[Pacemaker] Start order after manual migration

Dejan Muhamedagic dejanmm at fastmail.fm
Wed Jun 15 16:14:19 UTC 2011


Hi,

On Wed, Jun 15, 2011 at 10:26:36AM +0200, Pawel Warowny wrote:
> Hi
> 
> I have problem with starting order on node after manual migration.
> My cluster consists of two nodes that provides bunch of kvm virtualized
> hosts.
> 
> When I start node, first starting drbd, dlm (and the like) next libvirt
> and the last hosts provided by VirtualDomain resource.
> 
> When I stop node, the resources stopping in reverse order (first
> VirtualDomain, next libvirt and so on).
> 
> And that works.
> 
> But there is a problem with this scenario:
> - resource vr_debian_squeeze is running on node1 (because of location
> rule score 100)
> - then I manual migrate this resource to node2 (crm resource migration
>   vr_debian_squeeze)
> - next I'm stopping the whole node2 (/etc/init.d/pacemaker stop)
> - vr_debian_squeeze can't automatically migrate to node1, because of
>   manual migration constrain:
> location cli-standby-vr_debian_squeeze vr_debian_squeeze \
>         rule $id="cli-standby-rule-vr_debian_squeeze" -inf: #uname eq node1
> - so vr_debian_squeeze is stopped with the whole node2, and that's ok
> 
> And here is my problem:
> I'm starting again node2 (/etc/ini.t/pacemaker start) and the first
> resource which is starting on that node is vr_debian_squeeze!

That's really odd. Is it that at this time vr_debian_squeeze
runs on node1? The first rule in pacemaker is not to start the
same resource on two or more nodes. If that really happens,
please file a bugzilla. And turn off the cluster and send for
fire brigades ;-)

Thanks,

Dejan



> Of course
> it can't start because resources required by vr_debian_squeeze are not
> running yet.
> 
> For simplicity important parts from my configuration:
> 
> primitive cluster_fs ocf:heartbeat:Filesystem
> primitive libvirt lsb:libvirtd
> primitive vr_debian_squeeze ocf:heartbeat:VirtualDomain
> 
> clone cluster_fs-clone cluster_fs
> clone libvirt-clone libvirt
> 
> location vr_debian_squeeze_loc vr_debian_squeeze 100: lolek
> 
> colocation libvirt_with_cluster_fs inf: libvirt-clone cluster_fs-clone
> colocation vr_debian_squeeze_with_libvirt 1000: vr_debian_squeeze libvirt-clone
> 
> order cluster_fs_before_libvirt inf: cluster_fs-clone libvirt-clone
> order vr_debian_squeeze_after_libvirt inf: libvirt-clone vr_debian_squeeze
> 
> And after crm resource migration vr_debian_suqeeze there is also:
> location cli-standby-vr_debian_squeeze vr_debian_squeeze \
>         rule $id="cli-standby-rule-vr_debian_squeeze" -inf: #uname eq node1
> 
> Best regards
> -- 
> Pawel Warowny



> _______________________________________________
> 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