在java并发包中提供了若干锁的实现,它们是用于单个java虚拟机进程中的;而分布式锁能够在一组进程之间提供互斥机制,保证在任何时刻只有一个进程可以持有锁。

分布式环境中多个进程的锁则可以使用Zookeeper来实现。

下面这种方法是使用顺序节点实现共享锁,流程如下:

对于lock()操作,首先让所有参与争锁的客户端都在/_locks目录下创建临时顺序节点,然后获取该路径下的所有节点,如果客户端创建的节点序列号最小则获得锁。否则开始监视它前一个节点并进入等待状态。

对于unlock()操作,将自身创建的节点删除。此时后一个节点的监控将被触发,对应的客户端退出等待状态,取得锁。

下面是一个简单的示例:

---

/**
* 分布式锁
*/
public class DisLock implements Watcher { public static final String LOCK_ROOT = "/__locks__"; private ZooKeeper zk; //锁名称,标识竞争的是哪个锁
private String lockName; //当前创建的节点路径
private String path; //前一个节点的路径
private String prePath; //是否获取锁
private boolean acquired; //构造函数,连接zk,检查父节点存在
public DisLock(String lockName) throws KeeperException, InterruptedException, IOException {
this.lockName = "/" + lockName;
this.zk = new ZooKeeper("localhost:2181", 30000, this);
Assert.notNull(zk, "zookeeper is null");
if (zk.exists(LOCK_ROOT, false) == null) {
zk.create(LOCK_ROOT, new byte[0], ZooDefs.Ids.OPEN_ACL_UNSAFE,
CreateMode.PERSISTENT);
}
if (zk.exists(LOCK_ROOT + this.lockName, false) == null) {
zk.create(LOCK_ROOT + this.lockName, new byte[0], ZooDefs.Ids.OPEN_ACL_UNSAFE,
CreateMode.PERSISTENT);
}
} public void lock() {
if (tryLock()) {
return;
} else {
waitLock(); }
} //尝试获取锁
public boolean tryLock() {
if (acquired) {
return true;
}
try {
//创建临时节点,自动编号
path = zk.create(LOCK_ROOT + lockName + "/", new byte[0], ZooDefs.Ids.OPEN_ACL_UNSAFE,
CreateMode.EPHEMERAL_SEQUENTIAL); List<String> ls = zk.getChildren(LOCK_ROOT + lockName, false);
Collections.sort(ls);
if (path.equals(LOCK_ROOT + lockName + "/" + ls.get(0))) {
acquired = true;
return true;
} for (int i = 0; i < ls.size(); i++) {
if (path.equals(LOCK_ROOT + lockName + "/" + ls.get(i))) {
prePath = LOCK_ROOT + lockName + "/" + ls.get(i - 1);
break;
}
}
return false;
} catch (KeeperException e) {
e.printStackTrace();
} catch (InterruptedException e) {
e.printStackTrace();
}
return false;
} //释放锁
public void unlock() {
if (!acquired) {
return;
}
try {
zk.delete(path, -1);
acquired = false;
//System.out.println(Thread.currentThread().getName() + " free lock");
} catch (InterruptedException e) {
e.printStackTrace();
} catch (KeeperException e) {
e.printStackTrace();
}
} //设置监控并等待锁,前一个节点被删除后退出等待,得到锁
private synchronized void waitLock() {
try {
Stat s = zk.exists(prePath, true);
if (s == null) {
//等到锁,返回
acquired = true;
return;
}
this.wait();
} catch (InterruptedException e) {
e.printStackTrace();
} catch (KeeperException e) {
e.printStackTrace();
}
waitLock();
return;
} //监视节点变化,被监控节点被删除时激活等待锁的线程
@Override
public synchronized void process(WatchedEvent watchedEvent) {
if (watchedEvent.getType() == Event.EventType.NodeDeleted) {
//System.out.println("触发:"+watchedEvent.getPath());
this.notify();
}
}
}

---

其中waitLock和process方法需要加synchronized关键字,以便使用wait和notify方法。

测试方法:

Zookeeper的另一种应用场景是处理FIFO消息队列,利用zk的自动编号,存储数据和数据一致性能力,模拟生产者-消费者模型。

创建3个消费者从zk中获取数据,此时需要使用分布式锁,加锁获取数据的部分。出于模拟生产者和消费者都在不同的进程,所有不共享zk等对象。

示例代码如下:

生产者:

public class Producer extends Thread {

    private ZooKeeper zk;

    private Random ran = new Random();

    private static AtomicInteger count = new AtomicInteger(0);

    Producer() throws IOException {
this.zk = new ZooKeeper("localhost:2181", 30000, null);
} void produce(String str) throws KeeperException, InterruptedException {
String name = zk.create(ZkQueue.root + "/element", str.getBytes(), ZooDefs.Ids.OPEN_ACL_UNSAFE,
CreateMode.PERSISTENT_SEQUENTIAL);
//System.out.println(getName() + " create: " + str);
} @Override
public void run() {
try {
while (true) {
String msg = "msg" + count.getAndIncrement();
produce(msg);
Thread.sleep(ran.nextInt(1000));
}
} catch (KeeperException e) {
e.printStackTrace();
} catch (InterruptedException e) {
e.printStackTrace();
} finally {
}
}
}

---循环向zk中写入递增的信息,中间延迟随机毫秒

消费者:

public class Consumer extends Thread {

    private ZooKeeper zk;

    private DisLock lock;

    Consumer() throws IOException, KeeperException, InterruptedException {
lock = new DisLock("queue");
this.zk = new ZooKeeper("localhost:2181", 30000, null);
} private boolean consume() throws KeeperException, InterruptedException {
lock.lock();
try {
List<String> list = zk.getChildren(root, true);
if (list.isEmpty()) {
return true;
}
Collections.sort(list);
String first = list.get(0);
byte[] b = zk.getData(root + "/" + first, false, null);
zk.delete(root + "/" + first, -1);
String str = new String(b);
System.out.println(getName() + " get:" + str);
} finally {
lock.unlock();
}
return false;
} @Override
public void run() {
try {
while (true) {
if (consume()) {
Thread.sleep(1000);
}
}
} catch (KeeperException e) {
e.printStackTrace();
} catch (InterruptedException e) {
e.printStackTrace();
} finally {
}
}
}

---循环从zk中取节点然后删除节点,对整个过程加锁

启动类:

public class ZkQueue implements Watcher{

    @Override
public void process(WatchedEvent watchedEvent) {
System.out.printf("---->%s %s\n",watchedEvent.getPath(),watchedEvent.getType());
} private static ZooKeeper zk; public static final String root = "/zkqueue"; public static void main(String[] args) throws IOException, KeeperException, InterruptedException { zk = new ZooKeeper("localhost:2181", 30000, null); ZkUtils.delete(zk, "/__locks__/queue");//此处需保证该路径下没有子节点,否则可能获取到为最小 ZkUtils.delete(zk, root);
zk.create(root, "queue".getBytes(), ZooDefs.Ids.OPEN_ACL_UNSAFE, CreateMode.PERSISTENT); //模拟2个生产者
new Producer().start();
new Producer().start(); //模拟3个消费者
new Consumer().start();
new Consumer().start();
new Consumer().start(); Scanner s = new Scanner(System.in);
s.nextLine(); ZkUtils.delete(zk, root); // 关闭连接
zk.close();
} }

---

输出结果:

Thread-2 get lock
Thread-2 get:msg0
Thread-2 free lock
Thread-3 get lock
Thread-3 get:msg2
Thread-3 free lock
Thread-2 get lock
Thread-2 get:msg3
Thread-3 get lock
Thread-2 free lock
Thread-3 get:msg4
Thread-3 free lock
Thread-2 get lock
Thread-2 get:msg5
Thread-2 free lock
Thread-4 get lock
Thread-4 get:msg6
Thread-4 free lock
Thread-3 get lock
Thread-3 get:msg7
Thread-3 free lock
Thread-2 get lock
Thread-2 get:msg8
Thread-2 free lock
Thread-4 get lock
Thread-4 get:msg1
Thread-4 free lock
Thread-3 get lock
Thread-3 get:msg9
Thread-3 free lock
Thread-2 get lock
Thread-2 get:msg10
Thread-2 free lock
Thread-4 get lock
Thread-4 get:msg11
Thread-4 free lock
Thread-3 get lock
Thread-3 get:msg12
Thread-3 free lock
Thread-2 get lock
Thread-2 get:msg13
Thread-2 free lock
Thread-4 get lock
Thread-4 get:msg14
Thread-4 free lock
Thread-3 get lock
Thread-3 get:msg15
Thread-3 free lock
Thread-2 get lock
Thread-2 get:msg16
Thread-2 free lock
Thread-4 get lock
Thread-4 get:msg17
Thread-4 free lock
Thread-3 get lock
Thread-3 get:msg18
Thread-3 free lock
Thread-2 get lock
Thread-2 get:msg19
Thread-2 free lock
Thread-4 get lock
Thread-4 get:msg20
Thread-4 free lock
Thread-3 get lock
Thread-3 get:msg21
Thread-3 free lock
Thread-2 get lock
Thread-2 get:msg22
Thread-2 free lock
Thread-4 get lock
Thread-4 get:msg23
Thread-4 free lock
Thread-3 get lock
Thread-3 get:msg24
Thread-3 free lock
Thread-2 get lock
Thread-2 get:msg25
Thread-2 free lock
Thread-4 get lock
Thread-4 get:msg26
Thread-4 free lock
Thread-3 get lock
Thread-3 get:msg27
Thread-3 free lock
Thread-2 get lock
Thread-2 get:msg28
Thread-2 free lock
Thread-4 get lock
Thread-4 get:msg29
Thread-4 free lock
Thread-3 get lock
Thread-3 get:msg30
Thread-3 free lock
Thread-2 get lock
Thread-2 free lock
Thread-4 get lock
Thread-4 free lock
Thread-3 get lock
Thread-3 free lock
Thread-2 get lock
Thread-2 get:msg31
Thread-2 free lock
Thread-2 get lock
Thread-2 get:msg32
Thread-2 free lock
Thread-2 get lock
Thread-2 get:msg33
Thread-2 free lock
Thread-4 get lock
Thread-4 get:msg34
Thread-4 free lock
Thread-3 get lock
Thread-3 get:msg35
Thread-3 free lock
Thread-2 get lock
Thread-2 get:msg36
Thread-2 free lock
Thread-4 get lock
Thread-4 free lock
Thread-3 get lock
Thread-3 free lock
Thread-2 get lock
Thread-2 free lock
Thread-4 get lock
Thread-4 get:msg37
Thread-4 free lock
Thread-2 get lock
Thread-2 get:msg38
Thread-2 free lock
Thread-3 get lock
Thread-3 get:msg39
Thread-3 free lock
Thread-4 get lock
Thread-4 get:msg40
Thread-4 free lock
Thread-2 get lock
Thread-2 get:msg41
Thread-2 free lock
Thread-3 get lock
Thread-3 free lock
Thread-4 get lock
Thread-4 free lock
Thread-2 get lock
Thread-2 free lock
Thread-4 get lock
Thread-4 get:msg42
Thread-4 free lock
Thread-2 get lock
Thread-2 get:msg43
Thread-2 free lock
Thread-3 get lock
Thread-3 get:msg44
Thread-3 free lock
Thread-4 get lock
Thread-4 free lock
Thread-2 get lock
Thread-2 free lock
Thread-3 get lock
Thread-3 free lock
Thread-4 get lock
Thread-4 get:msg45
Thread-4 free lock
....

---

可见3个消费者线程随机获取到锁,数据在锁中被递增取出,没有重复和遗漏

end

Zookeeper--分布式锁和消息队列的更多相关文章

  1. redis分布式锁和消息队列

    最近博主在看redis的时候发现了两种redis使用方式,与之前redis作为缓存不同,利用的是redis可设置key的有效时间和redis的BRPOP命令. 分布式锁 由于目前一些编程语言,如PHP ...

  2. java-spring基于redis单机版(redisTemplate)实现的分布式锁+redis消息队列,可用于秒杀,定时器,高并发,抢购

    此教程不涉及整合spring整合redis,可另行查阅资料教程. 代码: RedisLock package com.cashloan.analytics.utils; import org.slf4 ...

  3. 分布式缓存重建并发冲突和zookeeper分布式锁解决方案

    如果缓存服务在本地的ehcache中都读取不到数据. 这个时候就意味着,需要重新到源头的服务中去拉去数据,拉取到数据之后,赶紧先给nginx的请求返回,同时将数据写入ehcache和redis中 分布 ...

  4. Zookeeper 分布式锁 (图解+秒懂+史上最全)

    文章很长,而且持续更新,建议收藏起来,慢慢读! 高并发 发烧友社群:疯狂创客圈(总入口) 奉上以下珍贵的学习资源: 疯狂创客圈 经典图书 : 极致经典 + 社群大片好评 < Java 高并发 三 ...

  5. ZooKeeper 分布式锁实现

    1 场景描述 在分布式应用, 往往存在多个进程提供同一服务. 这些进程有可能在相同的机器上, 也有可能分布在不同的机器上. 如果这些进程共享了一些资源, 可能就需要分布式锁来锁定对这些资源的访问. 2 ...

  6. ZooKeeper分布式锁浅谈(一)

    一.概述 清明节的时候写了一篇分布式锁概述,里面介绍了分布式锁实现的几种方式,其实那时候我一直沉迷于使用redis的悲观锁和乐观锁来实现分布式锁,直到一个血案的引发才让我重新认识了redis分布式锁的 ...

  7. 分布式锁(一) Zookeeper分布式锁

    什么是Zookeeper? Zookeeper(业界简称zk)是一种提供配置管理.分布式协同以及命名的中心化服务,这些提供的功能都是分布式系统中非常底层且必不可少的基本功能,但是如果自己实现这些功能而 ...

  8. Curator Zookeeper分布式锁

    Curator Zookeeper分布式锁 pom.xml中添加如下配置 <!-- https://mvnrepository.com/artifact/org.apache.curator/c ...

  9. [转载] zookeeper 分布式锁服务

    转载自http://www.cnblogs.com/shanyou/archive/2012/09/22/2697818.html 分布式锁服务在大家的项目中或许用的不多,因为大家都把排他放在数据库那 ...

  10. 跟着大神学zookeeper分布式锁实现-----来自Ruthless

    前几天分享了@Ruthless大神的Redis锁,发现和大家都学习了很多东西.因为分布式锁里面,最好的实现是zookeeper的分布式锁.所以在这里把实现方式和大家分享一下. zookeeper分布式 ...

随机推荐

  1. 连接GitHub的方法

    连接到GitHub 首先在本地创建 ssh key: ssh-keygen -t rsa -C "your_email@youremail.com" 后面的 your_email@ ...

  2. T-sql 中 truncate 、delete与drop操作数据库的区别

    1.truncate和不带where子句的delete.以及drop都会删除表内的数据. 2.drop.truncate都是DDL语句(数据定义语言),执行后会自动提交. 具体使用语法如下: --删除 ...

  3. Yii ExtendedActiveRecord 增强版 ActiveRecord 增加多数据库连接绑定功能

    ExtendedActiveRecord 继承自 CActiveRecord,因此基础功能与 CActiveRecord 无异 为添加对多数据库连接的支持,增加了对 connectionName() ...

  4. Poj 3318 Matrix Multiplication( 矩阵压缩)

    Matrix Multiplication Time Limit: 2000MS   Memory Limit: 65536K Total Submissions: 18928   Accepted: ...

  5. Solr后台管理界面配置

    配置来源 https://stackoverflow.com/questions/28043957/how-to-set-apache-solr-admin-password 注意:配置用户名密码后 ...

  6. 转载-lvs官方文档04-LVS集群的负载调度

    LVS集群的负载调度 章文嵩 (wensong@linux-vs.org) 2002 年 5 月 本文主要讲述了LVS集群的IP负载均衡软件IPVS在内核中实现的各种连接调度算法.针对请求的服务时间变 ...

  7. C与C++基础知识补遗

    本随笔用来记载项目开发中遇到的以前没掌握的C/C++基础知识 void * buffer; 无类型指针,可以指向任何类型数据.ANSI标准规定无类型指针不能进行算法,而GNU规定无类型指针算法操作与c ...

  8. Javascript-- jQuery动画篇(淡入淡出效果)

    淡入淡出效果 jQuery 提供了下面几种方法可以实现显示的淡入淡出效果: fadeIn() fadeOut() fadeToggle() fadeTo() fadeIn()方法 fadeIn()实现 ...

  9. 报错:Type mismatch: cannot convert from Object to Car

    问题描述: 一个非常简单的spring项目,用静态工厂方法配置bean实例.项目的目录结构如下: 代码如下: Car.java package com.tt.spring.beans.factory; ...

  10. Linux:Aircrack-ng

    Aircrack-ng 工具主要有 airmon-ng 处理网卡工作模式 airodump-ng 抓包 aircrack-ng 破解 aireplay-ng 发包,干扰 另外还要用到以下 linux ...