Re: Corrupt JFS inode table?
On 08/03/2006 02:01 PM, Jeff Cleverley wrote:
Greetings,
We have a Dell 2850 running the 2.6.15-1 kernel and have created 2 JFS
file systems from the non-boot internal drives. Both JFS file system
disks come from the same raid group. The other day the system quit
responding and we ended up having to power cycle the box. When it
booted, everything looked OK, file systems marked clean, mount points
correct, etc.
Shortly after the reboot we received a call from an engineer that said
some of their data had no permissions or size. Another user reported
that a file he opened did not contain the contents of the file that
should have been there. That was for one file system. The other JFS
file system appears to have come through cleanly. We ended up having to
restore the entire first file system. Below is an example of what
happened to the files that changed. The first one is the old file that
changed after the reboot and the second one is the one that was restored.
---------- 1 meg Debian-exim 0 Jul 19 13:47 cdsinit
-rwxr-xr-x 1 meg Debian-exim 17 Jul 19 13:47 cdsinit
As you can see, the ownership and time stamp stayed the same, but file
permissions were wiped and the size was nulled out in the first one. We
unmounted the file system and probably did about 6 fsck of the file
system with various options and even rebooted the box again. Nothing
changed. Not all data in the file system changed, just random
directories and files.
We really need to find out what happened, otherwise we don't feel we can
trust this server or file system with critical data. Any help debugging
or diagnosing the problem will be greatly appreciated. If there is
another list that might be more appropriate for this problem, please let
me know.
Thanks,
Jeff
Most likely, the following does not have anything to do with
your problem because your kernel is 2.6.15, and your
filesystems are jfs; however, you might be experiencing a
serious bug affecting XFS partitions.
http://lists.debian.org/debian-user/2006/07/msg02237.html
Reply to: