[Pacemaker] small gfs question

[Robert Lindgren]

On Tue, Jun 22, 2010 at 4:05 PM, Robert Lindgren
<robert.lindgren at gmail.com>wrote:

> Hi All,
>
> Sorry if this topic has been up before, I'm new to this list.
>
> I have a problem with gfs2_quotad turning up in D
> state, uninterpretable sleep, when I set one of my nodes in standby. Hence
> VirtualDomain resource agent stops working, since libvirt fails to read from
> gfs (drbd primary/primary). I'm running Ubuntu Lucid with the gfs tools from
> Cluster-stack ppa.
>
> If one one node is started gfs doesn't behave like this, and when one host
> is lost due to standby or powercord yank.
>
> Any hints would be appreciated.
>

Last things in dmesg are:

[86392.002282] block drbd0: conn( Unconnected -> WFConnection )
[86395.120629] dlm: closing connection to node 34212362
[86395.162508] GFS2: fsid=pcmk:pcmk.0: jid=1: Trying to acquire journal
lock...
[86520.420036] INFO: task kslowd000:2290 blocked for more than 120 seconds.

After that I see that gfs doesn't respond:
root at sugadaddy:/var/log# ps -efl | grep D
F S UID PID PPID C PRI NI ADDR SZ WCHAN STIME TTY TIME CMD
1 D root 2290 2 0 75 -5 - 0 dlm_lo Jun22 ? 00:00:00 [kslowd000]
1 D root 2310 2 0 80 0 - 0 dlm_lo Jun22 ? 00:00:00 [gfs2_quotad]
0 D root 18713 1147 0 80 0 - 4658 dlm_lo 15:54 ? 00:00:00
/usr/lib/libvirt/virt-aa-helper -r -u
libvirt-c0b6fc07-7195-4e31-7ce7-e12d5b71bdee

Looks to be the gfs2_quotad stalling but since parentpid 2 is kinda hard to
see if it's really quotad stalling or if it's waiting for the journal lock.

Not sure why I don't get the journal lock...
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.clusterlabs.org/pipermail/pacemaker/attachments/20100623/a9ca7eb8/attachment-0001.html>