本章是《Docker下RabbitMQ四部曲》系列的终篇，今天的我们一起来体验Rabbit’MQ集群的高可用能力，看看RabbitMQ集群中的部分节点宕机时，是否还能生产和消费消息；
前文链接
《Docker下RabbitMQ四部曲》的前三篇链接如下：
- 《Docker下RabbitMQ四部曲之一：极速体验(单机和集群)》；
- 《Docker下RabbitMQ四部曲之二：细说RabbitMQ镜像制作》；
- 《Docker下RabbitMQ四部曲之三：细说java开发》；
实战概要
今天实战的步骤如下：
- 制作docker-compose.yml文件，为每个容器配置好参数；
- 启动所有容器，包括RabbitMQ集群、消息生产者的web应用、消息消费者的web应用；
- 逐个停止集群中的RabbitMQ容器，每停止一个，就验证一次消息的生产和消费；
- 逐个恢复集群中的RabbitMQ容器，每恢复一个，就验证一次消息的生产和消费；
制作docker-compose.yml文件
本次实战会创建6个容器，整理如下：
容器名 | IP地址 | 作用 |
|---|---|---|
hacluster_rabbit1_1 | 172.19.0.2 | RabbitMQ主节点 |
hacluster_rabbit2_1 | 172.19.0.3 | RabbitMQ从节点，内存节点 |
hacluster_rabbit3_1 | 172.19.0.4 | RabbitMQ从节点 |
hacluster_producer_1 | 172.19.0.5 | web应用，负责生产消息 |
hacluster_consumer1_1 | 172.19.0.6 | web应用，负责消费消息 |
hacluster_consumer2_1 | 172.19.0.7 | web应用，负责消费消息 |
前面章节的实战中，我们也创建了上述六个容器，今天依然是六个，并且身份角色不变，变化的地方主要是以下三点：
- 负责生产消息的hacluster_producer_1容器，前面章节只连接了一个RabbitMQ容器，本章会连接三个；
- 负责消费消息的hacluster_consumer1_1，前面章节只连接了一个RabbitMQ容器，本章会连接三个；
- 负责消费消息的hacluster_consumer2_1，前面章节只连接了一个RabbitMQ容器，本章会连接三个；
基于以上总结，我们写出的docker-compose.yml文件内容如下：
代码语言：javascript复制version: '2'
services:
rabbit1:
image: bolingcavalry/rabbitmq-server:0.0.3
hostname: rabbit1
ports:
- "15672:15672"
environment:
- RABBITMQ_DEFAULT_USER=admin
- RABBITMQ_DEFAULT_PASS=888888
rabbit2:
image: bolingcavalry/rabbitmq-server:0.0.3
hostname: rabbit2
depends_on:
- rabbit1
links:
- rabbit1
environment:
- CLUSTERED=true
- CLUSTER_WITH=rabbit1
- RAM_NODE=true
- HA_ENABLE=true
ports:
- "15673:15672"
rabbit3:
image: bolingcavalry/rabbitmq-server:0.0.3
hostname: rabbit3
depends_on:
- rabbit2
links:
- rabbit1
- rabbit2
environment:
- CLUSTERED=true
- CLUSTER_WITH=rabbit1
ports:
- "15675:15672"
producer:
image: bolingcavalry/rabbitmqproducer:0.0.2-SNAPSHOT
hostname: producer
depends_on:
- rabbit3
links:
- rabbit1:rabbitmqhost1
- rabbit2:rabbitmqhost2
- rabbit3:rabbitmqhost3
ports:
- "18080:8080"
environment:
- mq.rabbit.address=rabbitmqhost1:5672,rabbitmqhost2:5672,rabbitmqhost3:5672
- mq.rabbit.username=admin
- mq.rabbit.password=888888
consumer1:
image: bolingcavalry/rabbitmqconsumer:0.0.5-SNAPSHOT
hostname: consumer1
depends_on:
- producer
links:
- rabbit1:rabbitmqhost1
- rabbit2:rabbitmqhost2
- rabbit3:rabbitmqhost3
environment:
- mq.rabbit.address=rabbitmqhost1:5672,rabbitmqhost2:5672,rabbitmqhost3:5672
- mq.rabbit.username=admin
- mq.rabbit.password=888888
- mq.rabbit.queue.name=consumer1.queue
consumer2:
image: bolingcavalry/rabbitmqconsumer:0.0.5-SNAPSHOT
hostname: consumer2
depends_on:
- consumer1
links:
- rabbit1:rabbitmqhost1
- rabbit2:rabbitmqhost2
- rabbit3:rabbitmqhost3
environment:
- mq.rabbit.address=rabbitmqhost1:5672,rabbitmqhost2:5672,rabbitmqhost3:5672
- mq.rabbit.username=admin
- mq.rabbit.password=888888
- mq.rabbit.queue.name=consumer2.queue以上的docker-compose.yml文件，有以下两点需要注意：
- rabbit2：增加了一个环境变量HA_ENABLE=true，《Docker下RabbitMQ四部曲之二：细说RabbitMQ镜像制作》一文中分析镜像制作的时候曾提到过，容器创建时startrabbit.sh脚本中会检查这个环境变量，如果为true，就会执行命令：rabbitmqctl set_policy HA ‘^(?!amq.).*’ ‘{“ha-mode”: “all”}’，该命令会将队列设置为镜像模式，在三个Rabbit MQ之间同步；
- producer、consumer1、consumer3这三个容器的环境变量mq.rabbit.address，都设置成了三个RabbitMQ容器的地址加端口：rabbitmqhost1:5672,rabbitmqhost2:5672,rabbitmqhost3:5672；
启动所有容器
在刚刚创建的docker-compose.yml文件所在目录下执行命令docker-compose up -d，即可创建所有容器，创建完成后执行以下操作来确认是否启动成功：
- 例如我的电脑IP地址是192.168.119.155，那么在浏览器输入地址：192.168.119.155:15672即可访问RabbitMQ的管理页面，用户名：admin，密码：888888，如下图：
- 点击"Exchanges"这个Tab页，如下图，看到交换机创建成功，HA模式：
- 点击"Queues"这个Tab页，如下图，看到两个队列创建成功，HA模式：
- 在浏览器输入：http://192.168.119.155:18080/send/aaa/bbb，即可控制容器hacluster_producer_1 生产一条消息，如下图：
- 在控制台输入docker logs -f hacluster_consumer1_1，即可看到hacluster_consumer1_1 消费消息的日志，如下：
2018-05-19 11:21:44.217 INFO 1 --- [ main] c.b.r.RabbitmqconsumerApplication : Started RabbitmqconsumerApplication in 29.099 seconds (JVM running for 39.398)
2018-05-19 11:36:21.332 INFO 1 --- [cTaskExecutor-1] c.b.r.receiver.FanoutReceiver : receive message : hello, aaa, bbb, 2018-05-19 11:36:21看来整个RabbitMQ集群的生产和消费是没有问题的，接下来通过停止容器的方式来模拟生产环境的宕机；
逐个停止集群中的RabbitMQ容器
- 先停hacluster_rabbit1_1 ，执行命令docker stop hacluster_rabbit1_1，如下：
root@maven:~# docker stop hacluster_rabbit1_1
hacluster_rabbit1_1- 去管理页面看看，由于hacluster_rabbit1_1容器已经停止了，所以我们要访问hacluster_rabbit2_1容器提供的web页面：http://192.168.119.155:15673，如下图红框，可以看见页面提示节点故障：
- 交换机和队列的页面并无异常；
- 在浏览器输入：http://192.168.119.155:18080/send/aaa/bbb，尝试发送一条消息，响应时间明显变长，但是依然会返回操作成功；
- 在控制台输入docker logs -f hacluster_producer_1，查看生产消息的web容器的日志，如下：
2018-05-19 11:43:22.681 WARN 1 --- [172.19.0.2:5672] c.r.c.impl.ForgivingExceptionHandler : An unexpected connection driver error occured (Exception message: Connection reset)
2018-05-19 11:43:22.703 ERROR 1 --- [172.19.0.2:5672] o.s.a.r.c.CachingConnectionFactory : Channel shutdown: connection error
2018-05-19 11:53:31.836 INFO 1 --- [io-8080-exec-10] o.s.a.r.c.CachingConnectionFactory : Attempting to connect to: [rabbitmqhost1:5672, rabbitmqhost2:5672, rabbitmqhost3:5672]
2018-05-19 11:53:46.878 INFO 1 --- [io-8080-exec-10] o.s.a.r.c.CachingConnectionFactory : Created new connection: connectionFactory#4ae3c1cd:1/SimpleConnection@44028da7 [delegate=amqp://admin@172.19.0.3:5672/, localPort= 37818]从日志中可以清晰的看到，停止hacluster_rabbit1_1容器是，消息生产者会立即报异常，但是不会自动重连，等到发送消息的时候，才会连接到新的RabbitMQ，这次连接的是hacluster_rabbit2_1 ； 6. 在控制台输入docker logs -f hacluster_consumer1_1，查看消费消息的web容器的日志，如下：
代码语言：javascript复制2018-05-19 11:38:14.945 INFO 1 --- [cTaskExecutor-1] c.b.r.receiver.FanoutReceiver : receive message : hello, aaa, bbb, 2018-05-19 11:38:14
2018-05-19 11:43:22.672 WARN 1 --- [172.19.0.2:5672] c.r.c.impl.ForgivingExceptionHandler : An unexpected connection driver error occured (Exception message: Connection reset)
2018-05-19 11:43:22.726 ERROR 1 --- [172.19.0.2:5672] o.s.a.r.c.CachingConnectionFactory : Channel shutdown: connection error
2018-05-19 11:43:23.163 INFO 1 --- [cTaskExecutor-1] o.s.a.r.l.SimpleMessageListenerContainer : Restarting Consumer@9f116cc: tags=[{amq.ctag-0csUBn5OQiTGEphcGI2p3A=consumer1.queue}], channel=Cached Rabbit Channel: PublisherCallbackChannelImpl: AMQChannel(amqp://admin@172.19.0.2:5672/,1), conn: Proxy@42dd311 Shared Rabbit Connection: SimpleConnection@5a1a52da [delegate=amqp://admin@172.19.0.2:5672/, localPort= 34240], acknowledgeMode=AUTO local queue size=0
2018-05-19 11:43:23.181 INFO 1 --- [cTaskExecutor-2] o.s.a.r.c.CachingConnectionFactory : Attempting to connect to: [rabbitmqhost1:5672, rabbitmqhost2:5672, rabbitmqhost3:5672]
2018-05-19 11:43:29.042 INFO 1 --- [cTaskExecutor-2] o.s.a.r.c.CachingConnectionFactory : Created new connection: connectionFactory#45f45fa1:1/SimpleConnection@2b9e231d [delegate=amqp://admin@172.19.0.3:5672/, localPort= 49624]
2018-05-19 11:53:46.899 INFO 1 --- [cTaskExecutor-2] c.b.r.receiver.FanoutReceiver : receive message : hello, aaa, bbb, 2018-05-19 11:53:31从日志上可以看出：RabbitMQ宕机的时候，消费者会立即重连到集群中的其他机器；(日志关键字：Created new connection) 7. 停掉RabbitMQ集群中的第二个容器，执行命令docker stop hacluster_rabbit2_1； 8. 访问管理页面的时候，要输入容器hacluster_rabbit3_1的地址：http://192.168.119.155:15675，基本情况如下图，两个节点的问题都能看到：
9. 在浏览器输入：http://192.168.119.155:18080/send/aaa/bbb，尝试发送一条消息，响应时间再次明显变长，但是依然会返回操作成功； 10. 在控制台输入docker logs -f hacluster_producer_1，查看生产消息的web容器的日志，如下，提示重连成功，这次连接到了容器hacluster_rabbit3_1 ：
代码语言：javascript复制2018-05-19 12:07:45.322 WARN 1 --- [172.19.0.3:5672] c.r.c.impl.ForgivingExceptionHandler : An unexpected connection driver error occured (Exception message: Connection reset)
2018-05-19 12:07:45.334 ERROR 1 --- [172.19.0.3:5672] o.s.a.r.c.CachingConnectionFactory : Channel shutdown: connection error
2018-05-19 12:07:45.336 ERROR 1 --- [172.19.0.3:5672] o.s.a.r.c.CachingConnectionFactory : Channel shutdown: connection error
2018-05-19 12:12:06.404 INFO 1 --- [nio-8080-exec-4] o.s.a.r.c.CachingConnectionFactory : Attempting to connect to: [rabbitmqhost1:5672, rabbitmqhost2:5672, rabbitmqhost3:5672]
2018-05-19 12:12:41.467 INFO 1 --- [nio-8080-exec-4] o.s.a.r.c.CachingConnectionFactory : Created new connection: connectionFactory#4ae3c1cd:2/SimpleConnection@6d23e50 [delegate=amqp://admin@172.19.0.4:5672/, localPort= 54310]- 在控制台输入docker logs -f hacluster_consumer1_1，查看消费消息的web容器的日志，如下：
2018-05-19 12:07:45.327 WARN 1 --- [172.19.0.3:5672] c.r.c.impl.ForgivingExceptionHandler : An unexpected connection driver error occured (Exception message: Connection reset)
2018-05-19 12:07:45.346 ERROR 1 --- [172.19.0.3:5672] o.s.a.r.c.CachingConnectionFactory : Channel shutdown: connection error
2018-05-19 12:07:45.348 ERROR 1 --- [172.19.0.3:5672] o.s.a.r.c.CachingConnectionFactory : Channel shutdown: connection error
2018-05-19 12:07:45.427 INFO 1 --- [cTaskExecutor-2] o.s.a.r.l.SimpleMessageListenerContainer : Restarting Consumer@317c5a8a: tags=[{amq.ctag-ZKT8Q4gcU9v7bA-lNOUEFQ=consumer1.queue}], channel=Cached Rabbit Channel: PublisherCallbackChannelImpl: AMQChannel(amqp://admin@172.19.0.3:5672/,1), conn: Proxy@42dd311 Shared Rabbit Connection: SimpleConnection@2b9e231d [delegate=amqp://admin@172.19.0.3:5672/, localPort= 49624], acknowledgeMode=AUTO local queue size=0
2018-05-19 12:07:45.432 INFO 1 --- [cTaskExecutor-3] o.s.a.r.c.CachingConnectionFactory : Attempting to connect to: [rabbitmqhost1:5672, rabbitmqhost2:5672, rabbitmqhost3:5672]
2018-05-19 12:08:07.352 INFO 1 --- [cTaskExecutor-3] o.s.a.r.c.CachingConnectionFactory : Created new connection: connectionFactory#45f45fa1:2/SimpleConnection@71dadbb0 [delegate=amqp://admin@172.19.0.4:5672/, localPort= 34416]
2018-05-19 12:12:56.869 INFO 1 --- [cTaskExecutor-3] c.b.r.receiver.FanoutReceiver : receive message : hello, aaa, bbb, 2018-05-19 12:12:06日志显示，也是连到了容器hacluster_rabbit3_1 ，并且消费消息成功； 12. 停掉RabbitMQ集群中的第三个容器（也是最后一个），执行命令docker stop hacluster_rabbit3_1； 13. 这次没有管理页面看了… 14. 在浏览器输入：http://192.168.119.155:18080/send/aaa/bbb，尝试发送一条消息，长时间等待后，页面提示错误如下图：
15. 查看容器hacluster_producer_1的日志，如下：
代码语言：javascript复制2018-05-19 12:18:27.812 WARN 1 --- [172.19.0.4:5672] c.r.c.impl.ForgivingExceptionHandler : An unexpected connection driver error occured (Exception message: Connection reset)
2018-05-19 12:18:27.813 ERROR 1 --- [172.19.0.4:5672] o.s.a.r.c.CachingConnectionFactory : Channel shutdown: connection error
2018-05-19 12:18:27.813 ERROR 1 --- [172.19.0.4:5672] o.s.a.r.c.CachingConnectionFactory : Channel shutdown: connection error
2018-05-19 12:18:55.836 INFO 1 --- [nio-8080-exec-7] o.s.a.r.c.CachingConnectionFactory : Attempting to connect to: [rabbitmqhost1:5672, rabbitmqhost2:5672, rabbitmqhost3:5672]
2018-05-19 12:19:50.921 ERROR 1 --- [nio-8080-exec-7] o.a.c.c.C.[.[.[/].[dispatcherServlet] : Servlet.service() for servlet [dispatcherServlet] in context with path [] threw exception [Request processing failed; nested exception is org.springframework.amqp.AmqpIOException: java.net.UnknownHostException: rabbitmqhost3] with root cause
java.net.UnknownHostException: rabbitmqhost3
at java.net.AbstractPlainSocketImpl.connect(AbstractPlainSocketImpl.java:184) ~[na:1.8.0_111]
at java.net.SocksSocketImpl.connect(SocksSocketImpl.java:392) ~[na:1.8.0_111]
at java.net.Socket.connect(Socket.java:589) ~[na:1.8.0_111]
at com.rabbitmq.client.impl.SocketFrameHandlerFactory.create(SocketFrameHandlerFactory.java:60) ~[amqp-client-5.1.2.jar!/:5.1.2]
at com.rabbitmq.client.ConnectionFactory.newConnection(ConnectionFactory.java:955) ~[amqp-client-5.1.2.jar!/:5.1.2]
at com.rabbitmq.client.ConnectionFactory.newConnection(ConnectionFactory.java:907) ~[amqp-client-5.1.2.jar!/:5.1.2]
at com.rabbitmq.client.ConnectionFactory.newConnection(ConnectionFactory.java:847) ~[amqp-client-5.1.2.jar!/:5.1.2]
at org.springframework.amqp.rabbit.connection.AbstractConnectionFactory.createBareConnection(AbstractConnectionFactory.java:449) ~[spring-rabbit-2.0.3.RELEASE.jar!/:2.0.3.RELEASE]
at org.springframework.amqp.rabbit.connection.CachingConnectionFactory.createConnection(CachingConnectionFactory.java:614) ~[spring-rabbit-2.0.3.RELEASE.jar!/:2.0.3.RELEASE]
at org.springframework.amqp.rabbit.connection.CachingConnectionFactory.createBareChannel(CachingConnectionFactory.java:564) ~[spring-rabbit-2.0.3.RELEASE.jar!/:2.0.3.RELEASE]
at org.springframework.amqp.rabbit.connection.CachingConnectionFactory.getCachedChannelProxy(CachingConnectionFactory.java:538) ~[spring-rabbit-2.0.3.RELEASE.jar!/:2.0.3.RELEASE]
at org.springframework.amqp.rabbit.connection.CachingConnectionFactory.getChannel(CachingConnectionFactory.java:520) ~[spring-rabbit-2.0.3.RELEASE.jar!/:2.0.3.RELEASE]
......
...如上所示，异常信息为连接RabbitMQ服务器失败； 16. 查看容器hacluster_consumer1_1的日志，如下：
代码语言：javascript复制2018-05-19 12:18:27.815 WARN 1 --- [172.19.0.4:5672] c.r.c.impl.ForgivingExceptionHandler : An unexpected connection driver error occured (Exception message: Connection reset)
2018-05-19 12:18:27.816 ERROR 1 --- [172.19.0.4:5672] o.s.a.r.c.CachingConnectionFactory : Channel shutdown: connection error
2018-05-19 12:18:27.816 ERROR 1 --- [172.19.0.4:5672] o.s.a.r.c.CachingConnectionFactory : Channel shutdown: connection error
2018-05-19 12:18:28.100 INFO 1 --- [cTaskExecutor-3] o.s.a.r.l.SimpleMessageListenerContainer : Restarting Consumer@5b0307b0: tags=[{amq.ctag-0UhQ6jE-D5Wl2ZPl4EWhDQ=consumer1.queue}], channel=Cached Rabbit Channel: PublisherCallbackChannelImpl: AMQChannel(amqp://admin@172.19.0.4:5672/,1), conn: Proxy@42dd311 Shared Rabbit Connection: SimpleConnection@71dadbb0 [delegate=amqp://admin@172.19.0.4:5672/, localPort= 34416], acknowledgeMode=AUTO local queue size=0
2018-05-19 12:18:28.104 INFO 1 --- [cTaskExecutor-4] o.s.a.r.c.CachingConnectionFactory : Attempting to connect to: [rabbitmqhost1:5672, rabbitmqhost2:5672, rabbitmqhost3:5672]
2018-05-19 12:19:23.178 ERROR 1 --- [cTaskExecutor-4] o.s.a.r.l.SimpleMessageListenerContainer : Failed to check/redeclare auto-delete queue(s).
org.springframework.amqp.AmqpIOException: java.net.UnknownHostException: rabbitmqhost3
at org.springframework.amqp.rabbit.support.RabbitExceptionTranslator.convertRabbitAccessException(RabbitExceptionTranslator.java:71) ~[spring-rabbit-2.0.3.RELEASE.jar!/:2.0.3.RELEASE]
at org.springframework.amqp.rabbit.connection.AbstractConnectionFactory.createBareConnection(AbstractConnectionFactory.java:476) ~[spring-rabbit-2.0.3.RELEASE.jar!/:2.0.3.RELEASE]
at org.springframework.amqp.rabbit.connection.CachingConnectionFactory.createConnection(CachingConnectionFactory.java:614) ~[spring-rabbit-2.0.3.RELEASE.jar!/:2.0.3.RELEASE]如上所示，也是连接失败，并且，日志的最后会发现应用在自动尝试重新连接RabbitMQ；
至此，RabbitMQ集群宕机模拟就完成了，结果说明在HA模式下，只要还有可用的节点，应用就会尝试连接，如果连接成功，消息的消费是不受影响的；
目前RabbitMQ集群的所有容器都停掉了，接下来我们逐个恢复刚才停下来的容器，看看服务能否恢复；
逐个恢复集群中的RabbitMQ容器
- 先恢复hacluster_rabbit1_1，执行命令docker start hacluster_rabbit1_1；
- 执行命令docker logs -f hacluster_rabbit1_1，查看容器日志，发现一直停留在下面的位置，不再更新：
- 浏览器访问管理页面：http://192.168.119.155:15672，结果页面无法打开；
- 在浏览器输入：http://192.168.119.155:18080/send/aaa/bbb，尝试发送一条消息，会显示报错页面；
- 去看生产者和消费者容器的日志，发现都是连接RabbitMQ失败的错误；
- 从上述现象可以发现：恢复集群服务时，只恢复一台机器是不够的；
- 再恢复hacluster_rabbit2_1，执行命令docker start hacluster_rabbit12_1；
- 管理页面依旧不可用，发送消息失败，并且生产者和消费者容器都无法连接到RabbitMQ容器上去；
- 再恢复hacluster_rabbit3_1，执行命令docker start hacluster_rabbit3_1，这样整个集群的所有容器都恢复了；
- hacluster_rabbit1_1的日志发生了变化，如下：
- hacluster_rabbit2_1的日志也更新了，如下：
- 管理页面可以正常打开，并且显示三个节点都正常了：
- 在浏览器输入：http://192.168.119.155:18080/send/aaa/bbb，尝试发送一条消息，消息的生产和消费都正常了；
至此，RabbitMQ的高可用实战已经完成了，从宕机到恢复我们都试了一遍，对RabbitMQ集群也有了更多直观的了解；
文章写到这里，《Docker下RabbitMQ四部曲》就全部结束了，希望这个系列能够对您在学习RabbitMQ过程中有所帮助，在docker实战中，也期待能对您自己定制的RabbitMQ镜像提供一些参考；
