Hello,

On 02/03/2012 04:35 PM, Felix Frank wrote:
> Hi,
> 
> On 02/03/2012 04:05 PM, Owen Le Blanc wrote:
>> This still
>> leaves open the question of why split brain keeps occuring?  The
>> cluster is managed by pacemaker, version 1.1.6, with corosync, version
>> 1.4.2.  There isn't actually any real data on the two drbd devices,
>> since this is only a test.  But it concerns me that it seems to go
>> into split brain about once a week.

Be sure to have STONITH enabled and configured, use resource-level
fencing in DRBD and redundant ring setup for corosync ... so try to
minimize the chance for split-brain and establish protection in case
your nodes get separated.

> 
> this is neigh impossible to answer without looking at your setup and
> workflows.
> 
> Split brain can only happen when the nodes get disconnected (this
> includes downtimes of either node).

... and additionally they need to get promoted to Primary on both sides
while not connected.

Regards,
Andreas

-- 
Need help with Pacemaker/DRBD/Corosync?
http://www.hastexo.com/now

> How often are your nodes getting separated thus?
> 
> ...this is not dual-primary, is it?
> 
> Prune your logs, it should be easy to determine when DRBD changes
> states. As to why, well - analyzing pacemaker activity from hindsight
> can be challenging, I guess. Your best bet is probably to make sure the
> nodes don't get separated, and that maintenance is done very carefully.
> 
> HTH,
> Felix
> _______________________________________________
> drbd-user mailing list
> [email protected]
> http://lists.linbit.com/mailman/listinfo/drbd-user




Attachment: signature.asc
Description: OpenPGP digital signature

_______________________________________________
drbd-user mailing list
[email protected]
http://lists.linbit.com/mailman/listinfo/drbd-user

Reply via email to