We just got a (severely) corrupted filesystem on our webserver and it
seems to me that an inode block has been overwritten.
The system was freshly installed (not upgraded) as Red Hat 4.2 system
one month ago (so, originally using kernel 2.0.27) and upgraded to
kernel 2.0.32 just after that installation.
The kernel very frequently (since the last 12 hours or so) logs the
following messages:
Jan 16 12:55:47 simba kernel: EXT2-fs warning (device 08:06):
ext2_free_inode: bit already cleared for inode 109321
If you look at the filesystem, you see for example entries like this:
[root@simba linux]# ls -al
total 1099678947
drwxrwsr-x 3 jos xos 1024 May 7 1996 .
drwxrwsr-x 11 jos xos 1024 Jan 9 14:25 ..
?r-------- 35327 30208 65408 71419655 Dec 19 1907 ipfwadm
This looks to me like an overwritten inode table...
There are currently no indications that there is a hardware problem.
Furthermore, we had a similar problem last year on one of our other
systems (that system was running kernel 2.0.27).
It worries me even more that I found a number of similar reports with
Dejanews (one time even with the same SW configuration, RH 4.2 with
2.0.32), so it really appears to be more than an occasional problem.
Unfortunately, I didn't see any response to those messages, except
for people suggesting that it might be caused by a hardware problem.
Any useful suggestions to avoid such problems in the future?
Will 2.0.33 solve the problem?
Thanks,
-- -- Jos Vos <jos@xos.nl> -- X/OS Experts in Open Systems BV | Phone: +31 20 6938364 -- Amsterdam, The Netherlands | Fax: +31 20 6948204