springboot使用zookeeper(curator)实现注册发现与负载均衡

最简单的实现服务高可用的方法就是集群化，也就是分布式部署，但是分布式部署会带来一些问题。比如：

1、各个实例之间的协同(锁)

2、负载均衡

3、热删除

这里通过一个简单的实例来说明如何解决注册发现和负载均衡。

1、先解决依赖，这里只给出zk相关的依赖，pom.xml如下

　　<dependency>
            <groupId>org.apache.zookeeper</groupId>
            <artifactId>zookeeper</artifactId>
            <version>3.4.8</version>
        </dependency>
        <dependency>
            <groupId>org.apache.curator</groupId>
            <artifactId>curator-recipes</artifactId>
            <version>2.9.1</version>
        </dependency>

        <dependency>
            <groupId>org.apache.curator</groupId>
            <artifactId>curator-client</artifactId>
            <version>2.9.1</version>
　　</dependency>

2、ZkClient

这里使用的是curator，curator是对zookeeper的简单封装，提供了一些集成的方法，或者是提供了更优雅的api，举例来说

zk的create(path, mode, acl, data)方法 == curator create().withMode(mode).forPath(path)调用链

package com.dqa.prometheus.client.zookeeper;

import org.apache.curator.RetryPolicy;
import org.apache.curator.framework.CuratorFramework;
import org.apache.curator.framework.CuratorFrameworkFactory;
import org.apache.curator.retry.ExponentialBackoffRetry;
import org.apache.zookeeper.CreateMode;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;

import java.net.InetAddress;
import java.util.ArrayList;
import java.util.List;

public class ZkClient {
    private final Logger logger = LoggerFactory.getLogger(this.getClass());

    private CuratorFramework client;
    private String zookeeperServer;
    private int sessionTimeoutMs;
    private int connectionTimeoutMs;
    private int baseSleepTimeMs;
    private int maxRetries;

    public void setZookeeperServer(String zookeeperServer) {
        this.zookeeperServer = zookeeperServer;
    }
    public String getZookeeperServer() {
        return zookeeperServer;
    }
    public void setSessionTimeoutMs(int sessionTimeoutMs) {
        this.sessionTimeoutMs = sessionTimeoutMs;
    }
    public int getSessionTimeoutMs() {
        return sessionTimeoutMs;
    }
    public void setConnectionTimeoutMs(int connectionTimeoutMs) {
        this.connectionTimeoutMs = connectionTimeoutMs;
    }
    public int getConnectionTimeoutMs() {
        return connectionTimeoutMs;
    }
    public void setBaseSleepTimeMs(int baseSleepTimeMs) {
        this.baseSleepTimeMs = baseSleepTimeMs;
    }
    public int getBaseSleepTimeMs() {
        return baseSleepTimeMs;
    }
    public void setMaxRetries(int maxRetries) {
        this.maxRetries = maxRetries;
    }
    public int getMaxRetries() {
        return maxRetries;
    }

    public void init() {
        RetryPolicy retryPolicy = new ExponentialBackoffRetry(baseSleepTimeMs, maxRetries);
        client = CuratorFrameworkFactory.builder().connectString(zookeeperServer).retryPolicy(retryPolicy)
                .sessionTimeoutMs(sessionTimeoutMs).connectionTimeoutMs(connectionTimeoutMs).build();
        client.start();
    }

    public void stop() {
        client.close();
    }

    public CuratorFramework getClient() {
        return client;
    }

    public void register() {
        try {
            String rootPath = "/" + "services";
            String hostAddress = InetAddress.getLocalHost().getHostAddress();
            String serviceInstance = "prometheus" + "-" +  hostAddress + "-";
            client.create().creatingParentsIfNeeded().withMode(CreateMode.EPHEMERAL_SEQUENTIAL).forPath(rootPath + "/" + serviceInstance);
        } catch (Exception e) {
            logger.error("注册出错", e);
        }
    }

    public List<String> getChildren(String path) {
        List<String> childrenList = new ArrayList<>();
        try {
            childrenList = client.getChildren().forPath(path);
        } catch (Exception e) {
            logger.error("获取子节点出错", e);
        }
        return childrenList;
    }

    public int getChildrenCount(String path) {
        return getChildren(path).size();
    }

    public List<String> getInstances() {
        return getChildren("/services");
    }

    public int getInstancesCount() {
        return getInstances().size();
    }
}

2、configuration如下

package com.dqa.prometheus.configuration;

import com.dqa.prometheus.client.zookeeper.ZkClient;
import org.springframework.beans.factory.annotation.Value;
import org.springframework.context.annotation.Bean;
import org.springframework.context.annotation.Configuration;

@Configuration
public class ZkConfiguration {
    @Value("${zookeeper.server}")
    private String zookeeperServer;
    @Value(("${zookeeper.sessionTimeoutMs}"))
    private int sessionTimeoutMs;
    @Value("${zookeeper.connectionTimeoutMs}")
    private int connectionTimeoutMs;
    @Value("${zookeeper.maxRetries}")
    private int maxRetries;
    @Value("${zookeeper.baseSleepTimeMs}")
    private int baseSleepTimeMs;

    @Bean(initMethod = "init", destroyMethod = "stop")
    public ZkClient zkClient() {
        ZkClient zkClient = new ZkClient();
        zkClient.setZookeeperServer(zookeeperServer);
        zkClient.setSessionTimeoutMs(sessionTimeoutMs);
        zkClient.setConnectionTimeoutMs(connectionTimeoutMs);
        zkClient.setMaxRetries(maxRetries);
        zkClient.setBaseSleepTimeMs(baseSleepTimeMs);
        return zkClient;
    }

}

配置文件如下

#============== zookeeper ===================
zookeeper.server=10.93.21.21:,10.93.18.34:,10.93.18.35:
zookeeper.sessionTimeoutMs=
zookeeper.connectionTimeoutMs=
zookeeper.maxRetries=
zookeeper.baseSleepTimeMs=

3、注册发现

是通过上面封装的ZkClient中的register方法实现的，调用如下。

package com.dqa.prometheus;

import com.dqa.prometheus.client.zookeeper.ZkClient;
import org.springframework.boot.SpringApplication;
import org.springframework.boot.autoconfigure.SpringBootApplication;
import org.springframework.boot.orm.jpa.EntityScan;
import org.springframework.context.ApplicationContext;
import org.springframework.scheduling.annotation.EnableAsync;
import org.springframework.scheduling.annotation.EnableScheduling;

@SpringBootApplication
@EnableAsync
@EnableScheduling
@EntityScan(basePackages="com.xiaoju.dqa.prometheus.model")
public class Application {
    public static void main(String[] args) {
        ApplicationContext context = SpringApplication.run(Application.class, args);
        ZkClient zkClient = context.getBean(ZkClient.class);
        zkClient.register();
    }
}

注册代码说明：

 public void register() {
        try {
            String rootPath = "/" + "services";
            String hostAddress = InetAddress.getLocalHost().getHostAddress();
            String serviceInstance = "prometheus" + "-" +  hostAddress + "-";
            client.create().creatingParentsIfNeeded().withMode(CreateMode.EPHEMERAL_SEQUENTIAL).forPath(rootPath + "/" + serviceInstance);
        } catch (Exception e) {
            logger.error("注册出错", e);
        }
    }

1、zk中的注册路径

/services/prometheus-10.93.21.21-00000000001

2、CreateMode有四种，选择EPHEMERAL_SEQUENTIAL的原因是，服务关闭的时候session超时，zk节点会自动删除，同时自增id可以实现锁和负载均衡，下面再说

、PERSISTENT

持久化目录节点，存储的数据不会丢失。

、PERSISTENT_SEQUENTIAL

顺序自动编号的持久化目录节点，存储的数据不会丢失，并且根据当前已近存在的节点数自动加 ，然后返回给客户端已经成功创建的目录节点名。

、EPHEMERAL

临时目录节点，一旦创建这个节点的客户端与服务器端口也就是session 超时，这种节点会被自动删除。 

、EPHEMERAL_SEQUENTIAL

临时自动编号节点，一旦创建这个节点的客户端与服务器端口也就是session 超时，这种节点会被自动删除，并且根据当前已近存在的节点数自动加 ，然后返回给客户端已经成功创建的目录节点名。

4、负载均衡

　　　　　/*
        *   我是第几个实例, 做负载均衡
        * */
        List<String> instanceList = zkClient.getInstances();
        Collections.sort(instanceList);
        String hostAddress = NetFunction.getAddressHost();
        int instanceNo = 0;
        if (hostAddress !=  null) {
            for (int i=0; i<instanceList.size(); i++) {
                if (instanceList.get(i).split("-")[1].equals(hostAddress)) {
                    instanceNo = i;
                }
            }
        } else {
            logger.info("获取本地IP失败");
        }
        logger.info("[分发] 实例总数={}, 我是第{}个实例", instanceCount, instanceNo);
        List<CheckTask> waitingTasks = checkTaskDao.getTasks(taskType, TaskStatus.WAITING.getValue());
        Iterator<CheckTask> waitingIterator = waitingTasks.iterator();
        while (waitingIterator.hasNext()) {
            if (waitingIterator.next().getTaskId().hashCode() % instanceCount != instanceNo) {
                waitingIterator.remove();
            }
        }

说明：

1、例如有3个实例(zkClient.getInstances())，那么通过IP我们把3个实例按照自增id排序分别标号为0，1，2

2、对第一个实例也就是instanceNo=0，只执行taskId.hashCode() % 3 == 0的任务，其他两个实例类似

3、当有一个实例挂掉，2个实例，instanceNo=0只执行taskId.hashCode() % 2 == 0的任务，实现热删除