Java实现ZooKeeper的zNode监控

上一篇文章已经完成了ZooKeeper的基本搭建和使用的介绍，现在开始用代码说话。参考 https://zookeeper.apache.org/doc/current/javaExample.html ，但对场景和代码都做了简化，只实现基本的Watcher功能。

1   场景设计

目的是体验ZooKeeper的Watcher功能。程序监控ZooKeeper的/watcher节点数据变化，当创建或修改数据时，控制台打印当前的数据内容和版本号；当/watcher被删除时，程序退出。

/watcher的创建、修改和删除操作，使用控制台或zkui操作。

2   搭建Maven项目

代码相对比较简单，就不用SpringBoot这个大杀器了，使用一个普通的Maven项目即可。

ZooKeeper客户端使用官方提供的Java库，org.apache.zookeeper: zookeeper: 3.5.5。日志框架使用习惯的slf4j+log4j2，ZooKeeper缺省使用的是log4j V1，因此在引入的时候需要excludes。另外，使用了lombok来简化一些代码。

以下是pom.xml的内容

<?xml version="1.0" encoding="UTF-8"?>

<project xmlns="http://maven.apache.org/POM/4.0.0" xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance"

     xsi:schemaLocation="http://maven.apache.org/POM/4.0.0 http://maven.apache.org/xsd/maven-4.0.0.xsd">

  <modelVersion>4.0.0</modelVersion>

  <groupId>tech.codestory.research</groupId>

  <artifactId>zookeeper</artifactId>

  <version>1.0.0-SNAPSHOT</version>

  <dependencies>

    <dependency>

      <groupId>org.apache.zookeeper</groupId>

      <artifactId>zookeeper</artifactId>

      <version>3.5.5</version>

      <exclusions>

        <exclusion>

          <groupId>log4j</groupId>

          <artifactId>log4j</artifactId>

        </exclusion>

        <exclusion>

          <groupId>org.slf4j</groupId>

          <artifactId>slf4j-log4j12</artifactId>

        </exclusion>

      </exclusions>

    </dependency>

    <dependency>

      <groupId>org.apache.logging.log4j</groupId>

      <artifactId>log4j-core</artifactId>

      <version>2.12.1</version>

    </dependency>

    <dependency>

      <groupId>org.apache.logging.log4j</groupId>

      <artifactId>log4j-api</artifactId>

      <version>2.12.1</version>

    </dependency>

    <dependency>

      <groupId>org.apache.logging.log4j</groupId>

      <artifactId>log4j-web</artifactId>

      <version>2.12.1</version>

    </dependency>

    <dependency>

      <groupId>org.apache.logging.log4j</groupId>

      <artifactId>log4j-slf4j-impl</artifactId>

      <version>2.12.1</version>

    </dependency>

    <dependency>

      <groupId>org.slf4j</groupId>

      <artifactId>slf4j-api</artifactId>

      <version>1.7.28</version>

    </dependency>

    <dependency>

      <groupId>org.slf4j</groupId>

      <artifactId>slf4j-ext</artifactId>

      <version>1.7.28</version>

    </dependency>

    <dependency>

      <groupId>org.projectlombok</groupId>

      <artifactId>lombok</artifactId>

      <version>1.18.8</version>

      <scope>provided</scope>

    </dependency>

  </dependencies>

</project>

3   log4j2.xml

在项目的 src/main/resources 下创建一个文件 log4j2.xml，内容为

<?xml version="1.0" encoding="UTF-8"?>

<configuration status="DEBUG" name="codeStoryLogger">

  <appenders>

    <Console name="stdout">

      <ThresholdFilter level="TRACE" onMatch="ACCEPT" onMismatch="DENY"/>

      <PatternLayout pattern="%d{HH:mm:ss:SSS} [%p] - %c{1}.%M(%L) - %m%n"/>

    </Console>

  </appenders>

  <ThresholdFilter level="TRACE"/>

  <loggers>

    <logger name="org.apache.zookeeper.ClientCnxn" level="ERROR" additivity="false">

      <appender-ref ref="stdout" />

    </logger>

    <logger name="org.apache.zookeeper" level="TRACE" additivity="false">

      <appender-ref ref="stdout"/>

    </logger>

    <logger name="tech.codestory" level="TRACE" additivity="false">

      <appender-ref ref="stdout"/>

    </logger>

    <root level="WARN">

      <appender-ref ref="stdout"/>

    </root>

  </loggers>

</configuration>

4   创建ZooKeeper连接

创建连接代码比较简单，只需要创建 ZooKeeper对象就行，

ZooKeeper构造函数的定义

/**

 * 创建一个 ZooKeeper 客户端对象

 * @param connectString 逗号分隔的 host:port 字符串，

 *    单点如 127.0.0.1:2181，

 *    集群如 192.168.5.128:2181,192.168.5.129:2181,192.168.5.130:2181，

 *    还可以指定根节点，如 127.0.0.1:2181/foo/bar

 * @param sessionTimeout 毫秒为单位的超时时间

 * @param watcher Watcher对象，用于接收 MatcherEvent

 * @throws IOException 网络错误时抛出异常

 * @throws IllegalArgumentException 如果Root路径设置错误

 */

public ZooKeeper(String connectString, int sessionTimeout, Watcher watcher)

  throws IOException;

写一段测试代码，创建zk对象后判断某一个zNode是否存在。

public class ZooKeeperWatcher implements Watcher {

  /** ZooKeeper的客户端连接 */

  ZooKeeper zk;

  public ZooKeeperWatcher(String hostPort, String zNode) throws KeeperException, IOException {

    zk = new ZooKeeper(hostPort, 3000, this);

    try {

      Stat exists = zk.exists(zNode, true);

      if(exists == null){

        log.info(“{} 不存在”, zNode)

      }

    } catch (InterruptedException e) {

      log.error("InterruptedException", e);

    }

  }

}

运行这段代码，发现会抛异常

java.net.SocketException: Socket is not connected

  at sun.nio.ch.Net.translateToSocketException(Net.java:162) ~[?:?]

  at sun.nio.ch.Net.translateException(Net.java:196) ~[?:?]

  at sun.nio.ch.Net.translateException(Net.java:202) ~[?:?]

  at sun.nio.ch.SocketAdaptor.shutdownInput(SocketAdaptor.java:400) ~[?:?]

  at org.apache.zookeeper.ClientCnxnSocketNIO.cleanup(ClientCnxnSocketNIO.java:198) [zookeeper-3.5.5.jar:3.5.5]

  at org.apache.zookeeper.ClientCnxn$SendThread.cleanup(ClientCnxn.java:1338) [zookeeper-3.5.5.jar:3.5.5]

  at org.apache.zookeeper.ClientCnxn$SendThread.cleanAndNotifyState(ClientCnxn.java:1276) [zookeeper-3.5.5.jar:3.5.5]

  at org.apache.zookeeper.ClientCnxn$SendThread.run(ClientCnxn.java:1254) [zookeeper-3.5.5.jar:3.5.5]

Caused by: java.nio.channels.NotYetConnectedException

  at sun.nio.ch.SocketChannelImpl.shutdownInput(SocketChannelImpl.java:917) ~[?:?]

  at sun.nio.ch.SocketAdaptor.shutdownInput(SocketAdaptor.java:398) ~[?:?]

  ... 4 more

NotYetConnectedException的字面意思是连接还没有创建好，网络搜索了一下，建立连接需要一些时间，创建zk对象后马上调用exists命令，这时候连接还没有创建好，就会抛异常。ZooKeeper在连接建立成功之后，会发送一个WatchedEvent事件，我们可以利用这个事件完成建立连接的过程。修改后的代码如下，顺便添加了slf4j-ext中的Profiler，用于记录所消耗的时间。

public class ZooKeeperWatcher implements Watcher {

  /** 等待连接建立成功的信号 */

  private CountDownLatch connectedSemaphore = new CountDownLatch(1);

  /** ZooKeeper的客户端连接 */

  ZooKeeper zk;

  public ZooKeeperWatcher(String hostPort, String zNode) throws KeeperException, IOException {

    Profiler profiler = new Profiler("连接到ZooKeeper");

    profiler.start("开始链接");

    zk = new ZooKeeper(hostPort, 3000, this);

    try {

          profiler.start("等待连接成功的Event");

      connectedSemaphore.await();

      Stat exists = zk.exists(zNode, true);

      if(exists == null){

        log.info(“{} 不存在”, zNode)

      }

    } catch (InterruptedException e) {

      log.error("InterruptedException", e);

    }

    profiler.stop();

    profiler.setLogger(log);

    profiler.log();

  }

  /** 收到ZooKeeper的WatchedEvent */

  @Override

  public void process(WatchedEvent event) {

    log.info("event = {}", event);

    if (Event.EventType.None.equals(event.getType())) {

      // 连接状态发生变化

      if (Event.KeeperState.SyncConnected.equals(event.getState())) {

        // 连接建立成功

        connectedSemaphore.countDown();

      }

    }

  }

}

修改代码之后的执行记录日志如下，可以看到等待连接成功的Event耗时9秒多。网络上有文章说关闭防火墙可以秒连，但我测试过，没发现有什么变化，使用systemctl stop firewalld之后重新执行程序，仍然需要9秒多。

[INFO] - ZooKeeperWatcher.process(61) - event = WatchedEvent state:SyncConnected type:None path:null

[DEBUG] - ZooKeeperWatcher.log(201) -

+ Profiler [连接到ZooKeeper]

|-- elapsed time                   [开始链接]    78.912 milliseconds.

|-- elapsed time           [等待连接成功的Event]  9330.606 milliseconds.

|-- Total                  [连接到ZooKeeper]  9409.926 milliseconds.

[INFO] - ZooKeeperWatcher.readNodeData(95) - /watcher 不存在

5   读取WatchedEvent

前面的代码，只是处理了建立连接成功时的Event，下面再来看看读取数据的过程。关键代码如下：

if (Event.EventType.NodeDataChanged.equals(event.getType())

    || Event.EventType.NodeCreated.equals(event.getType())) {

  String path = event.getPath();

  if (path != null && path.equals(zNode)) {

    // 节点数据被修改

    readNodeData();

  }

}

/** 读节点数据 */

private void readNodeData() {

  try {

    Stat stat = new Stat();

    byte[] data = zk.getData(zNode, true, stat);

    if (data != null) {

      log.info("{}, value={}, version={}", zNode, new String(data), stat.getVersion());

    }

  } catch (KeeperException e) {

    log.info("{} 不存在", zNode);   

  } catch (InterruptedException e) {

    log.error("InterruptedException", e);

  }

}

当接收到创建节点和修改节点的WatchedEvent，都会将数据读出并打印在控制台。

6   调整后的完整程序清单

对前面的代码做了部分调整，同时添加了退出系统的机制：节点被删除。

package tech.codestory.zookeeper.watcher;

import java.io.IOException;

import java.util.concurrent.CountDownLatch;

import org.apache.zookeeper.*;

import org.apache.zookeeper.data.Stat;

import org.slf4j.profiler.Profiler;

import lombok.extern.slf4j.Slf4j;

/**

 * 用于测试 ZooKeeper的 WatchedEvent用法

 * @author code story

 * @date 2019/8/13

 */

@Slf4j

public class ZooKeeperWatcher implements Watcher, Runnable {

  /** 等待连接建立成功的信号 */

  private CountDownLatch connectedSemaphore = new CountDownLatch(1);

  /** 退出系统的信号 */

  static Integer quitSemaphore = Integer.valueOf(-1);

  String zNode;

  ZooKeeper zk;

  public ZooKeeperWatcher(String hostPort, String zNode) throws KeeperException, IOException {

    this.zNode = zNode;

    Profiler profiler = new Profiler("连接到ZooKeeper");

    profiler.start("开始链接");

    zk = new ZooKeeper(hostPort, 3000, this);

    try {

      profiler.start("等待连接成功的Event");

      connectedSemaphore.await();

    } catch (InterruptedException e) {

      log.error("InterruptedException", e);

    }

    profiler.stop();

    profiler.setLogger(log);

    profiler.log();

    // 先读当前的数据

    readNodeData();

  }

  /** 收到ZooKeeper的WatchedEvent */

  @Override

  public void process(WatchedEvent event) {

    log.info("event = {}", event);

    if (Event.EventType.None.equals(event.getType())) {

      // 连接状态发生变化

      if (Event.KeeperState.SyncConnected.equals(event.getState())) {

        // 连接建立成功

        connectedSemaphore.countDown();

      }

    } else if (Event.EventType.NodeDataChanged.equals(event.getType())

        || Event.EventType.NodeCreated.equals(event.getType())) {

      String path = event.getPath();

      if (path != null && path.equals(zNode)) {

        // 节点数据被修改

        readNodeData();

      }

    } else if (Event.EventType.NodeDeleted.equals(event.getType())) {

      String path = event.getPath();

      if (path != null && path.equals(zNode)) {

        synchronized (quitSemaphore) {

          // 节点被删除，通知退出线程

          quitSemaphore.notify();

        }

      }

    }

  }

  /** 读节点数据 */

  private void readNodeData() {

    try {

      Stat stat = new Stat();

      byte[] data = zk.getData(zNode, true, stat);

      if (data != null) {

        log.info("{}, value={}, version={}", zNode, new String(data), stat.getVersion());

      }

    } catch (KeeperException e) {

      log.info("{} 不存在", zNode);

      try {

        // 目的是添加Watcher

        zk.exists(zNode, true);

      } catch (KeeperException ex) {

      } catch (InterruptedException ex) {

      }

    } catch (InterruptedException e) {

      log.error("InterruptedException", e);

    }

  }

  @Override

  public void run() {

    try {

      synchronized (quitSemaphore) {

        quitSemaphore.wait();

        log.info("{} 被删除，退出", zNode);

      }

    } catch (InterruptedException e) {

      log.error("InterruptedException", e);

    }

  }

  /** 主程序，代码中写死了Server地址和zNode名，也可以改成从args中读取 */

  public static void main(String[] args) {

    String hostPort = "192.168.5.128:2181";

    String zNode = "/watcher";

    try {

      new ZooKeeperWatcher(hostPort, zNode).run();

    } catch (Exception e) {

      log.error("new ZooKeeperExecutor()", e);

    }

  }

}

做一个测试，应用启动后创建节点，修改多次节点，最后删除节点，日志输出如下：

10:13:31:979 [INFO] - ZooKeeperWatcher.process(50) - event = WatchedEvent state:SyncConnected type:None path:null

10:13:31:982 [DEBUG] - ZooKeeperWatcher.log(201) -

+ Profiler [连接到ZooKeeper]

|-- elapsed time                   [开始链接]   210.193 milliseconds.

|-- elapsed time           [等待连接成功的Event]  9385.467 milliseconds.

|-- Total                  [连接到ZooKeeper]  9596.196 milliseconds.

10:13:31:996 [INFO] - ZooKeeperWatcher.readNodeData(84) - /watcher 不存在

10:15:43:451 [INFO] - ZooKeeperWatcher.process(50) - event = WatchedEvent state:SyncConnected type:NodeCreated path:/watcher

10:15:43:463 [INFO] - ZooKeeperWatcher.readNodeData(81) - /watcher, value=hello zk 00, version=0

10:15:50:906 [INFO] - ZooKeeperWatcher.process(50) - event = WatchedEvent state:SyncConnected type:NodeDataChanged path:/watcher

10:15:50:910 [INFO] - ZooKeeperWatcher.readNodeData(81) - /watcher, value=hello zk 01, version=1

10:15:56:004 [INFO] - ZooKeeperWatcher.process(50) - event = WatchedEvent state:SyncConnected type:NodeDataChanged path:/watcher

10:15:56:007 [INFO] - ZooKeeperWatcher.readNodeData(81) - /watcher, value=hello zk 02, version=2

10:16:00:246 [INFO] - ZooKeeperWatcher.process(50) - event = WatchedEvent state:SyncConnected type:NodeDataChanged path:/watcher

10:16:00:249 [INFO] - ZooKeeperWatcher.readNodeData(81) - /watcher, value=hello zk 03, version=3

10:16:06:399 [INFO] - ZooKeeperWatcher.process(50) - event = WatchedEvent state:SyncConnected type:NodeDataChanged path:/watcher

10:16:06:402 [INFO] - ZooKeeperWatcher.readNodeData(81) - /watcher, value=hello zk 10, version=4

10:16:10:217 [INFO] - ZooKeeperWatcher.process(50) - event = WatchedEvent state:SyncConnected type:NodeDataChanged path:/watcher

10:16:10:220 [INFO] - ZooKeeperWatcher.readNodeData(81) - /watcher, value=hello zk 11, version=5

10:16:14:444 [INFO] - ZooKeeperWatcher.process(50) - event = WatchedEvent state:SyncConnected type:NodeDataChanged path:/watcher

10:16:14:447 [INFO] - ZooKeeperWatcher.readNodeData(81) - /watcher, value=hello zk 12, version=6

10:16:20:118 [INFO] - ZooKeeperWatcher.process(50) - event = WatchedEvent state:SyncConnected type:NodeDeleted path:/watcher

10:16:20:118 [INFO] - ZooKeeperWatcher.run(101) - /watcher 被删除，退出