CentOS - Oct 2009 - EDAC Kernel Panic 2.6.9-78 and above

I've got a production system running CentOS 4 that was rock solid
until I upgraded from 2.6.9-55 to 2.6.9-78.0.13 (now running
2.6.9-89.0.11). The system now crashes intermittently after a few
weeks. I finally caught the panic message :

EDAC MC0: INTERNAL ERROR: channel-b out of range (4 >= 4)
Kernel panic - not syncing: MC0: Uncorrected Error

Looking at the kernel changelog, I see that EDAC support was added
for the Intel 5000 chipset in 2.6.9-68.20.EL which this server runs.

I'm trying to determine if this is a potential memory issue, or is
this related to some other hardware item. Also considering disabling
EDAC in the kernel (is "noedac" a valid option?) as a last resort. I
will run memtest86+ on the server as soon as possible to check the
memory, just formulating my game plan if it's something else.

Thoughts?

Chris

Chris Miller wrote:
> Thoughts?
Check your bios/system event log for any indication that it
is logging memory errors? Most modern server class motherboards
(past 5 years) do this, though not always reliably.

I've also had trouble with memtest86 myself, I prefer to run
ctcs:

http://sourceforge.net/projects/va-ctcs/

The software is really old and is picky what you build it on,
if I recall right I could only get it to build on RHEL/CentOS 4
not 5 (though the binaries work fine on 5).

It does a good torture test which in my experience can find
problems faster than memtest86(which can take days).

nate

Chris,
> I've got a production system running CentOS 4 that was rock solid
> until I upgraded from 2.6.9-55 to 2.6.9-78.0.13 (now running
> 2.6.9-89.0.11). The system now crashes intermittently after a few
> weeks. I finally caught the panic message :
> EDAC MC0: INTERNAL ERROR: channel-b out of range (4 >= 4)
> Kernel panic - not syncing: MC0: Uncorrected Error
> Looking at the kernel changelog, I see that EDAC support was added
> for the Intel 5000 chipset in 2.6.9-68.20.EL which this server runs.
Same issue here with a machine running centos 5.3. The problem began
with a kernel update that introduced the 5000 chipset. See the thread
"RAM errors after kernel-update" for more details. I couldn't
solve
the problem yet, but because the machine crashes every two days with
this kernel, I had to boot an earlier kernel without chipset support.

> I'm trying to determine if this is a potential memory issue, or is
> this related to some other hardware item. Also considering disabling
> EDAC in the kernel (is "noedac" a valid option?) as a last
resort. I
> will run memtest86+ on the server as soon as possible to check the
> memory, just formulating my game plan if it's something else.
Don't use the memtest86+ version that comes with the centos ISO. There is
a much newer version available from the authors website. Only the new
version identifies the chipset correctly.
-- 
Mit freundlichen Gr??en
Michael Schumacher
mailto:michael.schumacher at pamas.de