On 2012-03-27T09:17:15, Dominik Epple <[email protected]> wrote:
> Since I don't know what strange events may have lead to this situation
> (perhaps it was a manual manipulation), I created the following testcase to
> investigate this: configure a cluster, add some failover IP address primitive
> resource, let pacemaker start in on one host. Configure this IP address
> manually on the other host. Hope that pacemaker will detect it and react
> somehow (stop the resource, or do some stonith, or whatever).
>
> But even with some monitoring op defined, pacemaker seems only to check that
> the resource is running on the host it is scheduled for, and it does not
> check that the resource is not running on the host where it should not run.
> (I verified that the resource agent itself determines the status correctly,
> i.e. on the started resource, the monitoring action returns 0, on the stopped
> resource, it returns 7.)
>
> Is this observation correct? Is this the expected behavior? Is it possible to
> configure the resource such that it is checked that the resource is stopped
> where it is expected to be stopped?
Pacemaker will explicitly check for resources being active on start-up
on a node (we call that particular call to monitor "probe").
After that, we don't expect admins to manually mess things up. ;-)
You can, however, configure an explicit monitor for role="Stopped",
which will be able to detect primitives that are active where they
shouldn't be (within the configured interval). Yan added that on my
request for feature parity with another well-known cluster solution.
That will eventually catch them, but of course, if it really matters,
you'll already be looking at data/service corruption in the meantime.
Regards,
Lars
--
Architect Storage/HA
SUSE LINUX Products GmbH, GF: Jeff Hawn, Jennifer Guild, Felix Imendörffer, HRB
21284 (AG Nürnberg)
"Experience is the name everyone gives to their mistakes." -- Oscar Wilde
_______________________________________________
Pacemaker mailing list: [email protected]
http://oss.clusterlabs.org/mailman/listinfo/pacemaker
Project Home: http://www.clusterlabs.org
Getting started: http://www.clusterlabs.org/doc/Cluster_from_Scratch.pdf
Bugs: http://bugs.clusterlabs.org
