Pacemaker restarts multiple nodes when reassigning maser

[Anthony Landry]

Detailed bug description:
If the master rabbimq node is isolated and is not restarted by pacemaker, when it attempts to rejoin the cluster multiple nodes are restarted instead of just one.

Reproducibility:
 Yes, in an isolated environment

Workaround:
 manually modified cib.xml to contain the following bit of code:

<nvpair id="p_rabbitmq-server-meta_attributes-requires" name="requires" value="nothing"/>

Impact:
Monitoring services that require the use of persistent messages can lose the occasional message due to the rabbit resets from a split-brain environment.

[Anthony Landry (alandry)]

sla2 for 9.0-updates

[Anthony Landry (alandry)]

Assigned to me as I'm working on a workaround with the customer

[Anthony Landry (alandry)]

Addendum to this bug, this is a feature, and is intended working order for RMQ, it is requested to be changed in the pacemaker script to accommodate durable queues for monitoring purposes with the queue-sync command. However, this makes RMQ unstable and prone to crashing in MOS, therefore it is not recommended.

[Denis Meltsaykin (dmeltsaykin)]

Anthony, I'm not sure we want/can change this option since it was designed that way intentionally. How do you think can we close the bug report as Invalid?

[Anthony Landry (alandry)]

The customer hasn't responded to anything, was going to close out the
ticket today after I closed the salesforce

On Fri, Aug 31, 2018 at 5:11 AM Denis Meltsaykin <email address hidden>
wrote:

> Anthony, I'm not sure we want/can change this option since it was
> designed that way intentionally. How do you think can we close the bug
> report as Invalid?
>
> --
> You received this bug notification because you are subscribed to the bug
> report.
> https://bugs.launchpad.net/bugs/1786316
>
> Title:
> Pacemaker restarts multiple nodes when reassigning maser
>
> Status in Fuel for OpenStack:
> In Progress
>
> Bug description:
> Detailed bug description:
> If the master rabbimq node is isolated and is not restarted by
> pacemaker, when it attempts to rejoin the cluster multiple nodes are
> restarted instead of just one.
>
> Reproducibility:
> Yes, in an isolated environment
>
> Workaround:
> manually modified cib.xml to contain the following bit of code:
>
> <nvpair id="p_rabbitmq-server-meta_attributes-requires"
> name="requires" value="nothing"/>
>
> Impact:
> Monitoring services that require the use of persistent messages can lose
> the occasional message due to the rabbit resets from a split-brain
> environment.
>
> To manage notifications about this bug go to:
> https://bugs.launchpad.net/fuel/+bug/1786316/+subscriptions
>

[Denis Meltsaykin (dmeltsaykin)]

Closed as Invalid.