[Linux-HA] can't get "default_resource_stickiness" working

Chris Gallo chrisagallo at gmail.com
Mon Oct 2 07:08:31 MDT 2006


In your cib.xml, I don't actaully see that you specified a stickyness
for any of your resources. If you want all the resources to stay where
they get put, you will want to change the default to "INFINITY". Or if
you only want certain resources to stay put and others to migrate back
you should have something like the following for your resource
primitive

<primitive id="ldap" class="ocf" type="ldap" provider="ldap"
resource_stickiness="INFINTY">
And a real value, in your case 100 or less to have it migrate back.

I hope that helps.

-ChrisG

On 10/2/06, Andreas Kurz <akurz at sms.at> wrote:
> Hello!
>
> I am running a three node cluster with heartbeat 2.0.7 on CentOS 4.2.
>
> My idea was distributing the load between two nodes and balance the
> resources and let the third node be a quorum node without running a service.
>
> In addition to that I configured a pingd clone resource to monitor the
> gateway connectivity and to force a failover when a node loses the
> connectivity to the default gateway.
>
> Now I have some troubles getting "auto_failback off" running, although I
> have a "default_resource_stickiness" configured with a (in my opinion)
> suitable value. Whenever a dead node is online again it migrates its
> resources back instead of leaving them untouched on the current server.
>
> The log messages on the DC look like:
>
> pengine[31564]: 2006/10/02_12:58:52 info:
> choose_node_from_list:allocate.c 2 nodes with equal score (11000) for
> running the listed resources (chose clno02.intern.sms.at):
> pengine[31564]: 2006/10/02_12:58:52 info:
> choose_node_from_list:allocate.c 2 nodes with equal score (6600) for
> running the listed resources (chose clno01.intern.sms.at):
>
> For me it looks like the cluster is ignoring the
> "default_resource_stickiness" ????
>
> Maybe I am misinterpreting the manual and there is an obvious failure in
> my configuration, that I have attached to that mail.
>
> Thanks for your help!
>
> Regards,
> Andi
>
>
> _______________________________________________
> Linux-HA mailing list
> Linux-HA at lists.linux-ha.org
> http://lists.linux-ha.org/mailman/listinfo/linux-ha
> See also: http://linux-ha.org/ReportingProblems
>
>
>


More information about the Linux-HA mailing list