[Pacemaker] kernel BUG at fs/dlm/lock.c:242! after sync of GFS2 (2 node - active/active)

[Vladislav Bogdanov]

13.09.2010 16:21, Andrew Beekhof wrote:
> On Wed, Sep 8, 2010 at 6:00 AM, Vladislav Bogdanov <bubble at hoster-ok.com> wrote:
[...]
>> Could you paste output from drbd-overview? drbd device should be in
>> Primary state on node where you issue mount.
>> And...
>> Did you start corosync or openais? Later is required instead of "plain"
>> corosync to support GFS2/OCFS2.
>>
>> Andrew, shouldn't openais requirement for GFS2 be mentioned in docs?
> 
> Which ones though?  Its not strictly Pacemaker's job to explain the
> requirements of the resources you're running.
> I'm pretty sure clusters from scratch mentions it though, as does
> http://www.clusterlabs.org/wiki/FAQ#What_is_the_Project.27s_Relationship_with_OpenAIS.3F

Aha, missed that in bundled docs an pacemaker-explained.

> 
>> GFS2 needs CPKT, and cries about it when started under plain corosync.
>> Now I just start openais instead of corosync, but current openais
>> movement is to remove that initscript.
>>
>> BTW dlm seems to be really broken on F13 (and F12). At least for me. I
>> constantly get kernel panic with that kernels.
> 
> Just downgrading the kernel was enough?

Yes. No fancy staff like cgroups, but it at least works. Unlike F13 and
F12 kernels. Look at link below, I described what I get there.
Didn't try EL6-beta kernel yet.

> 
>> http://lists.linbit.com/pipermail/drbd-user/2010-August/014596.html


>>
>> Only downgrade to F11 kernel helped me to get it all work. I have no
>> power to investigate further but can provide more information if someone
>> is willing to fix it.
>>
>> Best,
>> Vladislav
>>