Ocfs2 users - May 2006 - Dinode signature problem

Dear All:

I am building an Oracle 10G R2 cluster as outlined in Jeffrey Hunter's
article 'Build your own Oracle 10g Release 2 Cluster on Linux and
Firewire'. However I'm experiencing problems with the OCFS2 filesystem
on the shared drive. After the system boots up and the device is mounted to the
directory /u02/oradata/orcl, the kernel displays the error message "OCFS2:
ERROR (device sda1): ocfs2_meta_lock_update: Dinode # 17 has bad signature
KNODE01  File system is now read only due to the potential of on-disk
corruption. Please run fsck.ocfs2 once the file system is unmounted.".
After running fsck.ocfs2 to fix the problem, the same message appears again. I
did notice this happens to dinode # 17 and 23 only. I upgraded ocfs2 to the
latest version, 1.2.1.1, but the problem still appears. I also checked the
Maxtor drive using their Powermax disk utility and it reported the drive is
clean. Any help with this problem is appreciated.

The shared drive (Maxtor DiamonMax 80G ATA/133) is in a Nspire NSP-ED352C
external drive enclosure with Firewire (Oxford 911 chipset) and USB interfaces.

System specs:
CentOS 4.2  
ocfs2 ver 1.2.1.1
Firewire cards with VIA chipsets 

Thank you very much.

Sincerely,

Valan Franklin, OCP
(905) 634 6137 Home
v.franklin at sympatico.ca
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
http://oss.oracle.com/pipermail/ocfs2-users/attachments/20060502/b39c1ef2/attachment.html

Hmmm... I've seen this issue once before and on a similar setup. As in, 
firewire
(Oxford 911 chipset). It literally is a one-bit corruption. The 
signature INODE01
becomes KNODE01. As far as I remember we never could duplicate it. As it was
only on one firewire setup and not on any fiber/iscsi/xen setups, we did 
not spend
much time on it.

First, backup:
dd if=/dev/sdX of=/tmp/backup bs=<blocksize> count=1000

To fix:
# dd if=/dev/sdX  of=blk17 count=1 skip=17 bs=<blocksize>
# dd if=/dev/sdX  of=blk23 count=1 skip=23 bs=<blocksize>

Use your favorite bin editor and change the first K to an I.

# dd of=/dev/sdX  if=blk17 count=1 seek=17 bs=<blocksize>
# dd of=/dev/sdX  if=blk23 count=1 seek=23 bs=<blocksize>

The fs should be umounted during this escapade.

Also, maybe you should look at iscsi for your shared storage needs.

Valan Franklin wrote:
> Dear All:
>  
> I am building an Oracle 10G R2 cluster as outlined in Jeffrey Hunter's 
> article 'Build your own Oracle 10g Release 2 Cluster on Linux and 
> Firewire'. However I'm experiencing problems with the OCFS2
filesystem
> on the shared drive. After the system boots up and the device is 
> mounted to the directory /u02/oradata/orcl, the kernel displays the 
> error message "OCFS2: ERROR (device sda1): ocfs2_meta_lock_update: 
> Dinode # 17 has bad signature KNODE01  File system is now read only 
> due to the potential of on-disk corruption. Please run fsck.ocfs2 once 
> the file system is unmounted.". After running fsck.ocfs2 to fix the 
> problem, the same message appears again. I did notice this happens to 
> dinode # 17 and 23 only. I upgraded ocfs2 to the latest version, 
> 1.2.1.1, but the problem still appears. I also checked the Maxtor 
> drive using their Powermax disk utility and it reported the drive is 
> clean. Any help with this problem is appreciated.
>  
> The shared drive (Maxtor DiamonMax 80G ATA/133) is in a Nspire 
> NSP-ED352C external drive enclosure with Firewire (Oxford 911 chipset) 
> and USB interfaces.
>  
> System specs:
> CentOS 4.2 
> ocfs2 ver 1.2.1.1
> Firewire cards with VIA chipsets
>  
> Thank you very much.
>  
> Sincerely,
>  
> Valan Franklin, OCP
> (905) 634 6137 Home
> v.franklin at sympatico.ca <mailto:v.franklin at sympatico.ca>
> ------------------------------------------------------------------------
>
> _______________________________________________
> Ocfs2-users mailing list
> Ocfs2-users at oss.oracle.com
> http://oss.oracle.com/mailman/listinfo/ocfs2-users
>