[Linux-cluster] gfs problem on FC6

[Robert Peterson]

Ashish Varman wrote:
> Hello all
>
> I am using a gfs filesystem on FC6 with a two node cluster.  Since I 
> needed gfs_grow, I checked out "cluster" from sources.redhat.com with 
> the tag "RHEL50".  I first tested gfs functionality without cluster ( 
> -p lock_nolock ).  I then made a two node cluster and formatted the 
> volume again with (-p lock_dlm).  On one node, the filesystem can be 
> mounted but on the other it fails.  gfs_fsck runs and gives no errors. 
> The errors from the mount command and in /var/log/messages are as follows
>
> [root at clnode31 ~]# mount /dev/mapper/ExVolGroup-ExLogVol /mnt/tmp/ -t gfs
> mount: wrong fs type, bad option, bad superblock on 
> /dev/mapper/ExVolGroup-ExLogVol,
>       missing codepage or other error
>       In some cases useful info is found in syslog - try
>       dmesg | tail  or so
>
> [root at clnode31 ~]# tail /var/log/messages
> Jan 20 18:45:45 clnode31 kernel: Trying to join cluster "lock_dlm", 
> "alpha:third"
> Jan 20 18:45:45 clnode31 kernel: dlm: third: recover 1
> Jan 20 18:45:45 clnode31 kernel: dlm: third: add member 1
> Jan 20 18:45:45 clnode31 kernel: dlm: third: total members 1 error 0
> Jan 20 18:45:45 clnode31 kernel: dlm: third: dlm_recover_directory
> Jan 20 18:45:45 clnode31 kernel: dlm: third: dlm_recover_directory 0 
> entries
> Jan 20 18:45:45 clnode31 kernel: dlm: third: recover 1 done: 0 ms
> Jan 20 18:45:45 clnode31 kernel: Joined cluster. Now mounting FS...
> Jan 20 18:45:45 clnode31 kernel: GFS: fsid=alpha:third.4294967295: 
> can't mount journal #4294967295
> Jan 20 18:45:45 clnode31 kernel: GFS: fsid=alpha:third.4294967295: 
> there are only 8 journals (0 - 7)
>
> What am I doing wrong ????
>
> Thanks in advance.
>
> Yours Sincerely
> Ashish Varman
Hi Ashish,

Hard to say offhand what is going on here.  I recommend checking the 
following
(off the top of my head; no particular order).

1. Make sure both nodes have /sbin/mount.gfs and /sbin/mount.gfs2
    It's okay to have /sbin/mount.gfs be a symlink to /sbin/mount.gfs2
2. Make sure both nodes are running the same version of the cluster software
    and GFS (i.e. via rpm -q).
3. Make sure the clustered bit is "on" for your vg.  Use vgs.  (Assuming,
    of course, that you're using lvm).  The vgs command should show
    something like "wz--nc", and the "c" at the end means it's clustered.
4. Make sure your /etc/cluster/cluster.conf file is the same on both nodes.
5. Make sure your gfs file system was made (gfs_mkfs) with that cluster 
name.
    Use you can check this with gfs_tool sb /dev/your_vg/your_lv table
6. If everything so far checks out, do the command "group_tool -v" to make
    sure both nodes are properly joined to the appropriate fence group and
    dlm group.
7. If everything looks good in #6, do the command "group_tool dump gfs"
    to see if there are any weird messages from the gfs_controld daemon
    regarding the mount that failed.

Regards,

Bob Peterson
Red Hat Cluster Suite