[Pacemaker] unordered groups - is this safe to use?

Andrew Beekhof andrew at beekhof.net
Thu Mar 3 08:26:20 UTC 2011


no

On Tue, Mar 1, 2011 at 11:19 AM, David Heidt <david.heidt at msales.com> wrote:
> Hello List,
> let me first explain my environment:
>
> * 2 Hosts with shared lvm-storage
> * Debian Squeeze x64  with official squeeze heartbeat and pacemaker
> * All resources are ocf:heartbeat:Xen
> * One host is meant to host all production resources, the other is a hot-standby
> * We want to run instant testing systems based on lvm snapshots on the
> standby node.
>
> This way, we can do heavy or dangerous tests without influencing the
> production-host and taking advantage of the unused standby-hardware
> In case of a hardware failure, all production resources have priority
> (that is the colocation at the end)
> To freely stop and restart resouces, i packed them into a group, but
> no resource depends on another one, so the group has to be uncolocated
> and unordered.
> Finally i came to this setup:
>
> <snip>
>
> group grp_production rsc_1 rsc_2 \
> meta ordered="false" collocated="false"
>
> group grp_testing rsc_1_test rsc_2_test \
> meta ordered="false" collocated="false"
>
> colocation coloc_production_first -inf: grp_testing grp_production
>
> </snip>
>
> But following this thread makes me think of redesigning my crm setup:
> http://oss.clusterlabs.org/pipermail/pacemaker/2011-January/008942.html
>
> Quote Andrew Beekhof: "Unordered and/or uncolocated groups are an abomination."
>
> What do you think? are resource_sets maybe a suitable alternative?
>
> Regards,
>
> David
>
> _______________________________________________
> 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://developerbugs.linux-foundation.org/enter_bug.cgi?product=Pacemaker
>




More information about the Pacemaker mailing list