RE: I have currently got a server that is running 99% cpu on one of my servers

["Seok Chan LEE" <chan blurryeyes com>]

Hmm..

My servers has gone to 99.9% cpu usage too. but they work
fine now with some work-arounds. Five RH7.2 + TUX servers never
crashed for last 3 weeks. 

My servers have
- Intel 8255x-based NICs
- Dual 1Ghz CPUs on STL2 boards

I've tried with the following workarounds. I din't find what's the problem
or what's the right work-around. Sorry, I'm lazy.

#1, Upgrade your BMC Firmware to 1.19 

If you're using STL2 board and the BMC firmware is older than 1.19,
you must upgrade it to the latest. Old BMC firmware has a bug
with handling WOL packets. Sorry, I don't know why this is
necessary. For more information, please see "#19 TCO packet
Processing Issue on Intel STL2 Server Boards" of

  ftp://download.intel.com/support/motherboards/server/stl2/spec_update.pdf 
  http://support.intel.com/support/motherboards/server/stl2/latest_bios.htm
  http://downloadfinder.intel.com/scripts-df/Detail_Desc.asp?ProductID=464&DwnldID=2829

#2. Turn off "WOL (Wakeup-On-Lan)" feature of Intel NIC

If you're using Intel 8255x-based NIC, you had better turn off
WOL. Run "brow.exe -all -be -wold -lwsd"

http://downloadfinder.intel.com/scripts-df/download.asp?url=/1305/eng/100PUTIL.EXE

#3. Use "e100" from intel, instead of "eepro100"

I've read some articles that eepro100 doesn't handle heavy traffic.
You may obtain Intel-supported "e100" driver. I use 1.8.37 Linux version.

http://appsr.intel.com/scripts-df/Detail_Desc.asp?ProductID=60&DwnldID=2896 

	unix$ cat /etc/modules.conf
	alias ...
	alias eth0 e100 RxDescriptors=369 TxDescriptors=512

#4. Fix some typo of tux source

I've made a patch for tux. (Sorry, I don't know this is right thing -_-)
  http://fischer.blurryeyes.com/~chan/tux/linux-2.4.9-tux2-fixes2.patch

If you want to try my kernel rpms, you may download
  http://fischer.blurryeyes.com/~chan/tux/kernel-2.4.9-31sayclub.src.rpm
  http://fischer.blurryeyes.com/~chan/tux/kernel-smp-2.4.9-31sayclub.i686.rpm

Before applying 2.4.9-31smp kernel, you had better run redhat "up2date"
to upgrade tux userspace rpm to 2.2.0-1.

You may also want e100.o (binary one) for the above kernel rpm
  http://fischer.blurryeyes.com/~chan/tux/intel-driver.tar

Un-tar it under /lib/modules/2.4.9-31smp/kernel/drivers/addon/e100.

> -----Original Message-----
> From: tux-list-admin@redhat.com 
> [mailto:tux-list-admin@redhat.com]On Behalf Of Alex Kramarov
> Sent: Sunday, March 24, 2002 2:55 PM
> To: mingo@elte.hu; tux-list@redhat.com
> Subject: I have currently got a server that is running 99% 
> cpu on one of my servers
> 
> 
> there is a debug log on 212.179.48.90/messages .
> 
> Mingo, please contact me if you want to take a look into that 
> machine (or if
> you don't - i don't want to let this machine go like this for 
> a long time).
> I do not have kernel profiling enabled on it. tux compiled as 
> a module into
> vanilla 2.4.17 kernel.
> 
> Alex.
> 
> 
> 
> _______________________________________________
> tux-list mailing list
> tux-list@redhat.com
> https://listman.redhat.com/mailman/listinfo/tux-list
>