Dan,
That is exactly what the problem was, the modules.conf file was being
munged in the systemimager script. When Roger was helping us push out
the initial machines using systemimager we modified the bottom of the
script to do some specific setup for our environment. The modules.conf
file was tweaked allowing us to re-associate the ethernet interfaces
with the appropriate modules. That same piece of code was left in the
Xw8000 machine script. We didn't notice any problems with RedHat 7.3 but
apparently RedHat 9 gets confused or does something different. Thanks
for your help, this was a real head scratcher.
Regards
-Richard
On Tue, 2003-09-30 at 22:00, dann frazier wrote:
> On Tue, Sep 30, 2003 at 01:18:06PM -0600, Richard N. Cleary wrote:
> > Here is what I have found so far. I was able to generate an error on the
> > console of the workstation that was freezing:
> >
> > Call Trace: [<c010c038>] wait_on_irq [kernel] 0xf8 (0xf590feac)
> >
> > There was more but I thought this was most interesting.
> >
> > What I have done on the machine that is freezing is turn off irqbalance
> > and it doesn't freeze but it is slow and every time I move the mouse it
> > takes a moment to wake up. The kernel I am booting is 2.4.20-20.9smp, I
> > tried booting the 2.4.20-20.9 with the irqbalance off and all appeared
> > ok.
> >
> > What is systemimager potentially doing with the irq's? The machine that
> > is the golden client is booting the 2.4.20-20.9smp kernel with
> > irqbalance on and it doesn't exibit any of this behavior. Again the
> > hardware is essentially identical.
>
> the systemimager part of sis only copies files - it knows nothing about
> the kernel, irqs, etc. after its done, it calls systemconfigurator, which
> under redhat will try to generate a new initrd. that is the only place
> where i can think of that the kernel could be configured differently.
> however, sc should be doing exactly what redhat does on a kernel install.
> i'd compare the modules loaded on a machine in this state vs. your original,
> and try to unload modules to see if something makes it go away.
>
> you can also compare the rate of increase in the fields in /proc/interrupts
> between the two to see if something is clearly interrupting the cpu too
> much.
--
Richard N. Cleary
Sandia National Laboratories
Dept. 9324 Infrastructure Computing Systems
Email: <[EMAIL PROTECTED]>
Phone: 505.845.7836
-------------------------------------------------------
This sf.net email is sponsored by:ThinkGeek
Welcome to geek heaven.
http://thinkgeek.com/sf
_______________________________________________
Sisuite-users mailing list
[EMAIL PROTECTED]
https://lists.sourceforge.net/lists/listinfo/sisuite-users
