[Pacemaker] Pacemaker resource management

Dejan Muhamedagic dejanmm at fastmail.fm
Tue Jun 8 17:29:25 UTC 2010


Hi,

On Sun, Jun 06, 2010 at 05:47:35PM +0300, Dan Frincu wrote:
> Hello all,
> 
> I have a couple of questions and I haven't found any relevant
> documentation about it so I would appreciate any answers on the
> matter.
> 
> I'm using drbd 8.3.2-6 with pacemaker 1.0.5-4.2, openais 0.80.5-15.2
> and heartbeat 3.0.0-33.3 for a high availability 2 node cluster for
> mysql and apache with drbd partitions.
> 
> What I want to know is if a a resource fails, such as apache,
> pacemaker tries to restart the service, it has to do with
> "common_apply_stickiness", from what I can see in the logs.

I guess that you mean default-resource-stickiness and
resource-stickiness.

> 1. How many times does pacemaker try to restart a resource before
> declaring it "down" and migrating the resource (and dependencies) to
> the other node?

migration-threshold

> 2. How can I alter this behavior, to be able to set the number of
> retries a resource is attempted to be restarted before migrating it
> to the other available node?

ditto. Note that if fail counts on both nodes reach
migration-threshold the resource won't run anywhere.

> I've noticed that sometimes, if there is a problem with the block
> device (drbd) the cluster will go into a stage where it migrates all
> resources in a group from A to B, however, when trying to start
> resources on B, there is a synchronization issue, one block device
> is still being in process of being updated from node A drbd0 to node
> B drbd0. In this case the group resources don't start until the
> synchronization is complete.
> 
> 3. Can I "force" a group of resources to migrate to another node if
> any of the resources fails to be brought up within a number of
> retries or after a timeout (including if the group is just being
> migrated from A to B, but one resource fails to start on B, to be
> migrated back to A)? How?

Again migration-threshold.

> 4. Is there a Resource Agent out there that can be configured to
> send SNMP traps?

ClusterMon

Thanks,

Dejan

> Thank you in advance for your replies.
> 
> Best regards.
> 
> 
> 
> 
> 
> 
> 
> 
> _______________________________________________
> 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