利用redis实现分布式锁,快速解决高并发时的线程安全问题
实际工作中,经常会遇到多线程并发时的类似抢购的功能,本篇描述一个简单的redis分布式锁实现的多线程抢票功能。
直接上代码。首先按照?T例,给出一个错误的示范:
我们可以看看,当20个线程一起来抢10张票的时候,会发生什么事。
package com.tiger.utils; public class TestMutilThread { // 总票量 public static int count = 10; public static void main(String[] args) { statrtMulti(); } public static void statrtMulti() { for (int i = 1; i <= 20; i++) { TicketRunnable tickrunner = new TicketRunnable(); Thread thread = new Thread(tickrunner, "Thread No: " + i); thread.start(); } } public static class TicketRunnable implements Runnable { @Override public void run() { System.out.println(Thread.currentThread().getName() + " start " + count); // TODO Auto-generated method stub // logger.info(Thread.currentThread().getName() // + " really start" + count); if (count <= 0) { System.out.println(Thread.currentThread().getName() + " ticket sold out ! No tickets remained!" + count); return; } else { count = count - 1; System.out.println(Thread.currentThread().getName() + " bought a ticket,now remaining :" + (count)); } } } }
测试结果,从结果可以看到,票数在不同的线程中已经出现混乱。
Thread No: 2 start 10 Thread No: 6 start 10 Thread No: 4 start 10 Thread No: 5 start 10 Thread No: 3 start 10 Thread No: 9 start 6 Thread No: 1 start 10 Thread No: 1 bought a ticket,now remaining :3 Thread No: 9 bought a ticket,now remaining :4 Thread No: 3 bought a ticket,now remaining :5 Thread No: 12 start 3 Thread No: 5 bought a ticket,now remaining :6 Thread No: 4 bought a ticket,now remaining :7 Thread No: 8 start 7 Thread No: 7 start 8 Thread No: 12 bought a ticket,now remaining :1 Thread No: 14 start 0 Thread No: 6 bought a ticket,now remaining :8 Thread No: 16 start 0 Thread No: 2 bought a ticket,now remaining :9 Thread No: 16 ticket sold out ! No tickets remained!0 Thread No: 14 ticket sold out ! No tickets remained!0 Thread No: 18 start 0 Thread No: 18 ticket sold out ! No tickets remained!0 Thread No: 7 bought a ticket,now remaining :0 Thread No: 15 start 0 Thread No: 8 bought a ticket,now remaining :1 Thread No: 13 start 2 Thread No: 19 start 0 Thread No: 11 start 3 Thread No: 11 ticket sold out ! No tickets remained!0 Thread No: 10 start 3 Thread No: 10 ticket sold out ! No tickets remained!0 Thread No: 19 ticket sold out ! No tickets remained!0 Thread No: 13 ticket sold out ! No tickets remained!0 Thread No: 20 start 0 Thread No: 20 ticket sold out ! No tickets remained!0 Thread No: 15 ticket sold out ! No tickets remained!0 Thread No: 17 start 0 Thread No: 17 ticket sold out ! No tickets remained!0
为了解决多线程时出现的混乱问题,这里?o出真正的测试类!!!
真正的测试类,这里启动20个线程,来抢10张票。
RedisTemplate 是用来实现redis操作的,由spring进行集成。这里是使用到了RedisTemplate,所以我以构造器的形式在外部将RedisTemplate传入到测试类中。
MultiTestLock 是用来实现加锁的工具类。
总票数使用volatile关键字,实现多线程时变量在系统内存中的可见性,这点可以去了解下volatile关键字的作用。
TicketRunnable用于模拟抢票功能。
其中由于lock与unlock之间存在if判断,为保证线程安全,这里使用synchronized来保证。
测试类:
package com.tiger.utils; import java.io.Serializable; import org.slf4j.Logger; import org.slf4j.LoggerFactory; import org.springframework.data.redis.core.RedisTemplate; public class MultiConsumer { Logger logger=LoggerFactory.getLogger(MultiTestLock.class); private RedisTemplate<Serializable, Serializable> redisTemplate; public MultiTestLock lock; //总票量 public volatile static int count = 10; public void statrtMulti() { lock = new MultiTestLock(redisTemplate); for (int i = 1; i <= 20; i++) { TicketRunnable tickrunner = new TicketRunnable(); Thread thread = new Thread(tickrunner, "Thread No: " + i); thread.start(); } } public class TicketRunnable implements Runnable { @Override public void run() { logger.info(Thread.currentThread().getName() + " start " + count); // TODO Auto-generated method stub if (count > 0) { // logger.info(Thread.currentThread().getName() // + " really start" + count); lock.lock(); synchronized (this) { if(count<=0){ logger.info(Thread.currentThread().getName() + " ticket sold out ! No tickets remained!" + count); lock.unlock(); return; }else{ count=count-1; logger.info(Thread.currentThread().getName() + " bought a ticket,now remaining :" + (count)); } } lock.unlock(); }else{ logger.info(Thread.currentThread().getName() + " ticket sold out !" + count); } } } public RedisTemplate<Serializable, Serializable> getRedisTemplate() { return redisTemplate; } public void setRedisTemplate( RedisTemplate<Serializable, Serializable> redisTemplate) { this.redisTemplate = redisTemplate; } public MultiConsumer(RedisTemplate<Serializable, Serializable> redisTemplate) { super(); this.redisTemplate = redisTemplate; } }
Lock工具类:
我们知道为保证线程安全,程序中执行的操作必须时原子的。redis后续的版本中可以使用set key同时设置expire超时时间。
想起上次去 电信翼支付 面试时,面试官问过一个问题:分布式锁如何防止死锁,问题关键在于我们在分布式中进行加锁操作时成功了,但是后续业务操作完毕执行解锁时出现失败。导致分布式锁无法释放。出现死锁,后续的加锁无法正常进行。所以这里设置expire超时时间的目的就是防止出现解锁失败的情况,这样,即使解锁失败了,分布式锁依然会在超时时间过了之后自动释放。
具体在代码中也有注释,也可以作为参考。
package com.tiger.utils; import java.io.Serializable; import java.util.Arrays; import java.util.Collections; import java.util.HashMap; import java.util.Iterator; import java.util.List; import java.util.Random; import java.util.concurrent.TimeUnit; import java.util.concurrent.locks.Condition; import java.util.concurrent.locks.Lock; import javax.sound.midi.MidiDevice.Info; import org.slf4j.Logger; import org.slf4j.LoggerFactory; import org.springframework.dao.DataAccessException; import org.springframework.data.redis.core.RedisOperations; import org.springframework.data.redis.core.RedisTemplate; import org.springframework.data.redis.core.SessionCallback; import org.springframework.data.redis.core.script.RedisScript; public class MultiTestLock implements Lock { Logger logger=LoggerFactory.getLogger(MultiTestLock.class); private RedisTemplate<Serializable, Serializable> redisTemplate; public MultiTestLock(RedisTemplate<Serializable, Serializable> redisTemplate) { super(); this.redisTemplate = redisTemplate; } @Override public void lock() { //这里使用while循环强制线程进来之后先进行抢锁操作。只有抢到锁才能进行后续操作 while(true){ if(tryLock()){ try { //这里让线程睡500毫秒的目的是为了模拟业务耗时,确保业务结束时之前设置的值正好打到超时时间, //实际生产中可能有偏差,这里需要经验 Thread.sleep(500l); // logger.info(Thread.currentThread().getName()+" time to awake"); return; } catch (InterruptedException e) { // TODO Auto-generated catch block e.printStackTrace(); } }else{ try { //这里设置一个随机毫秒的sleep目的时降低while循环的频率 Thread.sleep(new Random().nextInt(200)+100); } catch (InterruptedException e) { // TODO Auto-generated catch block e.printStackTrace(); } } } } @Override public boolean tryLock() { //这里也可以选用transactionSupport支持事务操作 SessionCallback<Object> sessionCallback=new SessionCallback<Object>() { @Override public Object execute(RedisOperations operations) throws DataAccessException { operations.multi(); operations.opsForValue().setIfAbsent("secret", "answer"); //设置超时时间要根据业务实际的可能处理时间来,是一个经验值 operations.expire("secret", 500l, TimeUnit.MILLISECONDS); Object object=operations.exec(); return object; } }; //执行两部操作,这里会拿到一个数组值 [true,true],分别对应上述两部操作的结果,如果中途出现第一次为false则表明第一步set值出错 List<Boolean> result=(List) redisTemplate.execute(sessionCallback); // logger.info(Thread.currentThread().getName()+" try lock "+ result); if(true==result.get(0)||"true".equals(result.get(0)+"")){ logger.info(Thread.currentThread().getName()+" try lock success"); return true; }else{ return false; } } @Override public boolean tryLock(long arg0, TimeUnit arg1) throws InterruptedException { // TODO Auto-generated method stub return false; } @Override public void unlock() { //unlock操作直接删除锁,如果执行完还没有达到超时时间则直接删除,让后续的线程进行继续操作。起到补刀的作用,确保锁已经超时或被删除 SessionCallback<Object> sessionCallback=new SessionCallback<Object>() { @Override public Object execute(RedisOperations operations) throws DataAccessException { operations.multi(); operations.delete("secret"); Object object=operations.exec(); return object; } }; Object result=redisTemplate.execute(sessionCallback); } @Override public void lockInterruptibly() throws InterruptedException { // TODO Auto-generated method stub } @Override public Condition newCondition() { // TODO Auto-generated method stub return null; } public RedisTemplate<Serializable, Serializable> getRedisTemplate() { return redisTemplate; } public void setRedisTemplate( RedisTemplate<Serializable, Serializable> redisTemplate) { this.redisTemplate = redisTemplate; } }
执行结果
可以看到,票数稳步减少,后续没有抢到锁的线程余票为0,无票可抢。
tips:
这其中也出现了一个问题,redis进行多部封装操作时,系统报错:ERR EXEC without MULTI
后经过查阅发现问题出在:
在spring中,多次执行MULTI命令不会报错,因为第一次执行时,会将其内部的一个isInMulti变量设为true,后续每次执行命令是都会检查这个变量,如果为true,则不执行命令。
而多次执行EXEC命令则会报开头说的"ERR EXEC without MULTI"错误。