Fedora Hangs on PE 600SC

[Hugh McGuirk]

I have since found that running the 2.6.0-test9 kernel, the hang problem
seems to have gone away.

I would like isolate what in the 2.4.22 kernel caused the hang, and if
there is a bug, help fix it.

Anyone got any ideas on a way to isolate or at least narrow down what
went wrong below...

Thanks,

Hugh.


On Wed, 2003-11-19 at 12:05, Hugh McGuirk wrote:
> Am testing Fedora on a Dell PE 600SC server.
> 
> The machine seems to consistantly hang, when i leave it running for
> maybe a day or day and a half.
> 
> I saw some other messages on this list relating to "re-exec init", but i
> do not see that in /var/log/messages. I see no problems there at all.
> 
> I come back to find it with the screen-saver still displaying ( but not
> running), and the Caps-lock and Scroll-lock Leds flashing.
> It is possible to manually turn the Num-lock led on and off, or press
> the reset button. I can ping it ( i think) but thats about it.
> There is nothing of interest in /var/log/messages
> 
> I thought the Ram chips might be bad, so i tested with MemTest86, and
> indeed that did give me errors on upper boundry of the installed RAM.
> Testing with Dell's own memory tester showed no problems.
> 
> So i thought the Kernel needed to be manually told to use just less than
> the actual amount of memory installed. It accepted the "mem=" kernel
> parameters i gave it, but the problem still happens.
> 
> I tried different memory, from another 600SC, but same problem.
> 
> I upgraded the BIOS, but same problem.
> 
> RedHat 9 is Dell pre-installed on another 600SC machine we have here, so
> i assume RedHat 9 is ok. We have yet to have it turned on for any length
> of time either, so im not completely sure there.
> 
> I'm about to try Mandrake 9.2 on the problem machine here, just to see
> if it has the same problem.
> 
> Is there any way to have much more verbose output from the kernel, to
> trace through or track down where things stop.
> 
> Is there some sort of debug kernel that might help, or tools to
> continually monitor what process has control of the CPU, although i
> suppose that switches far too fast to be able to log it.
> 
> Anyone else see this kind of thing?
> 
> Thanks,