[Pacemaker] asymmetric clusters, remote nodes, and monitor operations

[Lindsay Todd]

Hello Andrew,

I do have the symmetric-cluster=false property.

Since this is possibly a bug, I've submitted a crm_report as bug #5176, so
not only should you have a snapshot of logging information, you can also
see the configuration.  Thanks.

/Lindsay


On Wed, Sep 4, 2013 at 12:49 AM, Andrew Beekhof <andrew at beekhof.net> wrote:

>
> On 04/09/2013, at 6:18 AM, Lindsay Todd <rltodd.ml1 at gmail.com> wrote:
>
> > We've been attempting to set up an asymmetric pacemaker cluster using
> remote cluster nodes, with pacemaker 1.1.10 (actually, building from git
> lately, currently at a4eb44f).  We use location constraints to enable
> resources to start on nodes they should start on, and rely on asymmetry to
> otherwise keep resources from starting.
>
> You set symmetric-cluster=false or assumed that was the default
>
> >
> > But we get many monitor operation failures.
> >
> > Resource monitor operations run on the physical real hosts, and
> frequently fail because not all the components are present on those hosts.
>  For instance, the mysql resource agent's monitor operation fails as "not
> installed", since, well, mysql isn't installed on those systems, so the
> validate operation, which most or every path through that agent runs,
> always fails.  I don't see failures on the remote nodes, even ones without
> mysql installed.
> >
> > Previously I'd noticed LSB resources had failed monitor operations on
> systems that didn't have the LSB init script installed.
> >
> > Presumably these monitor operations are happening to ensure the resource
> is NOT running where it should not be???
>
> Correct. Although with symmetric-cluster=false it shouldn't show up as an
> error.
> Logs? crm_mon output?
>
> >  There doesn't seem to be a way to set up location constraints to
> prevent this from happening, at least that I've found.  I wrote an OCF
> wrapper RA to help me with LSB init scripts, but not sure what to do about
> other RA's like mysql short of maintaining my own version, unless there is
> a way to tune where "monitor" runs.  Or more likely I'm missing something
> ...
> >
> > It would seem to me that a "not installed" failure, OCF_ERR_INSTALLED,
> would not really be an error on a node that shouldn't be running that
> resource agent anyway, and is probably a pretty good indication that it
> isn't running.
> >
> > /Lindsay
> >
> > _______________________________________________
> > Pacemaker mailing list: Pacemaker at oss.clusterlabs.org
> > 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
>
>
> _______________________________________________
> Pacemaker mailing list: Pacemaker at oss.clusterlabs.org
> 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
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.clusterlabs.org/pipermail/pacemaker/attachments/20130904/89dabe1f/attachment.htm>