前两个版本的代码 都或多或少存在一定的问题,虽然可能微乎其微,但是程序需要严谨再严谨, 第一个版本问题: 局限于单机版,依赖于 Jvm的锁 第二个版本问题: 极端情况下,解锁逻辑的问题,线程B的锁,可能会被线程A解掉,这种情况实际上是不合理的。 1. 由于是客户端自己生成过期时间,所以需要强制要求分布式下每个客户端的时间必须同步。 2. 当锁过期的时候,如果多个客户端同时执行jedis.getSet()方法,那么虽然最终只有一个客户端可以加锁, 但是这个客户端的锁的过期时间可能被其他客户端覆盖。 3. 锁不具备拥有者标识,即任何客户端都可以解锁。 版本一: http://www.cnblogs.com/xifenglou/p/8807323.html
版本二: http://www.cnblogs.com/xifenglou/p/8883717.html所以基于以上问题,第三个版本出来了, Talk is cheap, show me the code!import org.springframework.util.StopWatch; import redis.clients.jedis.Jedis;import java.util.concurrent.CountDownLatch; import java.util.concurrent.CyclicBarrier; import java.util.concurrent.ExecutorService; import java.util.concurrent.Executors;/*** 使用RedisTool.tryGetDistributedLock* 实现 分布式锁* 终极版本*/ public class TicketRunnable3 implements Runnable {private CountDownLatch count;private CyclicBarrier barrier;private static final Integer Lock_Timeout = 10000;private static final String lockKey = "LockKey";private volatile static boolean working = true;public TicketRunnable3(CountDownLatch count, CyclicBarrier barrier) {this.count = count;this.barrier = barrier;}private int num = 20; // 总票数 此处可随意 写一个数,保证线程能运行起来,真正的共享变量不应该写死在程序中, 应该从redis中获取,这样模拟多进程多线程的并发访问public void sellTicket(Jedis jedis) {String name = Thread.currentThread().getName();try{boolean getLock = RedisTool.tryGetDistributedLock(jedis,lockKey, name,Lock_Timeout);if( getLock){if(!working)return;// Do your jobnum = Integer.parseInt(jedis.get("ticket"));if (num > 0) {num--;jedis.set("ticket",num+"");if(num!=0)System.out.println("================"+Thread.currentThread().getName()+"================= 售出票号" + (num+1)+",还剩" + num + "张票--" );else {System.out.println("================"+Thread.currentThread().getName()+"================= 售出票号" + (num+1)+",票已经票完!--");working = false;}}}else{//System.out.println();if(!working)return;System.out.println(Thread.currentThread().getName()+" Try to get the Lock,and wait 20 millisecond....");Thread.sleep(10);}}catch(Exception e){System.out.println(e);}finally {try {if(RedisTool.releaseDistributedLock(jedis,lockKey,name)){Thread.sleep(30);}}catch (Exception e ) {e.printStackTrace();}}}@Overridepublic void run() {System.out.println(Thread.currentThread().getName()+"到达,等待中...");Jedis jedis = new Jedis("localhost", 6379);try{barrier.await(); // 此处阻塞 等所有线程都到位后 一起进行抢票if(Thread.currentThread().getName().equals("pool-1-thread-1")){System.out.println("-----------------全部线程准备就绪,开始抢票------------------");}else {Thread.sleep(5);}while (working) {sellTicket(jedis);}count.countDown(); //当前线程结束后,计数器-1}catch (Exception e){e.printStackTrace();}}/**** @param args*/public static void main(String[] args) {int threadNum = 5; //模拟多个窗口 进行售票final CyclicBarrier barrier = new CyclicBarrier(threadNum);final CountDownLatch count = new CountDownLatch(threadNum); // 用于统计 执行时长StopWatch watch = new StopWatch();watch.start();TicketRunnable3 tickets = new TicketRunnable3(count,barrier);ExecutorService executorService = Executors.newFixedThreadPool(threadNum);//ExecutorService executorService = Executors.newCachedThreadPool();for (int i = 0; i < threadNum; i++) { //此处 设置数值 受限于 线程池中的数量executorService.submit(tickets);}try {count.await();executorService.shutdown();watch.stop();System.out.println("耗 时:" + watch.getTotalTimeSeconds() + "秒");} catch (InterruptedException e) {e.printStackTrace();}} }
import redis.clients.jedis.Jedis; import java.util.Collections;public class RedisTool {private static final String LOCK_SUCCESS = "OK";private static final String SET_IF_NOT_EXIST = "NX";private static final String SET_WITH_EXPIRE_TIME = "PX";private static final Long RELEASE_SUCCESS = 1L;/** * 尝试获取分布式锁* @param jedis Redis客户端* @param lockKey 锁* @param requestId 请求标识* @param expireTime 超期时间* @return 是否获取成功*/public static boolean tryGetDistributedLock(Jedis jedis, String lockKey, String requestId, int expireTime) {String result = jedis.set(lockKey, requestId, SET_IF_NOT_EXIST, SET_WITH_EXPIRE_TIME, expireTime);if (LOCK_SUCCESS.equals(result)) {System.out.println("=============="+Thread.currentThread().getName()+"=============== 获取到锁,开始工作!");return true;}return false;}/*** 释放分布式锁* @param jedis Redis客户端* @param lockKey 锁* @param requestId 请求标识* @return 是否释放成功*/public static boolean releaseDistributedLock(Jedis jedis, String lockKey, String requestId) {String script = "if redis.call('get', KEYS[1]) == ARGV[1] then return redis.call('del', KEYS[1]) else return 0 end";Object result = jedis.eval(script, Collections.singletonList(lockKey), Collections.singletonList(requestId));if (RELEASE_SUCCESS.equals(result)) {System.out.println("=============="+Thread.currentThread().getName()+"=============== 解锁成功!");return true;}return false;}}
解锁部分,我们将Lua代码传到jedis.eval()方法里,并使参数KEYS[1]赋值为lockKey,ARGV[1]赋值为requestId。eval()方法是将Lua代码交给Redis服务端执行。
那么这段Lua代码的功能是什么呢?其实很简单,首先获取锁对应的value值,检查是否与requestId相等,如果相等则删除锁(解锁)。那么为什么要使用Lua语言来实现呢?因为要确保上述操作是原子性的。源于Redis的特性,下面是官网对eval命令的部分解释:
简单来说,就是在eval命令执行Lua代码的时候,Lua代码将被当成一个命令去执行,并且直到eval命令执行完成,Redis才会执行其他命令。
运行结果如下:
欢迎留言,期待更深层次的探讨!
针对 上述代码,使用两个类 运行,
TicketRunnable3 TicketRunnable4 模拟多进程 多线程场景 ,
场景1: 运行时长 > 过期时长
此时: 锁自动失效, 线程均不用解锁,即使解锁也是失败!
代码及运行结果如下:
import org.springframework.util.StopWatch; import redis.clients.jedis.Jedis;import java.util.concurrent.CountDownLatch; import java.util.concurrent.CyclicBarrier; import java.util.concurrent.ExecutorService; import java.util.concurrent.Executors;/*** 使用RedisTool.tryGetDistributedLock* 实现 分布式锁* 终极版本*/ public class TicketRunnable4 implements Runnable {private CountDownLatch count;private CyclicBarrier barrier;private static final Integer Lock_Timeout = 3000; // 过期时间 代表 3秒后过期private static final Integer ExecuteTime = 5000;private static final Integer RetryInterval = 20;private static final String lockKey = "LockKey";private volatile static boolean working = true;public TicketRunnable4(CountDownLatch count, CyclicBarrier barrier) {this.count = count;this.barrier = barrier;}private int num = 20; // 总票数public void sellTicket(Jedis jedis) {String name = Thread.currentThread().getName();boolean gotLock = false;try{gotLock = RedisTool.tryGetDistributedLock(jedis,lockKey, name,Lock_Timeout);if( gotLock && working){// Do your jobnum = Integer.parseInt(jedis.get("ticket"));if (num > 0) {num--;jedis.set("ticket",num+"");if(num!=0)System.out.println("=============="+name+"=============== 售出票号" + (num+1)+",还剩" + num + "张票--" );else {System.out.println("=============="+name+"=============== 售出票号" + (num+1)+",票已经票完!--");return;}}if(num == 0){System.out.println("=============="+name+"============票已经被抢空啦");working = false;}Thread.sleep(ExecuteTime);}else{//System.out.println();//System.out.println(name+" Try to get the Lock,and wait "+RetryInterval+" millisecond....");Thread.sleep(RetryInterval);}}catch(Exception e){System.out.println(e);}finally {try {if(!gotLock||!working) //未获取到锁的线程不用解锁return;/*** 解锁成功后 sleep, 尝试让出cpu给其他线程机会* 解锁失败 说明锁已经失效 被其他线程获取到*/if(RedisTool.releaseDistributedLock(jedis,lockKey,name)){Thread.sleep(100);}}catch (Exception e ) {e.printStackTrace();}}}@Overridepublic void run() {String prefix = "#";String threadName = Thread.currentThread().getName();Thread.currentThread().setName(prefix+threadName);System.out.println(Thread.currentThread().getName()+"到达,等待中...");Jedis jedis = new Jedis("localhost", 6379);try{barrier.await(); // 此处阻塞 等所有线程都到位后 一起进行抢票if(Thread.currentThread().getName().equals(prefix+"pool-1-thread-2")){System.out.println("-----------------全部线程准备就绪,开始抢票------------------");}else {Thread.sleep(5);}while (working) {sellTicket(jedis);}count.countDown(); //当前线程结束后,计数器-1}catch (Exception e){e.printStackTrace();}}/**** @param args*/public static void main(String[] args) {int threadNum = 3; //模拟多个窗口 进行售票final CyclicBarrier barrier = new CyclicBarrier(threadNum);final CountDownLatch count = new CountDownLatch(threadNum); // 用于统计 执行时长StopWatch watch = new StopWatch();watch.start();TicketRunnable4 tickets = new TicketRunnable4(count,barrier);ExecutorService executorService = Executors.newFixedThreadPool(threadNum);//ExecutorService executorService = Executors.newCachedThreadPool();for (int i = 0; i < threadNum; i++) { //此处 设置数值 受限于 线程池中的数量executorService.submit(tickets);}try {count.await();executorService.shutdown();watch.stop();System.out.println("耗 时:" + watch.getTotalTimeSeconds() + "秒");} catch (InterruptedException e) {e.printStackTrace();}} }
TicketRunnable3 售出 10 9 8 6 5 3 2 1 票号。
TicketRunnable4售出 7 4 两个票号
合计10张票,模拟结束!
场景2: 运行时长 < 过期时长
此时: 此时需要有锁线程去释放锁,这样多线程再去竞争获取锁。
修改代码:
private static final Integer Lock_Timeout = 5000; // 将时间从3秒改为5秒 private static final Integer ExecuteTime = 3000; // 将执行时间5秒改为3秒运行结果如下:
一个进程售出 10 9 8 6 4 3 1 票号 另一进程售出 7 5 2 票号 此时 每个线程完成任务后,均需要释放锁,这样本地线程或是异地线程 才能获取到锁,这样才能有机会进行任务的执行!