redis的事务类似于队列操作，执行过程分为三步：

1. 开启事务
2. 入队操作
3. 执行事务

使用到的几个命令如下：

下面我们使用一个springboot的代码操作来说明这几个命令的含义：

package com.test.spring;

import org.springframework.beans.factory.annotation.Autowired;
import org.springframework.boot.SpringApplication;
import org.springframework.boot.autoconfigure.SpringBootApplication;
import org.springframework.dao.DataAccessException;
import org.springframework.data.redis.RedisSystemException;
import org.springframework.data.redis.connection.RedisConnection;
import org.springframework.data.redis.core.RedisCallback;
import org.springframework.data.redis.core.StringRedisTemplate;
import org.springframework.web.bind.annotation.RequestMapping;
import org.springframework.web.bind.annotation.RestController;
@SpringBootApplication
@RestController
public class TestSpringApplication {
	@Autowired
	private StringRedisTemplate stringRedisTemplate;

	@RequestMapping("/exec1")
	public String test1(){
		stringRedisTemplate.execute(new RedisCallback<Boolean>() {
			@Override
			public Boolean doInRedis(RedisConnection connection) throws DataAccessException {
				connection.multi();
				connection.commands().set("k1".getBytes(),"1".getBytes());
				connection.commands().set("k2".getBytes(),"2".getBytes());
				connection.exec();
				return true;
			}
		});

		return "k1="+stringRedisTemplate.opsForValue().get("k1")+",k2="+stringRedisTemplate.opsForValue().get("k2");
	}
	@RequestMapping("/exec2")
	public String exec2(){

		try{
			stringRedisTemplate.execute(new RedisCallback<Boolean>() {
				private int i=0;
				@Override
				public Boolean doInRedis(RedisConnection connection) throws DataAccessException {
					connection.multi();
					connection.commands().set("k1".getBytes(),"11".getBytes());
					connection.commands().set("k2".getBytes(),"22".getBytes());
					if(i==0){
						throw new RedisSystemException("一个异常",new RuntimeException("1"));
					}
					connection.exec();
					return true;
				}
			});
		}catch (Exception e){
			e.printStackTrace();
		}
		return "k1="+stringRedisTemplate.opsForValue().get("k1")+",k2="+stringRedisTemplate.opsForValue().get("k2");
	}

	@RequestMapping("/discard")
	public String discard(){
		try{
			stringRedisTemplate.execute(new RedisCallback<Boolean>() {
				@Override
				public Boolean doInRedis(RedisConnection connection) throws DataAccessException {
					connection.multi();
					connection.commands().set("k3".getBytes(),"3".getBytes());
					connection.commands().set("k4".getBytes(),"4".getBytes());
					connection.discard();
					connection.exec();
					return true;
				}
			});
		}catch (Exception e){
			e.printStackTrace();
		}

		return "k3="+stringRedisTemplate.opsForValue().get("k3")+",k4="+stringRedisTemplate.opsForValue().get("k4");
	}
	@RequestMapping("/watch1")
	public String watch1(){
		//开启一个线程
		new Thread(()->{
			stringRedisTemplate.execute(new RedisCallback<Boolean>() {
				@Override
				public Boolean doInRedis(RedisConnection connection) throws DataAccessException {
					try {
						connection.watch("k5".getBytes());
						connection.multi();
						connection.commands().set("k5".getBytes(), "5".getBytes());
						//休眠5秒钟在提交事务
						try {
							Thread.sleep(5000);
						} catch (InterruptedException e) {
						}
						connection.exec();
					}catch (Exception e){
						e.printStackTrace();
					}

					return true;
				}
			});
		}).start();

		//开启一个线程
		new Thread(()->{
			stringRedisTemplate.execute(new RedisCallback<Boolean>() {
				@Override
				public Boolean doInRedis(RedisConnection connection) throws DataAccessException {
					try {
						connection.multi();
						connection.commands().set("k5".getBytes(), "55".getBytes());
						connection.exec();
					}catch (Exception e){
						e.printStackTrace();
					}
					return true;
				}
			});
		}).start();
		return "success";
	}
	@RequestMapping("/watch2")
	public String watch2(){
		return "k5="+stringRedisTemplate.opsForValue().get("k5");
	}
	public static void main(String[] args) {
		SpringApplication.run(TestSpringApplication.class, args);
	}

}

1. exec1
   正常流程，使用curl进行测试会返回：k1=1,k2=2
2. exec2
   我们模拟在事务队列中发送异常，会发现这段设值不成功，测试返回：k1=1,k2=2
3. discard
   事务回滚，我们先回滚，再提交，后台会抛出：ERR EXEC without MULTI错误，说明设值失败
4. watch1、watch2
   这里我们模拟两个线程，第一个线程先监听key，然后等待5秒钟，但是第二个线程直接去修改这个key，当5秒结束时，第一个线程再去提交事务时，会发现已经失效了，然后我们再通过watch2去查询值，测试返回：k5=55，说明线程1事务失效

最后再说明一下unwatch，每次操作exec()后，底层会自动调用unwatch，所以我们可以不用显示去调用unwatch命令。