[plug] [OT-ish?] Need help diagnosing unstable computer

[indy at THE-TECH.MIT.EDU]

I may have missed this in the original message,
but have you ruled out an X related issue?

i.e. does it only crash from X, or does it crash from the 
console also?


Indy


On Thu, Aug 07, 2003 at 08:26:29PM +0800, Cameron Patrick wrote:
> On Thu, Aug 07, 2003 at 08:05:53PM +0800, James Devenish wrote:
> | In message <20030807114232.GA603 at erdos.home>
> | on Thu, Aug 07, 2003 at 07:42:32PM +0800, Cameron Patrick wrote:
> | > I'm reasonably sure that this is a hardware issue - the machine didn't
> | > respond to Ctrl-Alt-{Backspace,Delete} or SysRq-B, and the keyboard
> | > lights weren't doing the kernel panic thing.  The crashes seem to be
> | > occurring more-or-less randomly.
> | 
> | I don't suppose any useful log messages managed to make their way to the
> | disk before the hangs?
> 
> Not really :(  Last few lines of /var/log/kern.log before a crash and
> the first few afterwards:
> 
> Aug  7 20:13:58 erdos kernel: hub.c: 3 ports detected
> Aug  7 20:13:58 erdos kernel: blk: queue c03fa260, I/O limit 4095Mb (mask 0xffffffff)
> Aug  7 20:13:58 erdos kernel: hde: CHECK for good STATUS
> Aug  7 20:16:38 erdos kernel: klogd 1.4.1#11, log source = /proc/kmsg started.
> Aug  7 20:16:38 erdos kernel: Inspecting /boot/System.map-2.4.21-cjp-erdos-2
> Aug  7 20:16:38 erdos kernel: Loaded 21250 symbols from /boot/System.map-2.4.21-cjp-erdos-2.
> Aug  7 20:16:38 erdos kernel: Symbols match kernel version 2.4.21.
> Aug  7 20:16:38 erdos kernel: Loaded 206 symbols from 5 modules.
> Aug  7 20:16:38 erdos kernel: Linux version 2.4.21-cjp-erdos-2 (cameron at erdos) (gcc version 3.3.1 20030722 (Debian prerelease)) #1 Sun Aug 3 12:49:25 WST 2003
> 
> There's nothing remotely interesting in daemon.log either.
> 
> (Incidentally it died twice in the last 15 minutes - once while I was
> writing a reply and then again while loading up KDE. Aieee!)
> 
> Re. what Nema said: this, to me, doesn't smell like a memory issue, and
> as I mentioned I've run memtest86 on the machine for a short time to no
> avail.  I may leave memtest running over night and/or while I'm at Murdoch
> tomorrow, just in case...
> 
> Cameron.

-- 
Indranath Neogy
<indy at the-tech.mit.edu>