On 06/29/2012 04:02 PM, EXTERNAL Konold Martin (erfrakon, RtP2/TEF72) wrote: > Hi, > > I am experiencing an error situation which gets not detected by the cluster. > > I created a 2-node cluster using drbd and want to use ocfs2 on both nodes > simultaneously. (stripped off some monitor/meta stuff)
baaaaad idea ... pretty useless without the full configuration, especially the meta attributes in this case. Also share your drbd and corosycn configuration please. BTW: what is your use case to start with a "simple" dual-primary OCFS2 setup? > > primitive dlm ocf:pacemaker:controld > primitive o2cb ocf:ocfs2:o2cb > primitive resDRBD ocf:linbit:drbd \ > params drbd_resource="r0" \ > operations $id="resDRBD-operations" > primitive resource-fs ocf:heartbeat:Filesystem \ > params device="/dev/drbd_r0" directory="/SHARED" fstype="ocfs2" > ms msDRBD resDRBD > clone clone-dlm dlm > clone clone-fs resource-fs > clone clone-ocb o2cb > colocation colocation-dlm-drbd inf: clone-dlm msDRBD:Master > colocation colocation-fs-o2cb inf: clone-fs clone-ocb > colocation colocation-ocation-dlm inf: clone-ocb clone-dlm > order order-dlm-o2cb 0: clone-dlm clone-ocb > order order-drbd-dlm 0: msDRBD:promote clone-dlm:start > order order-o2cb-fs 0: clone-ocb clone-fs > > The cluster starts up happily. (everything green in crm_gui) but > > rt-lxcl9a:~ # drbd-overview > 0:r0/0 WFConnection Primary/Unknown UpToDate/DUnknown C r----- > rt-lxcl9b:~ # drbd-overview > 0:r0/0 StandAlone Primary/Unknown UpToDate/DUnknown r----- > > As you can see this is a split brain situation with both nodes having the > ocfs2 fs mounted but not in sync --> data loss will happen. > > 1. How to avoid split brain situations (I am confident that the cross > link using a 10GB cable was never interrupted)? logs should reveal what happend > 2. How to resolve this? > Switch cluster in maintenance mode and then follow > http://www.drbd.org/users-guide/s-resolve-split-brain.html ? at least you also need to stop the filesystem if it is running and you want to demote one Primary ... and then follow that link > 3. How to make the cluster aware of the split brain situation? (It > thinks everything is fine) setup fening method "resource-and-stonith" in drbd configuration, preferable use the "crm-fence-peer.sh" stonith script ... Pacemaker itself or better the DRBD resource agent will not react on such a situation. > 4. Should the DRBD(OCFS2 setup be maintained outside the cluster instead? better not ;-) Regards, Andreas -- Need help with Pacemaker? http://www.hastexo.com/now > > Mit freundlichen Grüßen / Best regards > > Martin Konold > > Robert Bosch GmbH > Automotive Electronics > Postfach 13 42 > 72703 Reutlingen > GERMANY > www.bosch.com<http://www.bosch.com> > > Tel. +49 7121 35 3322 > > Sitz: Stuttgart, Registergericht: Amtsgericht Stuttgart, HRB 14000; > Aufsichtsratsvorsitzender: Hermann Scholl; Geschäftsführung: Franz > Fehrenbach, Siegfried Dais; > Stefan Asenkerschbaumer, Bernd Bohr, Rudolf Colm, Volkmar Denner, Christoph > Kübel, Uwe Raschke, > Wolf-Henning Scheider, Werner Struth, Peter Tyroller > > > > _______________________________________________ > Linux-HA mailing list > [email protected] > http://lists.linux-ha.org/mailman/listinfo/linux-ha > See also: http://linux-ha.org/ReportingProblems >
signature.asc
Description: OpenPGP digital signature
_______________________________________________ Linux-HA mailing list [email protected] http://lists.linux-ha.org/mailman/listinfo/linux-ha See also: http://linux-ha.org/ReportingProblems
