On Wed, Jun 24, 2009 at 11:35:02AM +0200, Peter Luciak wrote:
> Noone ever encountered a problem when starting the cluster in degraded 
> mode with DRBD Peer Outdater? Or I'm doing something wrong here...
> 
> 
> I am facing the following problem in my v1 configuration with dopd and
> STONITH:
> 
> After consecutive shutdown of both nodes of cluster (ie. first active
> node is shutdown, becomes outdated, resources are migrated to passive
> node, then the passive node is shutdown as well) and starting the
> outdated node _only_, the following happens:
> 
> - node starts up with DRBD resources Secondary and Outdated (I have
> wfc-timeout set)
> - Heartbeat starts up, but obviously refuses to start drbd resources
> thus leaving cluster nonfunctional
> - during Heartbeat startup, the other node is STONITH'd which causes the
> other node to boot
> - after some time, both nodes are up in Secondary/Secondary
> UpToDate/UpToDate     
> 
> For STONITH I'm using iLO2:
> stonith_host srv1 external/riloe srv2-ilo 10.4.5.12 Administrator
> password 1 2.0 off
> 
> The whole thing seems logical, but my question is, how to avoid such
> situation? :) Or, is it possible to resolve this scenario automatically,
> without user intervention?

"avoid such situation" in the sense of
"do not outdate on intentional cluster shutdown"?

Put your drbd into "target-role=stopped" first.  Then crm should down
the drbd cleanly, before you shut down the cluster nodes.

Ah.
You mentioned you are "v1 configuration", i.e. haresources.
Well, just the same: stop your current "secondary" heartbeat first,
so it will not take over when you stop your "primary" heartbeat.
DRBD should then be Connected Secondary/Secondary.
 From there you can shutdown and reboot your boxes in any order.

-- 
: Lars Ellenberg
: LINBIT | Your Way to High Availability
: DRBD/HA support and consulting http://www.linbit.com

DRBD® and LINBIT® are registered trademarks of LINBIT, Austria.
_______________________________________________
Linux-HA mailing list
[email protected]
http://lists.linux-ha.org/mailman/listinfo/linux-ha
See also: http://linux-ha.org/ReportingProblems

Reply via email to