On Wed, Jan 27, 2010 at 09:46:17AM +0100, Florian Haas wrote:
> On 2010-01-27 00:25, Hunny Bunny wrote:
> > Hello folkz,
> > 
> > I'm very puzzled by the error messages in  my /var/log/warn shown below.
> > I'm not using any Pacemaker or Hearbeat CRM at this point.
> > 
> > I have a plain configuration of DRBD which replicates /dev/sda4 as a block 
> > device /dev/drbd0 between two nodes node1 and node2
> > 
> > On a top of /dev/drbd0 block I have OCFS2 partition which is mounted on 
> > node1 and node2
> > as /data. 
> > 
> > Also /dev/drbd0 is available as iscsi-target from the node2 to the node3
> 
> Won't work. Simple as that. Sorry.

And just to make that clear: that is not DRBD's fault.

btw, the heartbeat mentioned in that error message is the OCFS2
heartbeat. Each OCFS2 mount writes "I'm <some-uuid> and I'm still alive"
to a certain "slot", then reads from that slot.

If something is read from that slot that this node did not write,
this nodes concludes that "someone else" is heartbeating there.
And is quite NOISY about this fact, as conflicting slots here mean
coordination problems, which, on concurrently accessed shared data,
is quite synonymous with data corruption.

You typically can not mount any cluster file system directly as well as
via iSCSI.  DRBD or not, if your initiators mount it, you must not mount
OCFS2 directly on the iSCSI target.

-- 
: 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