[Pacemaker] multistate resource warning

Frank DiMeo Frank.DiMeo at bigbandnet.com
Wed Jan 6 10:52:41 EST 2010


When I remove the target-role=master specifier from the configuration, and then put the node where the master resources are running into standby, the resources on the remaining node do not get promoted to the master state.  Only with this specifier included will the slaves be promoted to master in this scenario.  Is that correct?

-Frank

> -----Original Message-----
> From: Lars Ellenberg [mailto:lars.ellenberg at linbit.com]
> Sent: Tuesday, January 05, 2010 11:21 AM
> To: pacemaker at oss.clusterlabs.org
> Subject: Re: [Pacemaker] multistate resource warning
> 
> On Tue, Jan 05, 2010 at 09:20:53AM -0500, Frank DiMeo wrote:
> > Hi All,
> >
> >
> >
> > I've got a simple multistate resource running on a two node cluster.
> > The enclosed configuration is happy accepted by the CIB, and I see
> the
> > resource running in the master/slave state on the two nodes, but when
> > I run a crm_verify, I get a warning:
> >
> >
> >
> > cibadmin -replace --xml-file 1rsc_basic.xml
> >
> > hacluster at ubuntu_2:~/clones$ crm_verify --live-check -V
> >
> > crm_verify[7640]: 2010/01/05_09:11:49 WARN: get_target_role: ms-
> world1
> > is not part of a master/slave resource, a target-role of 'Master'
> > makes no sense
> >
> > Warnings found during check: config may not be valid
> >
> >
> >
> > I think my configuration is valid...is this warning valid?
> 
> yes. as it says:
> target role of Master makes no sense.
> 
> where, when, or if, a resource is promoted does not depend on its
> target
> role, but on "master scores".
> 
> there has even been a bug in pacemaker (up to 1.0.3? 1.0.4?) where a
> "target-role=Master" resulted in the resource being (tried to be)
> promoted everywhere, disregarding any master-max settings.
> 
> as a side note, also try to avoid location constraints on Master roles.
> in our experience it can confuse the crm into believing that it would
> always be ok to promote on those nodes, which is not necessarily the
> case.
> ymmv.
> it probably depends on the behaviour of the RA as well.
> 
> 
> --
> : Lars Ellenberg
> : LINBIT | Your Way to High Availability
> : DRBD/HA support and consulting http://www.linbit.com
> 
> DRBD(r) and LINBIT(r) are registered trademarks of LINBIT, Austria.
> 
> _______________________________________________
> Pacemaker mailing list
> Pacemaker at oss.clusterlabs.org
> http://oss.clusterlabs.org/mailman/listinfo/pacemaker




More information about the Pacemaker mailing list