[cobalt-users] [RaQ-4i] Harddisk failure or something else

["Jelmer Jellema" <cobalt@xxxxxxxxxxxxxxx>]

Hi all,

Yesterday something weird happened. I was working on some on line tools on
the cobalt, and for testing purposes I created some sites and deleted them
again (in the GUI). Next thing you know:

- The "sitemanage" part of the interface was unavailable (File Not Found)
for every site
- The system stopped accepting e-mail for any user on the system

What happened:
- The /usr/admserv/html/.cobalt/siteManage directory (where the pages for
managing each site are created when you create the site) had disappeared
- In /etc/mail/virtusertable all the domain routes (mapping
someone@xxxxxxxxxxxx to someone@xxxxxxxxxxxxxxxx) had gone. Allthough I had
made a change to this file, I had done this days before and nicely below the
indicated line.

After getting back the siteManage stuff from backup, I found that all the
mail-domain aliases were gone in the GUI (and the database I guess) as well.
I had to retype them with a backup of the virtusertable as a guide.
meta-verify thought everything was great!

I did a e2fsck -n on all partitions, and found:

**/: nothing wrong
**/var:
Block bitmap
differences:  -34362 -34363 -34364 -34365 -34366 -34367 -37527 -37528 -37529
 -37530 -37531 -37538
Fix? no
**/home:
Block bitmap
differences:  -1780114 -1780115 -1780116 -1780117 -1780118 -1780127 -1780128
 -1780129 -1780130 -1780131 -1780132 -1780133 -1780134 -17832
Fix? no

Free blocks count wrong for group #217 (4912, counted=4904).
Fix? no

Free blocks count wrong (17496621, counted=17496613).
Fix? no


After this I did a shutdown -rF,  to force a full check. The system stayed
off line a while and when it came back, I could find nothing special in
/var/log/*, and a new e2fsck -n gave the same messages as above.

After this, I might have done the wrong thing: I ran e2fsck -n -c, which
checks for bad blocks and then writes the bad block list. I checked this on
a mounted system (/home), thought the -n made it quite safe (problem is as
allways distance: not easy getting to the system and boot it singe-user).

This did not find anything (it did'n't mention anything at least).

In the evening the logs began. The entries are known to this list, allthough
I could not find a solution in the archives:

To different types:
Some of these:
Apr  9 18:56:59 server01 kernel: EXT2-fs warning (device ide0(3,4)):
ext2_free_blocks: bit already cleared for block 2531693
Apr  9 18:56:59 server01 kernel: EXT2-fs warning (device ide0(3,4)):
ext2_free_blocks: bit already cleared for block 2531692

and a lot of those:
Apr 10 05:00:41 server01 kernel: attempt to access beyond end of device
Apr 10 05:00:41 server01 kernel: 03:04: rw=0, want=811278382, limit=18415656
Apr 10 05:00:41 server01 kernel: dev 03:04 blksize=1024 blocknr=811278381
sector=1622556762 size=1024 count=1
Apr 10 05:00:41 server01 kernel: attempt to access beyond end of device
Apr 10 05:00:41 server01 kernel: 03:04: rw=0, want=1883320122,
limit=18415656
Apr 10 05:00:41 server01 kernel: dev 03:04 blksize=1024 blocknr=1883320121
sector=-528327054 size=1024 count=1

(these seemed to come tonight a the time my backup began to run, using tar
to backup to NFS).

My first question is:
Is this: failing disk, failing kernel, failing controller, failing memory or
failing Jelmer (that's me).

What should I do. I'm particulary worried because of the disappearing files
and sudden changed in the system psql database...

Any help is greatly appreciated,
Jelmer
-----------------------------------------------------------------
Jelmer Jellema - Spin in het Web
www.spininhetweb.nl
Spin in het Web: Alle Touwtjes In Handen
-----------------------------------------------------------------

Spin in het Web is de producent van:
www.visinhetnet.nl: Niet Het Laatste Nieuws