[CentOS] CentOS 4.3 occasionally locking up accessing IDE drive

Sat Apr 1 19:47:37 UTC 2006
William L. Maltby <BillsCentOS at triad.rr.com>

On Sat, 2006-04-01 at 11:16 -0800, Bart Schaefer wrote:
> For those who haven't seen my several previous postings about problems
> with this (now not quite so) new PC, I have an ASUS P5N32-SLI Deluxe
> motherboard.  The boot drive and primary filesystems are on an SATA
> disk and I'm having no problem with that.  However, I recently plugged
> in a couple of IDE drives salvaged from my old PCs and I'm running
> into trouble with one of those.
> 
> The drive in question is a 20GB Maxtor 92049U6.  It had an old RH5.2
> ext2 filesystem on it when I first plugged it in, from which I tried
> to recover some data to back up to CD.  Mostly this worked, but I
> started encountering read errors accessing some files so I unmounted
> the partition and started a fsck on it.  At some point during the fsck
> (I was off doing something else on another workspace at the time), the
> system locked up hard, leaving the disk activity LED lit.  I had to
> reset the PC.
> 
> So at that point I booted single-user and ran the fsck from there.  It
> completed successfully after fixing a number of problems.  I continued
> into multi-user mode, finished doing my backups, repartitioned the
> drive, and started "mkfs -t ext3 -c" on the larger partition, to check
> for bad blocks.  Again at some point part way through the mkfs, the
> system locked up.
> 
> Back to single user mode, run the "mkfs", everything finishes fine. 
> Back to multi-user mode, start to copy some large files onto the
> drive.  MD5 sums fail to match for some of the copied files. 
> Unmounted and started up "fsck -y".  This succeeded, after fixing a
> number of errors, so (at this point just as a test case) I re-copied
> the files with bad MD5s.  Some of these came through OK this time,
> others still did not.  I decided perhaps this meant there were still
> bad blocks on the drive that a read-only test was not finding.
> 
> You'd think I'd have learned, but encouraged by the success of the
> previous fsck I optimistically started up another "fsck -c -c -y" on
> the suspect partition, and this time I waited around to watch it. 
> About 1.6GB into the 16GB partition, the system locked up again.
> 
> This time I booted into a hard disk diagnostic program instead of into
> CentOS.  After running overnight last night, a non-destructive
> read-write surface-scan reported no problems with the drive.  This
> leads me to suspect that the problem is with linux, but I don't know
> how to proceed with diagnosing it.  Suggestions would be appreciated.

Re the overnight diag, are environmental conditions similar to when you
encounter problems? Temp, power "brown out", etc? If *not*, try the diag
when conditions are similar to when you have the problem. Long shot, but
you've obviously gotten o the point of needing a long rifle.

Secondly, are your current HD configurations consistent with what is
actually on the drive?

"sfdisk -l /dev/hdXXX"

and then look at your BIOS settings for it. If the BIOS where the disk
was originally set up assigned different params than the current BIOS
(assuming you did auto-detect rather than set up manually) could this be
involved? I don't *think* BIOS settings are actually used in current
Linuxes, but I could be wrong. If sfdisk and/or the BIOS show different
params than the old BIOS, maybe a manual setup of the HD params will fix
your problem?

Since you have an (apparent) inconsistent behavior, how about cable
integrity, connectors, power etc.? Poorly seated or worn connector could
be very sensitive to temp changes and vibration. Have you visually
inspected all this, especially the power? Have you put volt meters on
the +5/+12 and their respective grounds when the system is under load to
see if maybe your PS is too weak?

HTH
Bill

P.S. Tried different IDE cable/power connectors?

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 189 bytes
Desc: This is a digitally signed message part
URL: <http://lists.centos.org/pipermail/centos/attachments/20060401/44312f72/attachment-0004.sig>