[Pacemaker] setup multimaster drbd with ocfs without o2cb and controld

[Andrew Beekhof]

On Fri, Jan 6, 2012 at 3:49 AM, thomas polnik <info+pacemaker at polnik.de> wrote:
> Hello,
>
> I want setup follow:
> 2 servers with multimaster drbd and ocfs filesystem and mount it.
>
> *system: gentoo, 2.6.39, pacemaker-1.1.5
>
> *Setup pacemaker:
>
> # misc settings
> property no-quorum-policy="ignore"
> property stonith-enabled="false"
> rsc_defaults resource-stickiness="200"
>
> # drbd setup
> primitive resDrbd ocf:linbit:drbd params drbd_resource="images" op start
> interval="0" timeout="240s" op stop interval="0" timeout="240s"
> ms msDrbd resDrbd meta master-max="2" clone-max="2" notify="true"
>
> # ocfs setup via lsb:ocfs2
> primitive resOcfs lsb:ocfs2 op monitor interval="20" timeout="40" meta
> is-managed="true" target-role="Started"
> clone clResOcfs resOcfs meta target-role="Started"
>
> # FS setup
> primitive resFsImages ocf:heartbeat:Filesystem \
>    params device="/dev/drbd/by-res/images" directory="/srv/images"
> fstype="ocfs2" options="rw,noatime" \
>    op start interval="0" timeout="60s" \
>    op stop interval="0" timeout="60s"
> clone clResFsImages resFsImages \
>    meta target-role="Started"
>
> # setup order of primitives
> order grDrbdOcfsFs inf: msDrbd:promote clResOcfs:start clResFsImages:start
>
> Problem:
> If one server goes down and comes back, the other node umount /srv/images,
> shutdown ocfs2 und stop drbd and then both nodes start the services again.
> After this, all services works fine again, but I have an outage duration of
> 5 seconds.

I think that was a bug in 1.1.5
Try 1.1.6?

>
> I think, this is not necessary, I don't know, why pacemaker shutdown all
> services on the running node.
>
> btw: I can not use ocf:ocfs2:o2cb and ocf:pacemaker:controld ([1]) because
> pacemaker-1.1.5 on a gentoo system does not offer me this agents. So I
> choose lsb:ocfs2 to use ocfs.

That means OCFS2 is using its own internal cluster comms.
So they and pacemaker may not agree on who the cluster members are at
some stage... then things are going to get /really/ interesting for
you.

>
> Has anybody a hint for me, how can I prevent it?
>
> Best regards,
> thomas.
>
>
> [1] http://www.drbd.org/users-guide/s-ocfs2-pacemaker.html
>
> _______________________________________________
> Pacemaker mailing list: Pacemaker at oss.clusterlabs.org
> http://oss.clusterlabs.org/mailman/listinfo/pacemaker
>
> Project Home: http://www.clusterlabs.org
> Getting started: http://www.clusterlabs.org/doc/Cluster_from_Scratch.pdf
> Bugs: http://bugs.clusterlabs.org