Dev: Luca

Goal:

Document RabbitMQ quorum queues for Greenfield deployments

• Provide high-level goal statement, providing user context and expected user outcome(s) for this Epic*

Historical context

In OSP17 and older versions we set a runtime policy via pacemaker resource agents to force all queues to be mirrored:
Resource: rabbitmq (class=ocf provider=heartbeat type=rabbitmq-cluster) Attributes: rabbitmq-instance_attributes set_policy='ha-all ^(?!amq\.).*

'

OpenStack services that consume the rabbitmq cluster have their queues and exchanges mirrored without explicit configurations required in the respective configuration files.

RHOSO before FR2
In RHOSO18 we decided not to set this policy anymore (true up to fr2, see below) to let each service decide the most appropriate queue configuration.
Most of the services seem to have lifted and shifted their oslo configuration from OSP17 so they are still operating under the assumption that there is a global rabbit policy in place, or that rabbitmq is providing a similar transparent configuration policy.
RHOSO FR2+
Because of customers pressure and disruptiveness of a minor update on API availability we had to find a solution to the lack of explicit queue configurations for each service that could:
Be applied to existing environments
Did not require downtime
Would provide resiliency to updates and rabbitmq pods disruptions

The only workaround that would satisfy all the three requirements was to apply the same policy as we have in OSP17.

Outcome : Capturing this change in docs

• Derived from one or more of the Use Cases/Scenarios or Acceptance Criteria from the parent Feature (or Initiative)
• 2-3 sentences... 

Acceptance Criteria:

• Documentation accepted by the PIDONE DFG