Curator是Netflix公司开源的一个Zookeeper客户端，与Zookeeper提供的原生客户端相比，Curator的抽象层次更高，简化了Zookeeper客户端的开发量。
1.引入依赖：
<dependency>
<groupId>org.apache.curator</groupId>
<artifactId>curator-framework</artifactId>
<version>4.0.0</version>
</dependency>
<dependency>
<groupId>org.apache.curator</groupId>
<artifactId>curator-recipes</artifactId>
<version>4.0.0</version>
</dependency>
2。 curator 简单操作实例
public class CuratorDemo {
public static void main(String[] args) throws Exception {
CuratorFramework curatorFramework = CuratorFrameworkFactory.builder() // 获取构造器
.connectString("192.168.254.135:2181," + "192.168.254.136:2181,192.168.254.137:2181") // 集群地址
.sessionTimeoutMs(4000) // session超时事件
.retryPolicy(new ExponentialBackoffRetry(1000, 3)) // 重试机制
.namespace("curator") // 命名构建 根节点
.build(); // 构造
curatorFramework.start();// 启动 连接
// 结果: /curator/mic/node1
// 原生api中，必须是逐层创建，也就是父节点必须存在，子节点才能创建
String path = curatorFramework.create() // 创建节点
.creatingParentsIfNeeded() // 如果需要创建父节点
.withMode(CreateMode.PERSISTENT) // 选择节点类型
.forPath("/wuzz/node1","1".getBytes()); // 路径及值
System.out.println(path);
Stat stat = new Stat();
curatorFramework.getData() // 获取
.storingStatIn(stat).forPath("/wuzz/node1");
System.out.println(stat);
curatorFramework.setData() // 修改
.withVersion(stat.getVersion()).forPath("/wuzz/node1", "xx".getBytes());
curatorFramework.delete()// 删除
.deletingChildrenIfNeeded().forPath("/mic/node1");
curatorFramework.close();// 关闭连接
}
}
监听事件实例：
public class CuratorWatcherDemo {
public static void main(String[] args) throws Exception {
CuratorFramework curatorFramework = CuratorFrameworkFactory.builder() // 获取构造器
.connectString("192.168.254.135:2181," +
"192.168.254.136:2181,192.168.254.137:2181") // 集群地址
.sessionTimeoutMs(4000) // session超时事件
.retryPolicy(new ExponentialBackoffRetry(1000, 3)) // 重试机制
.namespace("curator") // 命名构建 根节点
.build(); // 构造
curatorFramework.start();// 启动 连接
//当前节点的创建和删除事件监听 ---永久的
// addListenerWithNodeCache(curatorFramework,"/wuzz");
//子节点的增加、修改、删除的事件监听
addListenerWithPathChildCache(curatorFramework,"/wuzz");
//综合节点监听事件
// addListenerWithTreeCache(curatorFramework,"/wuzz");
System.in.read();
}
public static void addListenerWithTreeCache(CuratorFramework curatorFramework,String path) throws Exception {
TreeCache treeCache=new TreeCache(curatorFramework,path);
TreeCacheListener treeCacheListener=new TreeCacheListener() {
@Override
public void childEvent(CuratorFramework client, TreeCacheEvent event) throws Exception {
System.out.println(event.getType()+"->"+event.getData().getPath());
}
};
treeCache.getListenable().addListener(treeCacheListener);
treeCache.start();
}
/**
* PathChildCache 监听一个节点下子节点的创建、删除、更新
* NodeCache 监听一个节点的更新和创建事件
* TreeCache 综合PatchChildCache和NodeCache的特性
*/
public static void addListenerWithPathChildCache(CuratorFramework curatorFramework,String path) throws Exception {
PathChildrenCache pathChildrenCache=new PathChildrenCache(curatorFramework,path,true);
PathChildrenCacheListener pathChildrenCacheListener=new PathChildrenCacheListener() {
@Override
public void childEvent(CuratorFramework client, PathChildrenCacheEvent event) throws Exception {
System.out.println("Receive Event2:"+event.getType());
}
};
pathChildrenCache.getListenable().addListener(pathChildrenCacheListener);
pathChildrenCache.start(PathChildrenCache.StartMode.NORMAL);
}
// 监听一个节点的更新，创建/wuzz节点事件
public static void addListenerWithNodeCache(CuratorFramework curatorFramework,String path) throws Exception {
final NodeCache nodeCache=new NodeCache(curatorFramework,path,false);
NodeCacheListener nodeCacheListener=new NodeCacheListener() {
@Override
public void nodeChanged() throws Exception {
System.out.println("Receive Event1:"+nodeCache.getCurrentData().getPath());
}
};
nodeCache.getListenable().addListener(nodeCacheListener);
nodeCache.start();
}
}
简单实用，用起来会比原生的Api会方便许多。
基于Curator的Leader选举：
Curator提供了两种选举方案：Leader Latch 和 Leader Election。下面分别介绍这两种选举方案。
Leader Latch：使用 Leader Latch 方案进行Master选举，系统将随机从候选者中选出一台作为 leader，直到调用 close() 释放leadship，此时再重新随机选举 leader，否则其他的候选者无法成为 leader。
基本原理：选择一个根路径，例如"/leader_select"，多个机器同时向该根路径下创建临时顺序节点，如"/leader_latch/0002"，"/leader_latch/0000"，"/leader_latch/0001"，节点编号最小(这里为0000)的zk客户端成为leader，没抢到Leader的节点都监听前一个节点的删除事件，在前一个节点删除后进行重新抢主.可以看到节点变化类似：
使用示例如下：
public class LeaderLatchTest {
private static final String zkServerIps = "192.168.1.101:2181";
private static final String masterPath = "/testZK/leader_latch";
public static void main(String[] args) {
AtomicInteger atomicInteger = new AtomicInteger(1);
try {//启动5个线程
for (int i = 0; i < 5; i++) {
new Thread(new Runnable() {
@Override
public void run() {
// 创建客户端
CuratorFramework client = getClient();
try {
client.blockUntilConnected();
} catch (InterruptedException e) {
e.printStackTrace();
}
System.out.println(client.getState());
int number = atomicInteger.getAndIncrement();
final LeaderLatch latch = new LeaderLatch(client, masterPath, "client#" + number);
System.out.println("创建客户端：" + latch.getId());
// LeaderLatch 添加监听事件
latch.addListener(new LeaderLatchListener() {
@Override//抢主成功时触发
public void isLeader() {
System.out.println(latch.getId() + ": 我现在被选举为Leader！我开始工作了....");
try {//放弃leader。让剩下来的服务重新选主
latch.close();
} catch (IOException e) {
e.printStackTrace();
}
}
@Override//不会执行
public void notLeader() {
}
});
try {
latch.start();
System.in.read();
} catch (Exception e) {
System.out.println(e.getMessage());
} finally {
System.out.println("客户端 " + latch.getId() + " 关闭");
CloseableUtils.closeQuietly(latch);
CloseableUtils.closeQuietly(client);
}
}
}).start();
}
// countDownLatch.await(); // 等待，只有所有线程都退出
} catch (Exception e) {
e.printStackTrace();
}
}
private static synchronized CuratorFramework getClient() {
CuratorFramework client = CuratorFrameworkFactory.builder().connectString(zkServerIps)
.sessionTimeoutMs(6000).connectionTimeoutMs(3000) //.namespace("LeaderLatchTest")
.retryPolicy(new ExponentialBackoffRetry(1000, 3)).build();
client.start();
return client;
}
}
Leader Election：通过 Leader Election 选举方案进行 Master选举，需添加 LeaderSelectorListener 监听器对领导权进行控制，当节点被选为leader之后，将调用 takeLeadership 方法进行业务逻辑处理，处理完成会立即释放 leadship，重新进行Master选举，这样每个节点都有可能成为leader。autoRequeue() 方法的调用确保此实例在释放领导权后还可能获得领导权：利用Curator中InterProcessMutex分布式锁进行抢主，抢到锁的即为Leader。示例如下：
public class LeaderSelectorTest {
private static final String zkServerIps = "192.168.1.101:2181";
private static final String masterPath = "/testZK/leader_selector";
public static void main(String[] args) {
AtomicInteger atomicInteger = new AtomicInteger(1);
try {
for (int i = 0; i < 5; i++) {
new Thread(new Runnable() {
@Override
public void run() {
CuratorFramework client = getClient(); // 创建客户端
try {
client.blockUntilConnected();
} catch (InterruptedException e) {
e.printStackTrace();
}
int number = atomicInteger.getAndIncrement();
final String name = "client#" + number;
final LeaderSelector selector = new LeaderSelector(client, masterPath, new LeaderSelectorListener() {
@Override
public void takeLeadership(CuratorFramework client) throws Exception {
System.out.println(name + ": 我现在被选举为Leader！我开始工作了....");
Thread.sleep(3000);
}
@Override
public void stateChanged(CuratorFramework curatorFramework, ConnectionState connectionState) {
}
});
System.out.println("创建客户端：" + name);
try {//在抢到leader权限并释放后，自动加入抢主队列，重新抢主
selector.autoRequeue();
selector.start();
// 阻塞
System.in.read();
} catch (Exception e) {
System.out.println(e.getMessage());
} finally {
System.out.println("客户端 " + name + " 关闭");
CloseableUtils.closeQuietly(selector);
if (!client.getState().equals(CuratorFrameworkState.STOPPED)) {
CloseableUtils.closeQuietly(client);
}
}
}
}).start();
}
} catch (Exception e) {
e.printStackTrace();
}
}
private static synchronized CuratorFramework getClient() {
CuratorFramework client = CuratorFrameworkFactory.builder().connectString(zkServerIps)
.sessionTimeoutMs(10000).connectionTimeoutMs(10000) //.namespace("LeaderLatchTest")
.retryPolicy(new ExponentialBackoffRetry(1000, 3)).build();
client.start();
return client;
}
}