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 [email protected] 'echo -n "" > /var/lib/rabbitmq/.erlang.cookie'
rabbitmq-02:~$ rabbitmqctl cluster [email protected]
Clustering node '[email protected]' with ['[email protected]'] ...
...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 '[email protected]' ...
[{nodes,[{disc,['[email protected]']},{ram,['rabbitmq-02']}]},
 {running_nodes,['[email protected]','[email protected]']}]
 ...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 '[email protected]' ...
[{nodes,[{disc,['[email protected]']},{ram,['[email protected]']}]},
 {running_nodes,['[email protected]','[email protected]']}]
...done.
rabbitmq-02:~$ sudo rabbitmqctl stop_app
Stopping node '[email protected]' ...
...done.
rabbitmq-02:~$ sudo rabbitmqctl cluster [email protected] [email protected]
Clustering node '[email protected]' with ['[email protected]',
                                         '[email protected]'] ...
...done.
rabbitmq-02:~$ sudo rabbitmqctl start_app
Starting node '[email protected]' ...
...done.
rabbitmq-02:~$ sudo rabbitmqctl cluster_status
Cluster status of node '[email protected]' ...
[{nodes,[{disc,['[email protected]','[email protected]']}]},
 {running_nodes,['[email protected]','[email protected]']}]
...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!