基于Zookeeper实现分布式锁实践

1、什么是Zookeeper？

Zookeeper是一个分布式的，开源的分布式应用程序协调服务，是Hadoop和hbase的重要组件。

引用官网的图例：

特征：

1. zookeeper的数据机构是一种节点树的数据结构，zNode是基本的单位，znode是一种和unix文件系统相似的节点，可以往这个节点存储或向这个节点获取数据
2. 通过客户端可以对znode进行数据操作，还可以注册watcher监控znode的改变

2、Zookeeper节点类型

• 持久节点（Persistent）
• 持久顺序节点（Persistent_Sequential）
• 临时节点（Ephemeral）
• 临时顺序节点（Ephemeral_Sequential）

3、Zookeeper环境搭建

下载zookeeper，官网链接，https://zookeeper.apache.org/releases.html#download，去官网找到对应的软件下载到本地

修改配置文件，${ZOOKEEPER_HOME}\conf，找到zoo_sample.cfg文件，先备份一份，另外一份修改为zoo.cfg

解压后点击zkServer.cmd运行服务端：

4、Zookeeper基本使用

在cmd窗口或者直接在idea编辑器里的terminal输入命令：

zkCli.cmd -server 127.0.0.1:2181

输入命令help查看帮助信息：

ZooKeeper -server host:port -client-configuration properties-file cmd args
        addWatch [-m mode] path # optional mode is one of [PERSISTENT, PERSISTENT_RECURSIVE] - default is PERSISTENT_RECURSIVE
        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 [-b batch size]
        delquota [-n|-b|-N|-B] path
        get [-s] [-w] path
        getAcl [-s] path
        getAllChildrenNumber path
        getEphemerals path
        history
        listquota path
        ls [-s] [-w] [-R] path
        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]
        set [-s] [-v version] path data
        setAcl [-s] [-v version] [-R] path acl
        setquota -n|-b|-N|-B val path
        stat [-w] path
        sync path
        version
        whoami

create [-s] [-e] [-c] [-t ttl] path [data] [acl]，-s表示顺序节点，-e表示临时节点，若不指定表示持久节点，acl是来进行权限控制的

[zk: 127.0.0.1:2181(CONNECTED) 1] create -s /zk-test 0
Created /zk-test0000000000

查看

[zk: 127.0.0.1:2181(CONNECTED) 4] ls /
[zk-test0000000000, zookeeper]

设置修改节点数据

set /zk-test 123

获取节点数据

get /zk-test

ps，zookeeper命令详情查看help帮助文档，也可以去官网看看文档

ok，然后java写个例子，进行watcher监听

package com.example.concurrent.zkSample;

import org.I0Itec.zkclient.IZkDataListener;
import org.I0Itec.zkclient.ZkClient;

/** * <pre> * Zookeeper 例子 * </pre> * * <pre> * @author mazq * 修改记录 * 修改后版本: 修改人： 修改日期: 2021/12/09 16:57 修改内容: * </pre> */
public class ZookeeperSample {

    public static void main(String[] args) {
        ZkClient client = new ZkClient("localhost:2181");
        client.setZkSerializer(new MyZkSerializer());
        client.subscribeDataChanges("/zk-test", new IZkDataListener() {
            @Override
            public void handleDataChange(String dataPath, Object data) throws Exception {
                System.out.println("监听到节点数据改变!");
            }

            @Override
            public void handleDataDeleted(String dataPath) throws Exception {
                System.out.println("监听到节点数据被删除了");
            }
        });

        try {
            Thread.sleep(1000 * 60 * 2);
        } catch (InterruptedException e) {
            e.printStackTrace();
        }
    }
}

5、Zookeeper应用场景

Zookeeper有什么典型的应用场景：

1. 注册中心（Dubbo）
2. 命名服务
3. Master选举
4. 集群管理
5. 分布式队列
6. 分布式锁

6、Zookeeper分布式锁

Zookeeper适合用来做分布式锁，然后具体实现是利用什么原理？我们知道zookeeper是类似于unix的文件系统，文件系统我们也知道在一个文件夹下面，会有文件名称不能一致的特性的，也就是互斥的特性。同样zookeeper也有这个特性，在同个znode节点下面，子节点命名不能重复。所以利用这个特性可以来实现分布式锁

业务场景：在高并发的情况下面进行订单场景，这是一个典型的电商场景

自定义的Zookeeper序列化类：

package com.example.concurrent.zkSample;

import org.I0Itec.zkclient.exception.ZkMarshallingError;
import org.I0Itec.zkclient.serialize.ZkSerializer;

import java.io.UnsupportedEncodingException;

public class MyZkSerializer implements ZkSerializer {

    private String charset = "UTF-8";

    @Override
    public byte[] serialize(Object o) throws ZkMarshallingError {
        return String.valueOf(o).getBytes();
    }

    @Override
    public Object deserialize(byte[] bytes) throws ZkMarshallingError {
        try {
            return new String(bytes , charset);
        } catch (UnsupportedEncodingException e) {
            throw new ZkMarshallingError();
        }
    }
}

订单编号生成器类，因为SimpleDateFormat是线程不安全的，所以还是要加上ThreadLocal

package com.example.concurrent.zkSample;

import java.text.DateFormat;
import java.text.SimpleDateFormat;
import java.util.Date;
import java.util.concurrent.atomic.AtomicInteger;

public class OrderCodeGenerator {

    private static final String DATE_FORMAT = "yyyyMMddHHmmss";
    private static AtomicInteger ai  = new AtomicInteger(0);
    private static int i = 0;

    private static ThreadLocal<SimpleDateFormat> threadLocal = new ThreadLocal<SimpleDateFormat>() {
        @Override
        protected SimpleDateFormat initialValue() {
            return new SimpleDateFormat(DATE_FORMAT);
        }
    };

    public static DateFormat getDateFormat() {
        return (DateFormat) threadLocal.get();
    }

    public static String generatorOrderCode() {
        try {
            return getDateFormat().format(new Date(System.currentTimeMillis()))
                    + i++;
        } finally {
            threadLocal.remove();
        }
    }

}

pom.xml加上zookeeper客户端的配置：

<dependency>
    <groupId>com.101tec</groupId>
    <artifactId>zkclient</artifactId>
    <version>0.10</version>
</dependency>

实现一个zookeeper分布式锁，思路是获取节点，这个是多线程竞争的，能获取到锁，也就是创建节点成功，就执行业务，其它抢不到锁的线程，阻塞等待，注册watcher监听锁是否释放了，释放了，取消注册watcher，继续抢锁

package com.example.concurrent.zkSample;

import lombok.extern.slf4j.Slf4j;
import org.I0Itec.zkclient.IZkDataListener;
import org.I0Itec.zkclient.ZkClient;
import org.I0Itec.zkclient.exception.ZkNodeExistsException;

import java.util.concurrent.CountDownLatch;
import java.util.concurrent.TimeUnit;
import java.util.concurrent.locks.Condition;
import java.util.concurrent.locks.Lock;

@Slf4j
public class ZKDistributeLock implements Lock {

    private String localPath;
    private ZkClient zkClient;

    ZKDistributeLock(String localPath) {
        super();
        this.localPath = localPath;
        zkClient = new ZkClient("localhost:2181");
        zkClient.setZkSerializer(new MyZkSerializer());

    }

    @Override
    public void lock() {
        while (!tryLock()) {
            waitForLock();
        }
    }

    private void waitForLock() {
        // 创建countdownLatch协同
        CountDownLatch countDownLatch = new CountDownLatch(1);

        // 注册watcher监听
        IZkDataListener listener = new IZkDataListener() {
            @Override
            public void handleDataChange(String path, Object o) throws Exception {
                //System.out.println("zookeeper data has change!!!");
            }

            @Override
            public void handleDataDeleted(String s) throws Exception {
                // System.out.println("zookeeper data has delete!!!");
                // 监听到锁释放了，释放线程
                countDownLatch.countDown();
            }
        };
        zkClient.subscribeDataChanges(localPath , listener);

        // 线程等待
        if (zkClient.exists(localPath)) {
            try {
                countDownLatch.await();
            } catch (InterruptedException e) {
                e.printStackTrace();
            }
        }

        // 取消注册
        zkClient.unsubscribeDataChanges(localPath , listener);

    }

    @Override
    public void unlock() {
        zkClient.delete(localPath);
    }

    @Override
    public boolean tryLock() {
        try {
            zkClient.createEphemeral(localPath);
        } catch (ZkNodeExistsException e) {
            return false;
        }
        return true;
    }

    @Override
    public boolean tryLock(long time, TimeUnit unit) throws InterruptedException {
        return false;
    }

    @Override
    public void lockInterruptibly() throws InterruptedException {
    }

    @Override
    public Condition newCondition() {
        return null;
    }
}

订单服务api

package com.example.concurrent.zkSample;

public interface OrderService {
    void createOrder();
}

订单服务实现类，加上zookeeper分布式锁

package com.example.concurrent.zkSample;

import java.util.concurrent.locks.Lock;

public class OrderServiceInvoker implements OrderService{

    @Override
    public void createOrder() {
        Lock zkLock = new ZKDistributeLock("/zk-test");
        //Lock zkLock = new ZKDistributeImproveLock("/zk-test");
        String orderCode = null;
        try {
            zkLock.lock();
            orderCode = OrderCodeGenerator.generatorOrderCode();

        } finally {
            zkLock.unlock();
        }
        System.out.println(String.format("thread name : %s , orderCode : %s" ,
                Thread.currentThread().getName(),
                orderCode));
    }

}

因为搭建分布式环境比较繁琐，所以这里使用juc里的并发协同工具类，CyclicBarrier模拟多线程并发的场景，模拟分布式环境的高并发场景

package com.example.concurrent.zkSample;

import java.util.concurrent.BrokenBarrierException;
import java.util.concurrent.CyclicBarrier;

public class ConcurrentDistributeTest {

    public static void main(String[] args) {
        // 多线程数
        int threadSize = 30;
        // 创建多线程循环屏障
        CyclicBarrier cyclicBarrier = new CyclicBarrier(threadSize , ()->{
            System.out.println("准备完成！");
        }) ;

        // 模拟分布式集群的场景
        for (int i = 0 ; i < threadSize ; i ++) {
            new Thread(()->{
                OrderService orderService = new OrderServiceInvoker();
                // 所有线程都等待
                try {
                    cyclicBarrier.await();
                } catch (InterruptedException e) {
                    e.printStackTrace();
                } catch (BrokenBarrierException e) {
                    e.printStackTrace();
                }
                // 模拟并发请求
                orderService.createOrder();
            }).start();
        }
    }
}

跑多几次，没有发现订单号重复的情况，分布式锁还是有点效果的

thread name : Thread-6 , orderCode : 202112100945110
thread name : Thread-1 , orderCode : 202112100945111
thread name : Thread-13 , orderCode : 202112100945112
thread name : Thread-11 , orderCode : 202112100945113
thread name : Thread-14 , orderCode : 202112100945114
thread name : Thread-0 , orderCode : 202112100945115
thread name : Thread-8 , orderCode : 202112100945116
thread name : Thread-17 , orderCode : 202112100945117
thread name : Thread-10 , orderCode : 202112100945118
thread name : Thread-5 , orderCode : 202112100945119
thread name : Thread-2 , orderCode : 2021121009451110
thread name : Thread-16 , orderCode : 2021121009451111
thread name : Thread-19 , orderCode : 2021121009451112
thread name : Thread-4 , orderCode : 2021121009451113
thread name : Thread-18 , orderCode : 2021121009451114
thread name : Thread-3 , orderCode : 2021121009451115
thread name : Thread-9 , orderCode : 2021121009451116
thread name : Thread-12 , orderCode : 2021121009451117
thread name : Thread-15 , orderCode : 2021121009451118
thread name : Thread-7 , orderCode : 2021121009451219

注释加锁的代码，再加大并发数，模拟一下

package com.example.concurrent.zkSample;

import java.util.concurrent.locks.Lock;

public class OrderServiceInvoker implements OrderService{

    @Override
    public void createOrder() {
        //Lock zkLock = new ZKDistributeLock("/zk-test");
        //Lock zkLock = new ZKDistributeImproveLock("/zk-test");
        String orderCode = null;
        try {
            //zkLock.lock();
            orderCode = OrderCodeGenerator.generatorOrderCode();

        } finally {
            //zkLock.unlock();
        }
        System.out.println(String.format("thread name : %s , orderCode : %s" ,
                Thread.currentThread().getName(),
                orderCode));
    }

}

跑多几次，发现出现订单号重复的情况，所以分布式锁是可以保证分布式环境的线程安全的

7、公平式Zookeeper分布式锁

上面例子是一种非公平锁的方式，一旦监听到锁释放了，所有线程都会去抢锁，所以容易出现“惊群效应”：

• 巨大的服务器性能损耗
• 网络冲击
• 可能造成宕机

所以，需要改进分布式锁，改成一种公平锁的模式

• 公平锁：多个线程按照申请锁的顺序去获取锁，线程会在队列里排队，按照顺序去获取锁。只有队列第1个线程才能获取到锁，获取到锁之后，其它线程都会阻塞等待，等到持有锁的线程释放锁，其它线程才会被唤醒。
• 非公平锁：多个线程都会去竞争获取锁，获取不到就进入队列等待，竞争得到就直接获取锁；然后持有锁的线程释放锁之后，所有等待的线程就都会去竞争锁。

流程图：

代码改进：

package com.example.concurrent.zkSample;

import org.I0Itec.zkclient.IZkDataListener;
import org.I0Itec.zkclient.ZkClient;
import org.I0Itec.zkclient.exception.ZkNodeExistsException;

import java.util.Collections;
import java.util.List;
import java.util.concurrent.CountDownLatch;
import java.util.concurrent.TimeUnit;
import java.util.concurrent.locks.Condition;
import java.util.concurrent.locks.Lock;

public class ZKDistributeImproveLock implements Lock {

    private String localPath;
    private ZkClient zkClient;
    private String currentPath;
    private String beforePath;

    ZKDistributeImproveLock(String localPath) {
        super();
        this.localPath = localPath;
        zkClient = new ZkClient("localhost:2181");
        zkClient.setZkSerializer(new MyZkSerializer());
        if (!zkClient.exists(localPath)) {
            try {
                this.zkClient.createPersistent(localPath);
            } catch (ZkNodeExistsException e) {
            }
        }
    }

    @Override
    public void lock() {
        while (!tryLock()) {
            waitForLock();
        }
    }

    private void waitForLock() {
        CountDownLatch countDownLatch = new CountDownLatch(1);

        // 注册watcher
        IZkDataListener listener = new IZkDataListener() {
            @Override
            public void handleDataChange(String dataPath, Object data) throws Exception {
            }
            @Override
            public void handleDataDeleted(String dataPath) throws Exception {
                // 监听到锁释放，唤醒线程
                countDownLatch.countDown();
            }
        };
        zkClient.subscribeDataChanges(beforePath, listener);

        // 线程等待
        if (zkClient.exists(beforePath)) {
            try {
                countDownLatch.await();
            } catch (InterruptedException e) {
                e.printStackTrace();
            }
        }

        // 取消注册
        zkClient.unsubscribeDataChanges(beforePath , listener);

    }

    @Override
    public void unlock() {
        zkClient.delete(this.currentPath);
    }

    @Override
    public boolean tryLock() {
        if (this.currentPath == null) {
            currentPath = zkClient.createEphemeralSequential(localPath +"/" , "123");
        }
        // 获取Znode节点下面的所有子节点
        List<String> children = zkClient.getChildren(localPath);
        // 列表排序
        Collections.sort(children);
        if (currentPath.equals(localPath + "/" + children.get(0))) { // 当前节点是第1个节点
            return true;
        } else {
            //得到当前的索引号
            int index = children.indexOf(currentPath.substring(localPath.length() + 1));
            //取到前一个
            beforePath = localPath + "/" + children.get(index - 1);
        }
        return false;
    }

    @Override
    public boolean tryLock(long time, TimeUnit unit) throws InterruptedException {
        return false;
    }

    @Override
    public void lockInterruptibly() throws InterruptedException {
    }

    @Override
    public Condition newCondition() {
        return null;
    }
}

thread name : Thread-13 , orderCode : 202112100936140
thread name : Thread-3 , orderCode : 202112100936141
thread name : Thread-14 , orderCode : 202112100936142
thread name : Thread-16 , orderCode : 202112100936143
thread name : Thread-1 , orderCode : 202112100936144
thread name : Thread-9 , orderCode : 202112100936145
thread name : Thread-4 , orderCode : 202112100936146
thread name : Thread-5 , orderCode : 202112100936147
thread name : Thread-7 , orderCode : 202112100936148
thread name : Thread-2 , orderCode : 202112100936149
thread name : Thread-17 , orderCode : 2021121009361410
thread name : Thread-15 , orderCode : 2021121009361411
thread name : Thread-0 , orderCode : 2021121009361412
thread name : Thread-10 , orderCode : 2021121009361413
thread name : Thread-18 , orderCode : 2021121009361414
thread name : Thread-19 , orderCode : 2021121009361415
thread name : Thread-8 , orderCode : 2021121009361416
thread name : Thread-12 , orderCode : 2021121009361417
thread name : Thread-11 , orderCode : 2021121009361418
thread name : Thread-6 , orderCode : 2021121009361419

8、zookeeper和Redis锁对比？

Redis和Zookeeper都可以用来实现分布式锁，两者可以进行对比：

• 基于Redis实现分布式锁

• 实现比较复杂

• 存在死锁的可能

• 性能比较好，基于内存 ，而且保证的是高可用，redis优先保证的是AP(分布式CAP理论)

• 基于Zookeeper实现分布式锁

• 实现相对简单

• 可靠性高，因为zookeeper保证的是CP(分布式CAP理论)

• 性能相对较好 并发1~2万左右，并发太高，还是redis性能好

本博客代码可以在GitHub找到下载链接

附录参考资料

• https://juejin.cn/post/6844903677367418893
• https://blog.csdn.net/yangsnow_rain_wind/article/details/80749402