RabbitMQ cluster, controlled by pacemaker.

Registered by Sergey Vasilenko on 2014-05-15

Our life with RabbitMQ cluster gave us following issues:
* RabbitMQ cluster can't assemble after reboot controllers (with probability close to 50%)
* When closter lost node, which responsible for the integrity of the cluster, cluster died.
This issues can't be resolved without external control. As the dictator's perfect Pacemaker. On its OCF-scripts we can realize any start/stop/checking logic.

Feature Lead: Sergey Vasilenko
Developers: Sergey Vasilenko
Mandatory Design Reviewers: Vladimir Kuklin
QA: Nastya Urlapova

Whiteboard

Gerrit topic: https://review.openstack.org/#q,topic:bug/1318936,n,z

Addressed by: https://review.openstack.org/93956
    Rabbitmq ocf master/slave

The flow charts http://goo.gl/vaoYNd (will be included in the RabbitMQ clustering related technical bulletin)

(?)

Work Items

Work items:
1-st stage blueprint: DONE
POC OCF script: DONE
Code review: DONE
QA: DONE
describing behavior and implementation details: DONE
implement partition checking as additional monitor task: TODO

Dependency tree

* Blueprints in grey have been implemented.

This blueprint contains Public information 
Everyone can see this information.