高可用消息队列服务构建-RABBITMQ

openstack采用了一个松耦合的方式来进行各个组建之间的通信，具体来讲组建之间的通信是通过rabbitmq的消息队列进行的。

当openstack系统规模变大时，消息队列就成为了性能发挥的瓶颈。

网上有人给出了集群化的消息队列配置方案，如下：

原文：http://www.sebastien-han.fr/blog/2012/05/21/openstack-high-availability-rabbitmq/

Rabbitmq has his own buildin cluster management system. Here, we don’t need Pacemaker, everything is managed by RabbitMQ itself.

RabbitMQ or more generally the management queues layer is a critical component of OpenStack because every requests/queries use this layer to communicate.

CLUSTERING SETUP

1	$ sudo apt-get install rabbitmq-server

RabbitMQ generates a cookie for each server instance. This cookie must be the same on each member of the cluster:

1 2 3 4 5 6 7 8

rabbitmq-01:~$ $ sudo cat /var/lib/rabbitmq/.erlang.cookie ITCWRVSIDPHRSLGXBHCFc rabbitmq-02:~$ rabbitmqctl stop_app rabbitmq-02:~$ rabbitmqctl reset rabbitmq-01:~$ ssh localadmin@rabbitmq-02 'echo -n "" > /var/lib/rabbitmq/.erlang.cookie' rabbitmq-02:~$ rabbitmqctl cluster rabbit@rabbitmq-01 Clustering node 'rabbit@rabbitmq-02' with ['rabbit@rabbitmq-01'] ... ...done.

Check your cluster status, on the node 01 or 02, whatever:

1 2 3 4 5

rabbitmq-02:~$ sudo rabbitmqctl cluster_status Cluster status of node 'rabbit@rabbitmq-02' ... [{nodes,[{disc,['rabbit@rabbitmq-01']},{ram,['rabbitmq-02']}]}, {running_nodes,['rabbit@rabbitmq-01','rabbit@rabbitmq-02']}] ...done.

Eventually check your queues on both nodes, they should be identical:

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29

$ sudo rabbitmqctl list_queues name slave_pids synchronised_slave_pids Listing queues ... network_fanout_3a9877aa044c47b7846bbe01bf7faa4a volume.server-03 amq.gen-gXDebEl-MNze8Nv83Ox3wD volume_fanout_2df3d830f3ac40a98fe924bb032f6512 cert_fanout_4b1d7951b629470db0d13880b5814a00 amq.gen-waleDspYQ1IP5lKkpmkc9P volume scheduler.server-05 consoleauth_fanout_a0c57c1bca3645eda5b25788de9dd484 amq.gen-gxnO_iZvkdB9BaLHX6g9LT amq.gen-AsSmN5K6zqJ3-xDlJeP16t amq.gen-A9B2HzA2dZmHjP-bNv2rwb compute scheduler network cert.server-05 consoleauth.server-05 network.server2 compute.server2 amq.gen-wN5T2Ylyqbu41-_RzJ4vf8 amq.gen-wHYGWuERrgzEHbBPrmTfwS scheduler_fanout_d57ca4552c9542ed860a0ef8a9d48534 cert compute_fanout_b9c1dd93cb1a4282bb32c066e7d621d2 amq.gen-QtDR90yqOLXuaNEPHozmt_ consoleauth ...done.

Cluster nodes can be of two types: disk or ram. Disk nodes replicate data in ram and on disk, thus providing redundancy in the event of node failure and recovery from global events such as power failure across all nodes. Ram nodes replicate data in ram only and are mainly used for scalability. A cluster must always have at least one disk node.

You can also verify that the connection is well established between the node:

1 2	$ sudo netstat -plantu | grep 10.0. tcp 0 0 10.0.0.1:39958 10.0.0.2:46117 ESTABLISHED 5294/beam.smp

TIPS

CHANGE THE IP OR THE HOSTNAME OF A NODE

If you changed you IP address or your hostname, this is pretty nasty and harsh but it works:

1 2 3 4 5 6

$ sudo rabbitmqctl stop_app $ sudo dpkg-reconfigure rabbitmq-server Stopping rabbitmq-server: RabbitMQ is not running rabbitmq-server. Starting rabbitmq-server: SUCCESS rabbitmq-server.

The IP address and/or will be refresh in the rabbitmq database.

CONVERT RAM NODE TO DISK NODE

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20

rabbitmq-02:~$ sudo rabbitmqctl cluster_status Cluster status of node 'rabbit@server-02' ... [{nodes,[{disc,['rabbit@server-01']},{ram,['rabbit@server-02']}]}, {running_nodes,['rabbit@server-01','rabbit@server-02']}] ...done. rabbitmq-02:~$ sudo rabbitmqctl stop_app Stopping node 'rabbit@server-02' ... ...done. rabbitmq-02:~$ sudo rabbitmqctl cluster rabbit@server-01 rabbit@server-02 Clustering node 'rabbit@server-02' with ['rabbit@server-01', 'rabbit@server-02'] ... ...done. rabbitmq-02:~$ sudo rabbitmqctl start_app Starting node 'rabbit@server-02' ... ...done. rabbitmq-02:~$ sudo rabbitmqctl cluster_status Cluster status of node 'rabbit@server-02' ... [{nodes,[{disc,['rabbit@server-02','rabbit@server-01']}]}, {running_nodes,['rabbit@server-01','rabbit@server-02']}] ...done.

HAPROXY CONFIGURATION

Clustering doesn’t mean high-availability, this is why I put a load-balancer on top. Here HAProxy will balance the request only on one node, if this node fails the request will be route to the other node. It’s simple as that. The native port of HAproxy and the OpenStack queues ports are configured.

global
    log 127.0.0.1   local0
    #log loghost    local0 info
    maxconn 1024
    #chroot /usr/share/haproxy
    user haproxy
    group haproxy
    daemon
    #debug
    #quiet

defaults
    log     global
#log 127.0.0.1:514 local0 debug
    log 127.0.0.1   local1 debug
    mode    tcp
    option  tcplog
    option  dontlognull
    retries 3
    option redispatch
    maxconn 1024

# Default!
   timeout connect 5000ms
   timeout client 50000ms
   timeout server 50000ms

listen rabbitmq_cluster 0.0.0.0:4369
    mode tcp
    balance roundrobin
    server server-07_active 172.17.1.8:4369 check inter 5000 rise 2 fall 3
    server server-08_backup 172.17.1.9:4369 backup check inter 5000 rise 2 fall 3

listen rabbitmq_cluster_openstack 0.0.0.0:5672
    mode tcp
    balance roundrobin
    server server-07_active 172.17.1.8:5672 check inter 5000 rise 2 fall 3
    server server-08_backup 172.17.1.9:5672 backup check inter 5000 rise 2 fall 3

Et voilà!