[Ocfs2-users] Weird lock

[Nuno Fernandes]

On Thursday 10 April 2008 15:30:31 Erik Terpstra wrote:
> Hi Nuno,
>
> Nuno Fernandes wrote:
> > That's because thay are waiting for something that the kernel will
> > provide (probably a lock). Please do:
> >
> > echo t > /proc/sysrq-trigger
> >
> > Check /var/log/messages and paste it here.
>
> Thanks I will provide the information next time it locks.
>
> > The "another" node is also with high load? Is there any program running
> > at 100% CPU?
>
> We haven't checked if the other node has a high load, but that is
> unlikely, it seems like rebooting any node will fix it.
> I will check it anyway next time.
>
> When the system locks, the 'load' keeps rising but there is no
> particular process that seems to be the cause of it.
>
> > If yes, i think that the first node is waiting for a lock that the
> > "another" node has. Until that lock is released all processes remain in
> > "D" state and the load keeps on rising.
> >
> > Also do:
> >
> > echo t > /proc/sysrq-trigger
> >
> > Check /var/log/messages and paste it here.
> >
> > Do a:
> >
> > ps fax
> >
> > also and paste it here.
>
> Thanks, I will.
>
> > I also have that problem and i'm still trying to figure it out.
>
> Does rebooting an arbitrary node also fix your situation?
No it doesn't. I have to reboot the node that has the spinning process.

I'm upgrading to ocfs1.2.8 as i have 1.2.5 to see if it gets better :)

Regards,
Nuno Fernandes
>
> Regards,
>
> Erik.