[Linux-HA] DRBD not becoming primary when master node fails
Harald.Husemann at materna.de
Mon Jul 20 03:49:35 MDT 2009
I have a 2-node cluster, running Pacemaker (crm_resource version 1.04)
with heartbeat stack and DRBD 8.3.0 (compiled myself) on two CentOS 5.3
I've added several resources, DRBDs and other, and I can migrate them
between the cluster nodes without any problems.
But, when I switch off the master node of one of the DRBD resources,
DRBD fails to start the DRBD on the remaining node.
I've figured out that the problem is that DRBD is unable to set the
resource on the remaining node to primary state, since the connection to
the old primary is lost, and this causes the disk to be in "consistent"
and not in "UpToDate" state which is necessary for it to become primary.
Googlin' a bit showed up that it's possible to force primary mode by
using the option "overwrite-data-of-peer" with drbdadm, i. e.:
drbdadm -- --overwerite-data-of-peer primary <rsc>.
(It could be necessary to set the resource to "outdated" before).
I've tried this manually, and it worked - but I'd like Pacemaker to do
this for me...
I took a look into the drbd resource script, and it seems it does not
handle this situation - is there a newer version of it which takes care
Netzwerk- und Systemadministrator
Operation Management Center (OMC)
Information & Communications
Geschäftsführer: Dr. Winfried Materna, Helmut an de Meulen, Ralph Hartwig
Amtsgericht Dortmund HRB 5839
Tel: +49 231 9505 222
Fax: +49 231 9505 100
More information about the Linux-HA