blob: 7bacebe16b000d34632b420ac83e00323169fa42 [file] [log] [blame]
=========================
RabbitMQ messaging system
=========================
RabbitMQ is a complete and highly reliable enterprise messaging system based on the emerging AMQP standard.
Sample pillars
==============
Standalone Broker
-----------------
RabbitMQ as AMQP broker with admin user and vhosts
.. code-block:: yaml
rabbitmq:
server:
enabled: true
bind:
address: 0.0.0.0
port: 5672
secret_key: rabbit_master_cookie
admin:
name: adminuser
password: pwd
plugins:
- amqp_client
- rabbitmq_management
virtual_hosts:
- enabled: true
host: '/monitor'
user: 'monitor'
password: 'password'
RabbitMQ as a Stomp broker
.. code-block:: yaml
rabbitmq:
server:
enabled: true
secret_key: rabbit_master_cookie
bind:
address: 0.0.0.0
port: 5672
virtual_hosts:
- enabled: true
host: '/monitor'
user: 'monitor'
password: 'password'
plugins:
- rabbitmq_stomp
RabbitMQ cluster
----------------
RabbitMQ as base cluster node
.. code-block:: yaml
rabbitmq:
server:
enabled: true
bind:
address: 0.0.0.0
port: 5672
secret_key: rabbit_master_cookie
admin:
name: adminuser
password: pwd
cluster:
enabled: true
role: master
mode: disc
members:
- name: openstack1
host: 10.10.10.212
- name: openstack2
host: 10.10.10.213
HA Queues definition
.. code-block:: yaml
rabbitmq:
server:
enabled: true
...
virtual_hosts:
- enabled: true
host: '/monitor'
user: 'monitor'
password: 'password'
policies:
- name: HA
pattern: '^(?!amq\.).*'
definition: '{"ha-mode": "all"}'
Usage
=====
Check cluster status, example shows running cluster with 3 nodes: ctl-1, ctl-2, ctl-3
.. code-block:: yaml
> rabbitmqctl cluster_status
Cluster status of node 'rabbit@ctl-1' ...
[{nodes,[{disc,['rabbit@ctl-1','rabbit@ctl-2','rabbit@ctl-3']}]},
{running_nodes,['rabbit@ctl-3','rabbit@ctl-2','rabbit@ctl-1']},
{partitions,[]}]
...done.
Setup management user.
.. code-block:: yaml
> rabbitmqctl add_vhost vhost
> rabbitmqctl add_user user alive
> rabbitmqctl set_permissions -p vhost user ".*" ".*" ".*"
> rabbitmqctl set_user_tags user management
EPD process is Erlang Port Mapper Daemon. It's a feature of the Erlang runtime that helps Erlang nodes to find each other. It's a pretty tiny thing and doesn't contain much state (other than "what Erlang nodes are running on this system?") so it's not a huge deal for it to still be running.
Although it's running as user rabbitmq, it was started automatically by the Erlang VM when we started. We've considered adding "epmd -kill" to our shutdown script - but that would break any other Erlang apps running on the system; it's more "global" than RabbitMQ.
Read more
=========
* http://www.rabbitmq.com/admin-guide.html
* https://github.com/saltstack/salt-contrib/blob/master/states/rabbitmq_plugins.py
* http://docs.saltstack.com/ref/states/all/salt.states.rabbitmq_user.html
* http://stackoverflow.com/questions/14699873/how-to-reset-user-for-rabbitmq-management
* http://www.rabbitmq.com/memory.html
Clustering
==========
* http://www.rabbitmq.com/clustering.html#auto-config
* https://github.com/jesusaurus/hpcs-salt-state/tree/master/rabbitmq
* http://gigisayfan.blogspot.cz/2012/06/rabbit-mq-clustering-python-fabric.html
* http://docwiki.cisco.com/wiki/OpenStack_Havana_Release:_High-Availability_Manual_Deployment_Guide#RabbitMQ_Installation