On Tue, Aug 17, 2010 at 3:46 PM, Alain.Moulle <[email protected]> wrote:
> Hi,
>
> OK I finally got clone-o2cb started on both nodes and I manually can mount
> all my ocfs2 FS on both sides, but just to see, I launch a test which on
> both
> nodes in parallel which loops on the mount of all my ocfs2 FS , and then
> umount all of them.
> It works on one node, and as soon as I start the test on the second
> node, it works fine
> a few seconds and then returns an error on one FS (not necessarily
> always the same) :
> mount.ocfs2: Invalid argument while mounting /dev/sdc on /data. Check
> 'dmesg' for more information on this error.
>
> so this seems not robust at all ... because I thought ocfs2 was to
> enable these types of manipulations besides IO on
> both sides of course.
>
> So my question about Pacemaker and OCFS2 is now :
> I don't know the history arount pcmk use with Pacemaker,
> but why could we not write a RA ocf script which works with the Stack
> o2cb ?
> I mean, having a /etc/ocfs2/cluster.conf well configured, and an ocf script
> which start , stop and monitor the service o2cb and the drivers needed etc.
>
> Thanks a lot if someone could give an answer about this.

Because its having different components with a different idea of what
the current membership is leads to a whole world of hurt.
You _really_ do not want to go there.
_______________________________________________
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