFO  [main-EventThread:ClientCnxn$EventThread@524] - EventThread shut down for session: 0x10000138b4a0001

从server02上再看看这个目录下有什么。

[zk: server02:2181(CONNECTED) 1] ls /game
[3DO0000000001]

可以看到临时节点/game/3DO消失了。

JAVA API

在pom.xml文件里增加配置

      <dependency>
          <groupId>org.apache.zookeeper</groupId>
          <artifactId>zookeeper</artifactId>
          <version>3.5.8</version>
      </dependency>

创建节点

先试试创建持久化节点。所有代码如下。

import org.apache.zookeeper.*;
import org.junit.After;
import org.junit.Before;
import org.junit.Test;

import java.io.IOException;
import java.util.concurrent.CountDownLatch;

public class ZooKeeperTest {

    ZooKeeper client = null;

    @Before
    public void init() throws InterruptedException, IOException {
        String url = "server01:2181,server02:2181,server03:2181";
        CountDownLatch latch = new CountDownLatch(1);
        client = new ZooKeeper(url, 5000, (event) -> {if(event.getState() == Watcher.Event.KeeperState.SyncConnected) latch.countDown();});
        latch.await();
    }

    @Test
    public void createZnode() throws KeeperException, InterruptedException {
        //持久化节点
       String znode1 =  client.create("/game/ThreeKingdom","The story of many heros".getBytes(), ZooDefs.Ids.OPEN_ACL_UNSAFE, CreateMode.PERSISTENT);
    }

    @After
    public void destroy() throws InterruptedException {
        if(client != null)
            client.close();
    }
}

在初始化方法里，创建了zk的客户端，但是这个客户端并不是new ZooKeeper( )返回之后就立马可用，必须等到连接创建完毕后才能用。所以用到了Java的CountDownLatch，在watch事件状态为连接成功之后主线程才被唤醒。

执行createZnode( )方法之后，控制台输出如下

Process finished with exit code 0

通过命令行看看目录创建成功没有

[zk: server02:2181(CONNECTED) 4] ls /game
[3DO0000000001, ThreeKingdom]
[zk: server02:2181(CONNECTED) 5] get -s /game/ThreeKingdom
The story of many heros
cZxid = 0x250000000d
ctime = Thu Mar 11 11:51:04 CST 2021
mZxid = 0x250000000d
mtime = Thu Mar 11 11:51:04 CST 2021
pZxid = 0x250000000d
cversion = 0
dataVersion = 0
aclVersion = 0
ephemeralOwner = 0x0
dataLength = 23
numChildren = 0

删除节点

    @Test
    public void deleteZnode() throws KeeperException, InterruptedException {
        //删除节点，这里的version参数必须与zk服务器最新的版本号一直，否则会抛异常BadVersionException。
        //版本号是为了控制分布式更新的并发问题。如果不关心版本号，可以设置为-1
        client.delete("/game/ThreeKingdom", -1);

    }

删除之后再通过命令行看看，/game下已经没有三国的目录了。

[zk: server02:2181(CONNECTED) 8] ls /game
[3DO0000000001]

修改节点信息

    @Test
    public void setZnode() throws KeeperException, InterruptedException {
        //
        final Stat stat = client.setData("/game", "childhood memory".getBytes(), -1);
        System.out.println(stat.toString());
    }

控制台输出结果：

158913789957,158913789974,1615430142891,1615444561053,1,5,0,0,16,1,158913789971

Process finished with exit code 0

来看看setData返回的Stat包含哪些字段信息

public String toString() {
        try {
            ByteArrayOutputStream s = new ByteArrayOutputStream();
            CsvOutputArchive a_ = new CsvOutputArchive(s);
            a_.startRecord(this, "");
            a_.writeLong(this.czxid, "czxid");
            a_.writeLong(this.mzxid, "mzxid");
            a_.writeLong(this.ctime, "ctime");
            a_.writeLong(this.mtime, "mtime");
            a_.writeInt(this.version, "version");
            a_.writeInt(this.cversion, "cversion");
            a_.writeInt(this.aversion, "aversion");
            a_.writeLong(this.ephemeralOwner, "ephemeralOwner");
            a_.writeInt(this.dataLength, "dataLength");
            a_.writeInt(this.numChildren, "numChildren");
            a_.writeLong(this.pzxid, "pzxid");
            a_.endRecord(this, "");
            return new String(s.toByteArray(), "UTF-8");
        } catch (Throwable var3) {
            var3.printStackTrace();
            return "ERROR";
        }
    }

通过命令行获取/game的状态信息，与IDE控制台返回的一致

[zk: server02:2181(CONNECTED) 10] get -s /game
childhood memory
cZxid = 0x2500000005
ctime = Thu Mar 11 10:35:42 CST 2021
mZxid = 0x2500000016
mtime = Thu Mar 11 14:36:01 CST 2021
pZxid = 0x2500000013
cversion = 5
dataVersion = 1
aclVersion = 0
ephemeralOwner = 0x0
dataLength = 16
numChildren = 1

获取节点信息

    @Test
    public void getZnode() throws KeeperException, InterruptedException {
        //
        final byte[] data = client.getData("/game", false, new Stat());
        System.out.println(new String(data));
    }

控制台输出结果

childhood memory

Process finished with exit code 0

注册监听

演示一下对一个已经存在的节点注册监听。先定义一个监听类，监听节点改变事件。

public class ZnodeChangeWatcher implements Watcher {
    @Override
    public void process(WatchedEvent watchedEvent) {
        if(Event.EventType.NodeDataChanged == watchedEvent.getType()){
            System.out.println("znode:"+watchedEvent.getPath()+"发生变化");
        }
    }
}

再添加一个测试方法，监听/game/ThreeKingdom节点。

    @Test
    public void watchZnode() throws KeeperException, InterruptedException {
        //
        Watcher watcher = new ZnodeChangeWatcher();
        client.exists("/game/ThreeKingdom",watcher);
        Thread.sleep(Integer.MAX_VALUE);
    }

通过命令行修改这个节点的值，先看看原值是The story of many heros，修改为The Story of three heros。

[zk: localhost:2181(CONNECTED) 1] get -s /game/ThreeKingdom
The story of many heros
cZxid = 0xf00000005
ctime = Sat Mar 13 22:43:41 CST 2021
mZxid = 0xf00000005
mtime = Sat Mar 13 22:43:41 CST 2021
pZxid = 0xf00000005
cversion = 0
dataVersion = 0
aclVersion = 0
ephemeralOwner = 0x0
dataLength = 23
numChildren = 0
[zk: localhost:2181(CONNECTED) 2] set /game/ThreeKingdom "The Story of three heros"

切换到IDE控制台，看看输出结果。

znode:/game/ThreeKingdom发生变化