20160914 镜像到内存节点测试

目的

验证在内存节点上做镜像，消息是否会丢失

环境

3 台服务器（IP： 64,65,66）配置集群，没有任何镜像策略。

过程

1 查看集群环境，如果没有内存节点，使用 change_cluster_node_type 命令改变其中一个节点为内存节点

$ rabbitmqctl cluster_status
  
[{nodes,[{disc,['rabbit@TK-RABBITMQ1','rabbit@TK-RABBITMQ2']},
         {ram,['rabbit@TK-RABBITMQ3']}]},
 {running_nodes,['rabbit@TK-RABBITMQ2','rabbit@TK-RABBITMQ3',
                 'rabbit@TK-RABBITMQ1']},
 {cluster_name,<<"rabbit@TK-RABBITMQ1">>},
 {partitions,[]},
 {alarms,[{'rabbit@TK-RABBITMQ2',[]},
          {'rabbit@TK-RABBITMQ3',[]},
          {'rabbit@TK-RABBITMQ1',[]}]}]

2 新建一个 node 方式，名字为 test_policy ，内容为镜像到节点TK-RABBITMQ3 的镜像

$ rabbitmqctl set_policy test_policy "^test\."  '{"ha-mode":"nodes","ha-params":["rabbit@TK-RABBITMQ3"]}'
  
Setting policy "test_policy" for pattern "^test\\." to "{\"ha-mode\":\"nodes\",\"ha-params\":[\"rabbit@TK-RABBITMQ3\"]}" with priority "0" ...

3 使用客户端代码新建name为test.one Node为rabbit@TK-RABBITMQ1的队列
   使用客户端代码新建name为test.two Node为rabbit@TK-RABBITMQ2的队列
   使用客户端代码新建name为test.three Node为rabbit@TK-RABBITMQ3的队列

4 查看队列信息

$ rabbitmqctl list_queues name policy messages_ready  pid slave_pids
  
test.one    test_policy 10  <rabbit@TK-RABBITMQ3.3.27052.123> []
test.three  test_policy 10  <rabbit@TK-RABBITMQ3.3.27041.123> []
test.two    test_policy 10  <rabbit@TK-RABBITMQ3.3.27038.123> []

由于镜像是在 TK-RABBITMQ3 上，虽然建队列时是指定在其它节点上，但是主队列都在此节点上。

5 使用 rabbitmqctl stop_app 关闭节点 TK-RABBITMQ3 应用，再用 rabbitmqctl start_app 启动应用

6 查看队列信息

$ rabbitmqctl list_queues name policy messages_ready  pid slave_pids
  
test.one    test_policy 10  <rabbit@TK-RABBITMQ3.3.27052.123> []
test.three  test_policy 10  <rabbit@TK-RABBITMQ3.3.27041.123> []
test.two    test_policy 10  <rabbit@TK-RABBITMQ3.3.27038.123> []

7 关闭节点 TK-RABBITMQ1 和 TK-RABBITMQ2

8 使用 rabbitmqctl stop_app关闭节点 TK-RABBITMQ3

9 启动节点TK-RABBITMQ3

$ rabbitmqctl start_app
  
Starting node 'rabbit@TK-RABBITMQ3' ...
 
BOOT FAILED
===========
Error description:
   {could_not_start,rabbit,
       {{failed_to_cluster_with,
            ['rabbit@TK-RABBITMQ1','rabbit@TK-RABBITMQ2'],
            "Mnesia could not connect to any nodes."},
        {rabbit,start,[normal,[]]}}}
Log files (may contain more information):
   /home/rabbitmq/rmq/rabbitmq_server-3.6.2/var/log/rabbitmq/rabbit@TK-RABBITMQ3.log
   /home/rabbitmq/rmq/rabbitmq_server-3.6.2/var/log/rabbitmq/rabbit@TK-RABBITMQ3-sasl.log
Error: {could_not_start,rabbit,
           {{failed_to_cluster_with,
                ['rabbit@TK-RABBITMQ1','rabbit@TK-RABBITMQ2'],
                "Mnesia could not connect to any nodes."},
            {rabbit,start,[normal,[]]}}}

10 启动节点 TK-RABBITMQ1 ，再启动节点 TK-RABBITMQ3

11 查看队列信息

$ rabbitmqctl list_queues name policy messages_ready  pid slave_pids
  
test.one    test_policy 10  <rabbit@TK-RABBITMQ3.3.27052.123> []
test.three  test_policy 10  <rabbit@TK-RABBITMQ3.3.27041.123> []
test.two    test_policy 10  <rabbit@TK-RABBITMQ3.3.27038.123> []

结论

• 任意时候必须有一个 disk node 在运行。
• 在内存节点上做镜像，消息不会丢失

 

更多实验可点击：Rabbitmq 实验

 

转载于:https://my.oschina.net/u/3568300/blog/3006757