[Linux-cluster] gfs2 mount hangs

[David Teigland]

On Tue, Aug 18, 2009 at 10:32:54AM +0800, Wengang Wang wrote:
> there is a line with "...error -1 2", but seems the line is out during 
> the second(hang) mount.
> 
> 2) for node desk(the first mount is on this node).
> [root at desk ~]# gfs_control dump
> 1250589068 logging mode 3 syslog f 160 p 6 logfile p 6 
> /var/log/cluster/gfs_controld.log
> 1250589068 gfs_controld 3.0.0 started
> 1250589068 logging mode 3 syslog f 160 p 6 logfile p 6 
> /var/log/cluster/gfs_controld.log
> 1250589068 group_mode 3 compat 0
> 1250589068 setup_cpg 13
> 1250589068 run protocol from nodeid 1
> 1250589068 daemon run 1.1.1 max 1.1.1 kernel run 1.1.1 max 1.1.1
> 1250590628 uevent add gfs2 /fs/gfs2/testgfs2:1
> 1250590628 uevent change gfs2 /fs/gfs2/testgfs2:1
> 1250590628 recovery_uevent mg not found 1
> 1250590628 uevent change gfs2 /fs/gfs2/testgfs2:1
> 1250590628 recovery_uevent mg not found 1
> 1250590628 uevent change gfs2 /fs/gfs2/testgfs2:1
> 1250590628 recovery_uevent mg not found 1
> 1250590628 uevent online gfs2 /fs/gfs2/testgfs2:1

That's the problem, there are a bunch of steps missing that come from
mount.gfs2.  My guess is that you're missing /sbin/mount.gfs2 on this node.

It also seems that gfs2 no longer properly handles this error condition
(missing/empty hostdata string).  In the past you would have gotten a nice
message in /var/log/messages explaining that you were missing mount.gfs2.

Dave