[ClusterLabs] Need help for specific resource manipulation

Dejan Muhamedagic dejanmm at fastmail.fm
Mon Dec 21 06:59:38 EST 2015


Hi,

On Tue, Dec 15, 2015 at 04:34:25PM +0100, Ulrich Windl wrote:
> Hi!
> 
> I have two scenarios where I could need help from the experts:
> 
> 1) For a master/slave resource, is it possible to stop just the
> master or slave (and let the cluster do the recovery)? For

On a specific node? Or something like slave stop/master stop? You
could reduce the max-nodes and then one slave should be stopped.
Otherwise, if you know where the slave is running, you can create
a location constraint with -inf as a score for that resource and
that node.

> clone resources the failed operations get suffixes like ":0",
> ":1", ... Would it work with master/slave also? (SLES11 SP4)

I guess so, but not sure.

> 2) In a three-node cluster with nodes (O1, O2, N3), node N3 has
> a somewhat newer CPU than the O1 and O2 nodes (which have
> identical CPUs). There is a problem with Xen PVMs: If a PVM is
> created on one of the older nodes, it can be live migrated from
> old to new and back; however if the PVM is created on N3, live
> migration to the older nodes fails, and the resource is
> effectively destroyed hard and then restarted by the cluster (a
> service request is in progress already).  Thus my question: Can
> I formulate location preferences that say a PVM should not be
> created on N3, but can be migrated to N3 once it has been
> started? Obviously any plain location constraint won't do that.

I don't think so, apart from providing location preferences which
would make it very unlikely for a resource to start on that node.

Thanks,

Dejan

> Regards,
> Ulrich
> 
> 
> 
> _______________________________________________
> Users mailing list: Users at clusterlabs.org
> http://clusterlabs.org/mailman/listinfo/users
> 
> Project Home: http://www.clusterlabs.org
> Getting started: http://www.clusterlabs.org/doc/Cluster_from_Scratch.pdf
> Bugs: http://bugs.clusterlabs.org




More information about the Users mailing list