redis实现消息队列&发布/订阅模式使用
在项目中用到了redis作为缓存，再学习了ActiveMq之后想着用redis实现简单的消息队列，下面做记录。
Redis的列表类型键可以用来实现队列，并且支持阻塞式读取，可以很容易的实现一个高性能的优先队列。同时在更高层面上，Redis还支持"发布/订阅"的消息模式，可以基于此构建一个聊天系统。
一、redis的列表类型天生支持用作消息队列。(类似于MQ的队列模型--任何时候都可以消费，一条消息只能消费一次)
list操作参考:https://www.cnblogs.com/qlqwjy/p/7789125.html
在Redis中，List类型是按照插入顺序排序的字符串链表。和数据结构中的普通链表一样，我们可以在其头部(left)和尾部(right)添加新的元素。在插入时，如果该键并不存在，Redis将为该键创建一个新的链表。与此相反，如果链表中所有的元素均被移除，那么该键也将会被从数据库中删除。List中可以包含的最大元素数量是4294967295。
从元素插入和删除的效率视角来看，如果我们是在链表的两头插入或删除元素，这将会是非常高效的操作，即使链表中已经存储了百万条记录，该操作也可以在常量时间内完成。然而需要说明的是，如果元素插入或删除操作是作用于链表中间，那将会是非常低效的。相信对于有良好数据结构基础的开发者而言，这一点并不难理解。(类似于java的ArrayList)
redis对list的操作命令中。L表示从左边(头部)开始插与弹出，R表示从右边(尾部)开始插与弹出。
1.redis中简单的操作list，简单的在命令行操作实现队列
(1)从左向右插入，从右向左弹出:
127.0.0.1:6379> lpush mylist a b c d (integer) 4 127.0.0.1:6379> lrange mylist 0 -1 1) "d" 2) "c" 3) "b" 4) "a" 127.0.0.1:6379> rpop mylist "a" 127.0.0.1:6379> rpop mylist "b"
执行完 lpush mylist a b c d 之后数据结构如下:(满足先进先出的队列模式)
执行完第一次:rpop mylist之后数据结构如下:
(2)从右向左插入，从左向右弹出:
127.0.0.1:6379> rpush mylist2 a b c d (integer) 4 127.0.0.1:6379> lrange mylist2 0 -1 1) "a" 2) "b" 3) "c" 4) "d" 127.0.0.1:6379> lpop mylist2 "a" 127.0.0.1:6379> lpop mylist2 "b"
执行完:rpush mylist2 a b c d之后的数据结构如下
第一次执行完 lpop mylist2 之后数据结构如下:(满足先进先出的队列模式)
2.JAVA程序实现消息队列
redis.properties
redis.url=localhost
redis.port=6379
redis.maxIdle=30
redis.minIdle=10
redis.maxTotal=100
redis.maxWait=10000
获取连接的工具类:
import redis.clients.jedis.Jedis; import redis.clients.jedis.JedisPool; import redis.clients.jedis.JedisPoolConfig; import java.io.IOException; import java.io.InputStream; import java.util.Properties; /** * @Author: qlq * @Description * @Date: 21:32 2018/10/9 */ public class JedisPoolUtils { private static JedisPool pool = null; static { //加载配置文件 InputStream in = JedisPoolUtils.class.getClassLoader().getResourceAsStream("redis.properties"); Properties pro = new Properties(); try { pro.load(in); } catch (IOException e) { e.printStackTrace(); } //获得池子对象 JedisPoolConfig poolConfig = new JedisPoolConfig(); poolConfig.setMaxIdle(Integer.parseInt(pro.get("redis.maxIdle").toString()));//最大闲置个数 poolConfig.setMaxWaitMillis(Integer.parseInt(pro.get("redis.maxWait").toString()));//最大闲置个数 poolConfig.setMinIdle(Integer.parseInt(pro.get("redis.minIdle").toString()));//最小闲置个数 poolConfig.setMaxTotal(Integer.parseInt(pro.get("redis.maxTotal").toString()));//最大连接数 pool = new JedisPool(poolConfig, pro.getProperty("redis.url"), Integer.parseInt(pro.get("redis.port").toString())); } //获得jedis资源的方法 public static Jedis getJedis() { return pool.getResource(); } public static void main(String[] args) { Jedis jedis = getJedis(); System.out.println(jedis); } }
(1)消息生产者:(开启5个线程生产消息)
import redis.clients.jedis.Jedis; /** * @Author: qlq * @Description * @Date: 21:29 2018/10/9 */ public class MessageProducer extends Thread { public static final String MESSAGE_KEY = "message:queue"; private volatile int count; public void putMessage(String message) { Jedis jedis = JedisPoolUtils.getJedis(); Long size = jedis.lpush(MESSAGE_KEY, message); System.out.println(Thread.currentThread().getName() + " put message,size=" + size + ",count=" + count); count++; } @Override public synchronized void run() { for (int i = 0; i < 5; i++) { putMessage("message" + count); } } public static void main(String[] args) { MessageProducer messageProducer = new MessageProducer(); Thread t1 = new Thread(messageProducer, "thread1"); Thread t2 = new Thread(messageProducer, "thread2"); Thread t3 = new Thread(messageProducer, "thread3"); Thread t4 = new Thread(messageProducer, "thread4"); Thread t5 = new Thread(messageProducer, "thread5"); t1.start(); t2.start(); t3.start(); t4.start(); t5.start(); } }
结果：(证明了redis是单线程操作，只能一个一个操作)
thread1 put message,size=1,count=0 thread1 put message,size=2,count=1 thread1 put message,size=3,count=2 thread1 put message,size=4,count=3 thread1 put message,size=5,count=4 thread3 put message,size=6,count=5 thread3 put message,size=7,count=6 thread3 put message,size=8,count=7 thread3 put message,size=9,count=8 thread3 put message,size=10,count=9 thread4 put message,size=11,count=10 thread4 put message,size=12,count=11 thread4 put message,size=13,count=12 thread4 put message,size=14,count=13 thread4 put message,size=15,count=14 thread5 put message,size=16,count=15 thread5 put message,size=17,count=16 thread5 put message,size=18,count=17 thread5 put message,size=19,count=18 thread5 put message,size=20,count=19 thread2 put message,size=21,count=20 thread2 put message,size=22,count=21 thread2 put message,size=23,count=22 thread2 put message,size=24,count=23 thread2 put message,size=25,count=24
redis后台查看:
127.0.0.1:6379> lrange message:queue 0 -1 1) "message24" 2) "message23" 3) "message22" 4) "message21" 5) "message20" 6) "message19" 7) "message18" 8) "message17" 9) "message16" 10) "message15" 11) "message14" 12) "message13" 13) "message12" 14) "message11" 15) "message10" 16) "message9" 17) "message8" 18) "message7" 19) "message6" 20) "message5" 21) "message4" 22) "message3" 23) "message2" 24) "message1" 25) "message0"
(2)消息消费者:(开启两个线程消费消息)
import redis.clients.jedis.Jedis; /** * @Author: qlq * @Description * @Date: 22:34 2018/10/9 */ public class MessageConsumer implements Runnable { public static final String MESSAGE_KEY = "message:queue"; private volatile int count; public void consumerMessage() { Jedis jedis = JedisPoolUtils.getJedis(); String message = jedis.rpop(MESSAGE_KEY); System.out.println(Thread.currentThread().getName() + " consumer message,message=" + message + ",count=" + count); count++; } @Override public void run() { while (true) { consumerMessage(); } } public static void main(String[] args) { MessageConsumer messageConsumer = new MessageConsumer(); Thread t1 = new Thread(messageConsumer, "thread6"); Thread t2 = new Thread(messageConsumer, "thread7"); t1.start(); t2.start(); } }
结果:(满足先进先出的规则)--虽然消息已经消费完了，但是仍然在不停的rpop，所以造成浪费
thread6 consumer message,message=message0,count=0 thread6 consumer message,message=message1,count=1 thread6 consumer message,message=message2,count=2 thread6 consumer message,message=message3,count=3 thread7 consumer message,message=message4,count=4 thread6 consumer message,message=message5,count=5 thread7 consumer message,message=message6,count=6 thread6 consumer message,message=message7,count=7 thread7 consumer message,message=message8,count=8 thread6 consumer message,message=message9,count=9 thread7 consumer message,message=message10,count=10 thread6 consumer message,message=message11,count=11 thread7 consumer message,message=message12,count=12 thread6 consumer message,message=message13,count=13 thread7 consumer message,message=message14,count=14 thread6 consumer message,message=message15,count=15 thread7 consumer message,message=message16,count=16 thread6 consumer message,message=message17,count=16 thread7 consumer message,message=message18,count=18 thread6 consumer message,message=message19,count=19 thread7 consumer message,message=message20,count=20 thread6 consumer message,message=message21,count=20 thread7 consumer message,message=message22,count=22 thread6 consumer message,message=message23,count=22 thread7 consumer message,message=message24,count=24 thread6 consumer message,message=null,count=25 thread7 consumer message,message=null,count=26 thread6 consumer message,message=null,count=27 thread7 consumer message,message=null,count=28 thread6 consumer message,message=null,count=28 thread7 consumer message,message=null,count=30 thread6 consumer message,message=null,count=31
...
但上述例子中消息消费者有一个问题存在，即需要不停的调用rpop方法查看List中是否有待处理消息。每调用一次都会发起一次连接，这会造成不必要的浪费。也许你会使用Thread.sleep()等方法让消费者线程隔一段时间再消费，但这样做有两个问题：
1）、如果生产者速度大于消费者消费速度，消息队列长度会一直增大，时间久了会占用大量内存空间。
2）、如果睡眠时间过长，这样不能处理一些时效性的消息，睡眠时间过短，也会在连接上造成比较大的开销。
补充:brpop和blpop实现阻塞读取(重要)
也就是上面的操作需要一直调用rpop命令或者lpop命令才可以实现不停的监听且消费消息。为了解决这一问题，redis提供了阻塞命令 brpop和blpop。下面以brpop命名为例进行试验:
brpop命令可以接收多个键，其完整的命令格式为 BRPOP key [key ...] timeout,如:brpop key1 0。意义是同时检测多个键，如果所有键都没有元素则阻塞，如果其中一个有元素则从该键中弹出该元素(会按照key的顺序进行读取，可以实现具有优先级的队列)。例如下面试验:
开启两个客户端，第一个客户端中采用brpop阻塞读取两个键:
127.0.0.1:6379> brpop mylist1 mylist2 0
第二个客户端增加mylist1 :
127.0.0.1:6379> lpush mylist1 1 2 (integer) 2
则在第一个客户端显示:
127.0.0.1:6379> brpop mylist1 mylist2 0 1) "mylist1" 2) "1" (56.31s)
也就是brpop会阻塞队列，并且每次也是弹出一个消息，如果没有消息会阻塞。
如果多个键都有元素则按照从左到右读取第一个键中的一个元素，例如我们现在queue1和queue2各自添加一个元素：
127.0.0.1:6379> lpush queue1 1 2 (integer) 2 127.0.0.1:6379> lpush queue2 3 4 (integer) 2
然后执行brpop命令:(会返回读取的key和value，第一个是返回的key，第二个是value)
127.0.0.1:6379> brpop queue1 queue2 2 1) "queue1" 2) "1"
借此特性可以实现区分优先级的任务队列。也就是brpop会按照key的顺序依次读取一个数据。
改造上面代码实现阻塞读取:
import redis.clients.jedis.Jedis; import java.util.List; /** * @Author: qlq * @Description * @Date: 22:34 2018/10/9 */ public class MessageConsumer implements Runnable { public static final String MESSAGE_KEY = "message:queue"; private volatile int count; private Jedis jedis = JedisPoolUtils.getJedis(); public void consumerMessage() { List<String> brpop = jedis.brpop(0, MESSAGE_KEY);//0是timeout,返回的是一个集合，第一个是消息的key，第二个是消息的内容 System.out.println(brpop); } @Override public void run() { while (true) { consumerMessage(); } } public static void main(String[] args) { MessageConsumer messageConsumer = new MessageConsumer(); Thread t1 = new Thread(messageConsumer, "thread6"); Thread t2 = new Thread(messageConsumer, "thread7"); t1.start(); t2.start(); } }
然后可以运行Customer，清空控制台，可以看到程序没有任何输出，阻塞在了brpop这儿。然后在打开Redis的客户端，输入指令client list，可以查看当前的连接个数。
当启动生产者生产消息之后，消费者会自动消费消息，而且消费者会阻塞直到有消息。
[message:queue, message0]
[message:queue, message1]
[message:queue, message2]
[message:queue, message3]
[message:queue, message4]
[message:queue, message5]
[message:queue, message6]
[message:queue, message7]
[message:queue, message8]
[message:queue, message9]
[message:queue, message10]
[message:queue, message11]
[message:queue, message12]
[message:queue, message13]
[message:queue, message14]
[message:queue, message15]
[message:queue, message16]
[message:queue, message17]
[message:queue, message18]
[message:queue, message19]
[message:queue, message20]
[message:queue, message21]
[message:queue, message22]
[message:queue, message23]
[message:queue, message24]
二、发布/订阅模式(类似于MQ的主题模式-只能消费订阅之后发布的消息，一个消息可以被多个订阅者消费)
1.客户端发布/订阅
1.1 普通的发布/订阅
除了实现任务队列外，redis还提供了一组命令可以让开发者实现"发布/订阅"(publish/subscribe)模式。"发布/订阅"模式同样可以实现进程间的消息传递，其原理如下:
"发布/订阅"模式包含两种角色，分别是发布者和订阅者。订阅者可以订阅一个或者多个频道(channel),而发布者可以向指定的频道(channel)发送消息，所有订阅此频道的订阅者都会收到此消息。
(1)发布消息
发布者发布消息的命令是 publish,用法是 publish channel message，如向 channel1.1说一声hi
127.0.0.1:6379> publish channel:1 hi (integer) 0
这样消息就发出去了。返回值表示接收这条消息的订阅者数量。发出去的消息不会被持久化，也就是有客户端订阅channel:1后只能接收到后续发布到该频道的消息，之前的就接收不到了。
(2)订阅频道
订阅频道的命令是 subscribe，可以同时订阅多个频道，用法是 subscribe channel1 [channel2 ...],例如新开一个客户端订阅上面频道:(不会收到消息，因为不会收到订阅之前就发布到该频道的消息)
127.0.0.1:6379> subscribe channel:1 Reading messages... (press Ctrl-C to quit) 1) "subscribe" 2) "channel:1" 3) (integer) 1
执行上面命令客户端会进入订阅状态，处于此状态下客户端不能使用除subscribe、unsubscribe、psubscribe和punsubscribe这四个属于"发布/订阅"之外的命令，否则会报错。
进入订阅状态后客户端可能收到3种类型的回复。每种类型的回复都包含3个值，第一个值是消息的类型，根据消类型的不同，第二个和第三个参数的含义可能不同。
消息类型的取值可能是以下3个:
(1)subscribe。表示订阅成功的反馈信息。第二个值是订阅成功的频道名称，第三个是当前客户端订阅的频道数量。
(2)message。表示接收到的消息，第二个值表示产生消息的频道名称，第三个值是消息的内容。
(3)unsubscribe。表示成功取消订阅某个频道。第二个值是对应的频道名称，第三个值是当前客户端订阅的频道数量，当此值为0时客户端会退出订阅状态，之后就可以执行其他非"发布/订阅"模式的命令了。
(3)第一个客户端重新向channel:1发送一条消息
127.0.0.1:6379> publish channel:1 hi (integer) 1
返回值表示订阅此频道的数量
c
上面订阅的客户端:
127.0.0.1:6379> subscribe channel:1 Reading messages... (press Ctrl-C to quit) 1) "subscribe" 2) "channel:1" 3) (integer) 1 1) "message" 2) "channel:1" 3) "hi"
红字部分表示成功的收到消息(依次是消息类型，频道，消息内容)
1.2 按照规则发布/订阅
除了可以使用subscribe命令订阅指定的频道外，还可以使用psubscribe命令订阅指定的规则。规则支持通配符格式。命令格式为 psubscribe pattern [pattern ...]订阅多个模式的频道。
通配符中?表示1个占位符，*表示任意个占位符(包括0)，?*表示1个以上占位符。
例如:
(1)订阅者订阅三个通配符频道
127.0.0.1:6379> psubscribe c? b* d?* Reading messages... (press Ctrl-C to quit) 1) "psubscribe" 2) "c?" 3) (integer) 1 1) "psubscribe" 2) "b*" 3) (integer) 2 1) "psubscribe" 2) "d?*" 3) (integer) 3
(2)新开一个客户端发送到指定频道
C:\Users\liqiang>redis-cli 127.0.0.1:6379> publish c m1 (integer) 0 127.0.0.1:6379> publish c1 m1 (integer) 1 127.0.0.1:6379> publish c11 m1 (integer) 0 127.0.0.1:6379> publish b m1 (integer) 1 127.0.0.1:6379> publish b1 m1 (integer) 1 127.0.0.1:6379> publish b11 m1 (integer) 1 127.0.0.1:6379> publish d m1 (integer) 0 127.0.0.1:6379> publish d1 m1 (integer) 1 127.0.0.1:6379> publish d11 m1 (integer) 1
上面返回值为1表示被订阅者所接受，可以匹配上面的通配符。
订阅者客户端:
127.0.0.1:6379> psubscribe c? b* d?* Reading messages... (press Ctrl-C to quit) 1) "psubscribe" 2) "c?" 3) (integer) 1 1) "psubscribe" 2) "b*" 3) (integer) 2 1) "psubscribe" 2) "d?*" 3) (integer) 3 1) "pmessage" 2) "c?" 3) "c1" 4) "m1" 1) "pmessage" 2) "b*" 3) "b" 4) "m1" 1) "pmessage" 2) "b*" 3) "b1" 4) "m1" 1) "pmessage" 2) "b*" 3) "b11" 4) "m1" 1) "pmessage" 2) "d?*" 3) "d1" 4) "m1" 1) "pmessage" 2) "d?*" 3) "d11" 4) "m1"
注意:
(1)使用psubscribe命令可以重复订阅同一个频道，如客户端执行了psubscribe c? c?*。这时向c1发布消息客户端会接受到两条消息，而同时publish命令的返回值是2而不是。.同样的，如果有另一个客户端执行了subscribe c1 和psubscribe c?*的话，向c1发送一条消息该客户顿也会受到两条消息(但是是两种类型:message和pmessage)，同时publish命令也返回2.
(2)punsubscribe命令可以退订指定的规则，用法是: punsubscribe [pattern [pattern ...]],如果没有参数则会退订所有规则。
(3)使用punsubscribe只能退订通过psubscribe命令订阅的规则，不会影响直接通过subscribe命令订阅的频道；同样unsubscribe命令也不会影响通过psubscribe命令订阅的规则。另外需要注意punsubscribe命令退订某个规则时不会将其中的通配符展开，而是进行严格的字符串匹配，所以punsubscribe * 无法退订c*规则，而是必须使用punsubscribe c*才可以退订。
2.Java程序实现发布者订阅者模式
1.生产者
import redis.clients.jedis.Jedis; /** * @Author: qlq * @Description * @Date: 21:29 2018/10/9 */ public class MessageProducer extends Thread { public static final String CHANNEL_KEY = "channel:1"; private volatile int count; public void putMessage(String message) { Jedis jedis = JedisPoolUtils.getJedis(); Long publish = jedis.publish(CHANNEL_KEY, message);//返回订阅者数量 System.out.println(Thread.currentThread().getName() + " put message,count=" + count+",subscriberNum="+publish); count++; } @Override public synchronized void run() { for (int i = 0; i < 5; i++) { putMessage("message" + count); } } public static void main(String[] args) { MessageProducer messageProducer = new MessageProducer(); Thread t1 = new Thread(messageProducer, "thread1"); Thread t2 = new Thread(messageProducer, "thread2"); Thread t3 = new Thread(messageProducer, "thread3"); Thread t4 = new Thread(messageProducer, "thread4"); Thread t5 = new Thread(messageProducer, "thread5"); t1.start(); t2.start(); t3.start(); t4.start(); t5.start(); } }
结果:
thread1 put message,count=0,subscriberNum=0
thread1 put message,count=1,subscriberNum=0
thread1 put message,count=2,subscriberNum=0
thread1 put message,count=3,subscriberNum=0
thread1 put message,count=4,subscriberNum=0
thread4 put message,count=5,subscriberNum=0
thread4 put message,count=6,subscriberNum=0
thread4 put message,count=7,subscriberNum=0
thread4 put message,count=8,subscriberNum=0
thread4 put message,count=9,subscriberNum=0
thread5 put message,count=10,subscriberNum=0
thread5 put message,count=11,subscriberNum=0
thread5 put message,count=12,subscriberNum=0
thread5 put message,count=13,subscriberNum=0
thread5 put message,count=14,subscriberNum=0
thread2 put message,count=15,subscriberNum=0
thread2 put message,count=16,subscriberNum=0
thread2 put message,count=17,subscriberNum=0
thread2 put message,count=18,subscriberNum=0
thread2 put message,count=19,subscriberNum=0
thread3 put message,count=20,subscriberNum=0
thread3 put message,count=21,subscriberNum=0
thread3 put message,count=22,subscriberNum=0
thread3 put message,count=23,subscriberNum=0
thread3 put message,count=24,subscriberNum=0
2.消费者
(1)subscribe实现订阅消费消息(开启两个线程订阅消息)
import redis.clients.jedis.Jedis; import redis.clients.jedis.JedisPubSub; /** * @Author: qlq * @Description * @Date: 22:34 2018/10/9 */ public class MessageConsumer implements Runnable { public static final String CHANNEL_KEY = "channel:1";//频道 public static final String EXIT_COMMAND = "exit";//结束程序的消息 private MyJedisPubSub myJedisPubSub = new MyJedisPubSub();//处理接收消息 public void consumerMessage() { Jedis jedis = JedisPoolUtils.getJedis(); jedis.subscribe(myJedisPubSub, CHANNEL_KEY);//第一个参数是处理接收消息，第二个参数是订阅的消息频道 } @Override public void run() { while (true) { consumerMessage(); } } public static void main(String[] args) { MessageConsumer messageConsumer = new MessageConsumer(); Thread t1 = new Thread(messageConsumer, "thread5"); Thread t2 = new Thread(messageConsumer, "thread6"); t1.start(); t2.start(); } } /** * 继承JedisPubSub，重写接收消息的方法 */ class MyJedisPubSub extends JedisPubSub { @Override /** JedisPubSub类是一个没有抽象方法的抽象类,里面方法都是一些空实现 * 所以可以选择需要的方法覆盖,这儿使用的是SUBSCRIBE指令，所以覆盖了onMessage * 如果使用PSUBSCRIBE指令，则覆盖onPMessage方法 * 当然也可以选择BinaryJedisPubSub,同样是抽象类，但方法参数为byte[] **/ public void onMessage(String channel, String message) { System.out.println(Thread.currentThread().getName()+"-接收到消息:channel=" + channel + ",message=" + message); //接收到exit消息后退出 if (MessageConsumer.EXIT_COMMAND.equals(message)) { System.exit(0); } } }
我们再次启动生产者生产消息,生产者控制台:
thread5 put message,count=0,subscriberNum=2 thread5 put message,count=1,subscriberNum=2 thread5 put message,count=2,subscriberNum=2 thread5 put message,count=3,subscriberNum=2 thread5 put message,count=4,subscriberNum=2 thread3 put message,count=5,subscriberNum=2 thread3 put message,count=6,subscriberNum=2 thread3 put message,count=7,subscriberNum=2 thread3 put message,count=8,subscriberNum=2 thread3 put message,count=9,subscriberNum=2 thread2 put message,count=10,subscriberNum=2 thread2 put message,count=11,subscriberNum=2 thread2 put message,count=12,subscriberNum=2 thread2 put message,count=13,subscriberNum=2 thread2 put message,count=14,subscriberNum=2 thread4 put message,count=15,subscriberNum=2 thread4 put message,count=16,subscriberNum=2 thread4 put message,count=17,subscriberNum=2 thread4 put message,count=18,subscriberNum=2 thread4 put message,count=19,subscriberNum=2 thread1 put message,count=20,subscriberNum=2 thread1 put message,count=21,subscriberNum=2 thread1 put message,count=22,subscriberNum=2 thread1 put message,count=23,subscriberNum=2 thread1 put message,count=24,subscriberNum=2 Process finished with exit code 0
消费者控制台:
thread6-接收到消息:channel=channel:1,message=message0 thread5-接收到消息:channel=channel:1,message=message0 thread5-接收到消息:channel=channel:1,message=message1 thread6-接收到消息:channel=channel:1,message=message1 thread5-接收到消息:channel=channel:1,message=message2 thread6-接收到消息:channel=channel:1,message=message2 thread5-接收到消息:channel=channel:1,message=message3 thread6-接收到消息:channel=channel:1,message=message3 thread5-接收到消息:channel=channel:1,message=message4 thread6-接收到消息:channel=channel:1,message=message4 thread5-接收到消息:channel=channel:1,message=message5 thread6-接收到消息:channel=channel:1,message=message5 thread5-接收到消息:channel=channel:1,message=message6 thread6-接收到消息:channel=channel:1,message=message6 thread5-接收到消息:channel=channel:1,message=message7 thread6-接收到消息:channel=channel:1,message=message7 thread5-接收到消息:channel=channel:1,message=message8 thread6-接收到消息:channel=channel:1,message=message8 thread5-接收到消息:channel=channel:1,message=message9 thread6-接收到消息:channel=channel:1,message=message9 thread5-接收到消息:channel=channel:1,message=message10 thread6-接收到消息:channel=channel:1,message=message10 thread5-接收到消息:channel=channel:1,message=message11 thread6-接收到消息:channel=channel:1,message=message11 thread5-接收到消息:channel=channel:1,message=message12 thread6-接收到消息:channel=channel:1,message=message12 thread5-接收到消息:channel=channel:1,message=message13 thread6-接收到消息:channel=channel:1,message=message13 thread5-接收到消息:channel=channel:1,message=message14 thread6-接收到消息:channel=channel:1,message=message14 thread5-接收到消息:channel=channel:1,message=message15 thread6-接收到消息:channel=channel:1,message=message15 thread5-接收到消息:channel=channel:1,message=message16 thread6-接收到消息:channel=channel:1,message=message16 thread5-接收到消息:channel=channel:1,message=message17 thread6-接收到消息:channel=channel:1,message=message17 thread5-接收到消息:channel=channel:1,message=message18 thread6-接收到消息:channel=channel:1,message=message18 thread5-接收到消息:channel=channel:1,message=message19 thread6-接收到消息:channel=channel:1,message=message19 thread5-接收到消息:channel=channel:1,message=message20 thread6-接收到消息:channel=channel:1,message=message20 thread5-接收到消息:channel=channel:1,message=message21 thread6-接收到消息:channel=channel:1,message=message21 thread5-接收到消息:channel=channel:1,message=message22 thread6-接收到消息:channel=channel:1,message=message22 thread5-接收到消息:channel=channel:1,message=message23 thread6-接收到消息:channel=channel:1,message=message23 thread5-接收到消息:channel=channel:1,message=message24 thread6-接收到消息:channel=channel:1,message=message24
(2)psubscribe实现订阅消费消息(开启两个线程订阅消息)
import redis.clients.jedis.Jedis; import redis.clients.jedis.JedisPubSub; /** * @Author: qlq * @Description * @Date: 22:34 2018/10/9 */ public class MessageConsumer implements Runnable { public static final String CHANNEL_KEY = "channel*";//频道 public static final String EXIT_COMMAND = "exit";//结束程序的消息 private MyJedisPubSub myJedisPubSub = new MyJedisPubSub();//处理接收消息 public void consumerMessage() { Jedis jedis = JedisPoolUtils.getJedis(); jedis.psubscribe(myJedisPubSub, CHANNEL_KEY);//第一个参数是处理接收消息，第二个参数是订阅的消息频道 } @Override public void run() { while (true) { consumerMessage(); } } public static void main(String[] args) { MessageConsumer messageConsumer = new MessageConsumer(); Thread t1 = new Thread(messageConsumer, "thread5"); Thread t2 = new Thread(messageConsumer, "thread6"); t1.start(); t2.start(); } } /** * 继承JedisPubSub，重写接收消息的方法 */ class MyJedisPubSub extends JedisPubSub { @Override public void onPMessage(String pattern, String channel, String message) { System.out.println(Thread.currentThread().getName()+"-接收到消息:pattern="+pattern+",channel=" + channel + ",message=" + message); //接收到exit消息后退出 if (MessageConsumer.EXIT_COMMAND.equals(message)) { System.exit(0); } } }
重写JedisPubSub 的onPMessage方法即可
启动生产者生产消息之后查看消费者控制台:
thread6-接收到消息:pattern=channel*,channel=channel:1,message=message0 thread5-接收到消息:pattern=channel*,channel=channel:1,message=message0 thread5-接收到消息:pattern=channel*,channel=channel:1,message=message1 thread6-接收到消息:pattern=channel*,channel=channel:1,message=message1 thread6-接收到消息:pattern=channel*,channel=channel:1,message=message2 thread5-接收到消息:pattern=channel*,channel=channel:1,message=message2 thread6-接收到消息:pattern=channel*,channel=channel:1,message=message3 thread5-接收到消息:pattern=channel*,channel=channel:1,message=message3 thread6-接收到消息:pattern=channel*,channel=channel:1,message=message4 thread5-接收到消息:pattern=channel*,channel=channel:1,message=message4 thread6-接收到消息:pattern=channel*,channel=channel:1,message=message5 thread5-接收到消息:pattern=channel*,channel=channel:1,message=message5 thread6-接收到消息:pattern=channel*,channel=channel:1,message=message6 thread5-接收到消息:pattern=channel*,channel=channel:1,message=message6 thread6-接收到消息:pattern=channel*,channel=channel:1,message=message7 thread5-接收到消息:pattern=channel*,channel=channel:1,message=message7 thread6-接收到消息:pattern=channel*,channel=channel:1,message=message8 thread5-接收到消息:pattern=channel*,channel=channel:1,message=message8 thread6-接收到消息:pattern=channel*,channel=channel:1,message=message9 thread5-接收到消息:pattern=channel*,channel=channel:1,message=message9 thread6-接收到消息:pattern=channel*,channel=channel:1,message=message10 thread5-接收到消息:pattern=channel*,channel=channel:1,message=message10 thread6-接收到消息:pattern=channel*,channel=channel:1,message=message11 thread5-接收到消息:pattern=channel*,channel=channel:1,message=message11 thread6-接收到消息:pattern=channel*,channel=channel:1,message=message12 thread5-接收到消息:pattern=channel*,channel=channel:1,message=message12 thread6-接收到消息:pattern=channel*,channel=channel:1,message=message13 thread5-接收到消息:pattern=channel*,channel=channel:1,message=message13 thread6-接收到消息:pattern=channel*,channel=channel:1,message=message14 thread5-接收到消息:pattern=channel*,channel=channel:1,message=message14 thread6-接收到消息:pattern=channel*,channel=channel:1,message=message15 thread5-接收到消息:pattern=channel*,channel=channel:1,message=message15 thread6-接收到消息:pattern=channel*,channel=channel:1,message=message16 thread5-接收到消息:pattern=channel*,channel=channel:1,message=message16 thread6-接收到消息:pattern=channel*,channel=channel:1,message=message17 thread5-接收到消息:pattern=channel*,channel=channel:1,message=message17 thread6-接收到消息:pattern=channel*,channel=channel:1,message=message18 thread5-接收到消息:pattern=channel*,channel=channel:1,message=message18 thread6-接收到消息:pattern=channel*,channel=channel:1,message=message19 thread5-接收到消息:pattern=channel*,channel=channel:1,message=message19 thread6-接收到消息:pattern=channel*,channel=channel:1,message=message20 thread5-接收到消息:pattern=channel*,channel=channel:1,message=message20 thread6-接收到消息:pattern=channel*,channel=channel:1,message=message21 thread5-接收到消息:pattern=channel*,channel=channel:1,message=message21 thread6-接收到消息:pattern=channel*,channel=channel:1,message=message22 thread5-接收到消息:pattern=channel*,channel=channel:1,message=message22 thread6-接收到消息:pattern=channel*,channel=channel:1,message=message23 thread5-接收到消息:pattern=channel*,channel=channel:1,message=message23 thread5-接收到消息:pattern=channel*,channel=channel:1,message=message24 thread6-接收到消息:pattern=channel*,channel=channel:1,message=message24
补充:订阅的时候subscribe()和psubscribe()的第二个参数支持可变参数，也就是可以实现订阅多个频道。
至此实现了两种方式的消息队列:
redis自带的list类型(lpush和rpop或者brpop，rpush和lpop或者blpop)---blpop和brpop是阻塞读取。
"发布/订阅"模式(publish channel message 和 subscribe channel [channel ...] 或者 psubscribe pattern [pattern ...] 通配符订阅多个频道)
补充:
1.发布订阅执行订阅之后该线程处于阻塞状态，线程不会终止，如果终止线程需要退订，需要调用JedisPubSub的unsubscribe()方法
例如:
package plainTest; import cn.xm.redisChat.util.JedisPoolUtils; import redis.clients.jedis.Jedis; import redis.clients.jedis.JedisPubSub; /** * @Author: qlq * @Description * @Date: 23:36 2018/10/13 */ public class Test111 { public static void main(String[] args) { Jedis jedis = JedisPoolUtils.getJedis(); System.out.println("订阅前"); jedis.subscribe(new JedisPubSub() { @Override public void onMessage(String channel, String message) { super.onMessage(channel, message); } }, "c1"); System.out.println("订阅后"); } }
结果只会打印订阅前，而且线程不会终止。
为了使线程可以停止，必须退订，而且退订只能调用 JedisPubSub.unsubscribe()方法，例如:收到quit消息之后会退订，线程会回到主线程打印订阅后。
package plainTest; import cn.xm.redisChat.util.JedisPoolUtils; import redis.clients.jedis.Jedis; import redis.clients.jedis.JedisPubSub; /** * @Author: qlq * @Description * @Date: 23:36 2018/10/13 */ public class Test111 { public static void main(String[] args) { Jedis jedis = JedisPoolUtils.getJedis(); System.out.println("订阅前"); jedis.subscribe(new JedisPubSub() { @Override public void onMessage(String channel, String message) { if("quit".equals(message)){ unsubscribe("c1"); } System.out.println(message); } @Override public void unsubscribe(String... channels) { super.unsubscribe(channels); } }, "c1"); System.out.println("订阅后"); } }
2.BRPOP:当给定列表内没有任何元素可供弹出的时候，连接将被BRPOP命令阻塞，直到等待超时或发现可弹出元素为止。(每次只弹出一个元素，当没有元素的时候处于阻塞，当弹出一个元素之后就会解除阻塞)
package plainTest; import cn.xm.redisChat.util.JedisPoolUtils; import redis.clients.jedis.Jedis; import java.util.List; /** * @Author: qlq * @Description * @Date: 23:36 2018/10/13 */ public class Test111 { public static void main(String[] args) { Jedis jedis = JedisPoolUtils.getJedis(); System.out.println("brpop之前"); List<String> messages = jedis.brpop(0,"list1"); System.out.println(messages); System.out.println("brpop之后"); } }
没有元素的时候只会打印brpop之前。
三、基于Springboot封装上面pub/sub 模式的消息队列
这里需要注意避免消息的重复消费，每个实例都会监听到消息。 也就是上面的pub/sub 模式只要是sub 就会收到消息。
1. RedisTemplate
@Configuration public class RedisConfig { @Bean public RedisTemplate<String, Object> redisTemplate(RedisConnectionFactory factory) { RedisTemplate<String, Object> template = new RedisTemplate<>(); template.setConnectionFactory(factory); StringRedisSerializer stringRedisSerializer = new StringRedisSerializer(); GenericJackson2JsonRedisSerializer genericJackson2JsonRedisSerializer = new GenericJackson2JsonRedisSerializer(); template.setKeySerializer(stringRedisSerializer); template.setHashKeySerializer(stringRedisSerializer); template.setValueSerializer(genericJackson2JsonRedisSerializer); template.setHashValueSerializer(genericJackson2JsonRedisSerializer); template.afterPropertiesSet(); return template; } }
用json 存对象， 发送到上面的channel。
2. 消费者:最后发现是使用Spring 新起的线程进行处理，不占用业务线程
@Slf4j public class RedisMqConsumer implements MessageListener { @Autowired private RedisTemplate<String, Object> redisTemplate; @Override public void onMessage(Message message, byte[] pattern) { RedisSerializer<?> valueSerializer = redisTemplate.getValueSerializer(); XXXDto pushMqMessageDTO = (XXXDto) valueSerializer.deserialize(message.getBody()); } }
3. 配置适配器和container 容器
@Configuration public class RedisMqConfiguration { /** * 监听器 * * @return */ @Bean public RedisMqConsumer redisConsumer() { return new RedisMqConsumer(); } /** * 消息适配器 */ @Bean public MessageListenerAdapter messageListenerAdapter(ProjectRemindRedisMqConsumer redisConsumer) { return new MessageListenerAdapter(redisConsumer); } /** * 创建消息监听容器 * * @param redisConnectionFactory * @param messageListenerAdapter * @return */ @Bean public RedisMessageListenerContainer redisMessageListenerContainer(RedisConnectionFactory redisConnectionFactory, MessageListenerAdapter messageListenerAdapter) { RedisMessageListenerContainer redisMessageListenerContainer = new RedisMessageListenerContainer(); redisMessageListenerContainer.setConnectionFactory(redisConnectionFactory); redisMessageListenerContainer.addMessageListener(messageListenerAdapter, new PatternTopic(IConstant.RedisMq.CHANNEL_NAME)); return redisMessageListenerContainer; } }
4. 测试生产者代码
@RestController @RequestMapping("/msg") @Slf4j public class RedisMQProducer { @Autowired private RedisTemplate<String, Object> redisTemplate; @GetMapping("/send") public void send(@RequestBody XXXDto pushMqMessageDTO) { String messageId = "you message id"; // 检测是否重复发送，用sennx 进行加锁检测 Boolean success = redisTemplate.opsForValue().setIfAbsent(messageId, 1, 20, TimeUnit.SECONDS); if (BooleanUtils.isTrue(success)) { redisTemplate.convertAndSend(IConstant.RedisMq.CHANNEL_NAME, pushMqMessageDTO); } log.info("send failed"); } }
补充：redis 也可以实现延迟队列
延时队可以通过redis的zset(有序列表)来实现。思路是将消息序列化成一个字符串作为zset的value，这个消息的到期时间作为score 处理 然后用多个线程轮询zset 获取到期的任务进行处理。
比如： 一个消息预计在11585 时间戳被消费，则其分数设置为该时间戳即可。
【当你用心写完每一篇博客之后,你会发现它比你用代码实现功能更有成就感!】
