Subject: Re: multiprocessor i386 1.6ZC system crash
To: Lubos Vrbka <lubos.vrbka@jh-inst.cas.cz>
From: Scott Ellis <scotte@intrepid.dnsalias.com>
List: current-users
Date: 10/13/2003 11:16:04
On Mon, Oct 13, 2003 at 07:52:30PM +0200, Lubos Vrbka wrote:
> hi guys,
> 
> i've got -current running on 2 processor i386 (p3 800) machine. from 
> time to time the whole machine freezes and i can do nothing with it - no 
> keyboard response, no network response... the only thing i can do is to 
> reset... (hapenned to me twice today :o(, happens randomly ~once per week).
> 
> i inspected logs but found nothing. seems that the last thing my 
> computer was doing before crash was restarting syslogd.
> from my observations, the hangups happen only when i i'm working (i'm 
> running X from 1.6.1 stable) on the machine, mainly when accesing 
> harddrive. it's interesting that i ran several few hours long 
> calculations that use harddrive quite a lot, but no crash...
> 

I wanted to add that I've seen similar hard-lockups (can't even drop into
DDB via the serial console, it's so frozen).  I've also seen it
"almost" lock up, where the IDE activity light would be on solid, yet there
was no IDE activity.  When my machine freezes, it usually has the IDE light
lit solid (makes a visual "is it locked up" check easy!). :-)

I can't correlate the lockups to specific tasks, as doing lots of
SMB traffic, and many concurrent builds doesn't seem to tickle the lockup.

My machine is a dual P3-600EB with 512M of RAM.  While this is a pretty
generic "me too" report, I wanted you to know that you're not alone. :-)

(And yes, running in !MP mode makes a stable machine, although my box
currently seems to run ~15 days between lockups.)

-- 
   //////////////////////////////////////////////////////////////////////
  //    Scott Ellis     //             scotte@warped.com              //
 //////////////////////////////////////////////////////////////////////
// WARNING: This signature warps  time and space in its vicinity    //