[sclug] System catatonia

[Alex Butcher]

On Wed, 14 Sep 2005, James Wyper wrote:

> Every so often, my linux system (Mandrake 10.1; 2.6.8 Kernel), which I
> leave running constantly, pretty much stops responding.  I think the
> problem is related to disc i/o as while the mouse will move in KDE, and
> it still functions as a firewall / router, I can't start any new
> programs.  The system always comes back up OK when I alt-sysreq-b.
>
> It's running on 1999 vintage hardware, so this could be a hardware
> problem, but - coincidentally or not - it only started after I upgraded
> my kernel to 2.6.

Is that a kernel.org kernel, the original 10.1 release kernel, or Mandrake's
latest errata kernel for 10.1?

If either of the former options, upgrade to the latest errata kernel and see
if that fixes it.

  2.6.8 is VERY old and was quite buggy. Unless Mandrakesoft have backported
all the fixes from later kernels, I'd expect it to be kinda crappy. Further,
kernel.org kernels aren't really subject to the sort of QA that distro
kernels get and thus tend to experience unpleasant regressions from time to
time.

> I've looked in /var/log/messages, /var/log/syslog, and
> /var/log/kernel/* without finding anything (except that /messages stops
> recording anything at the time the machine hangs).
>
> Is there anywhere else I can look for after-the-fact evidence of a
> problem that I can use for diagnosis?  How can I do a thorough test of
> my (internal IDE) disk drive?

I'd expect you to get some IDE layer errors in the log if it was the disc,
but it's conceivable that may not happen.

Enabling magic SysRq and using it in the event of a hang can sometimes prove
useful.

To do a thorough test of your HDD, use 'badblocks'. There's a read-only
mode, a read-write mode, and a mode that purports to be able to do a
non-destructive write test as long as no partitions are mounted (i.e. you'd
need to run it from a boot disc of some kind).

I'd also run memtest86+ on your system.

> Thanks,
> James.

HTH,
Alex.
-- 
Alex Butcher      Brainbench MVP for Internet Security: www.brainbench.com
Bristol, UK                      Need reliable and secure network systems?
PGP/GnuPG ID:0x271fd950                         <http://www.assursys.com/>