I have done everything in the world I can think of but My dual Xeon but
still spikes to ungodly cpu usage even with no one in the server. The
specifications of the box is Dual 2.8Ghz Xeon with a 1GB DDR Ram, 36GB scsi
drive, and a single nic Running RH9. Currently I have 6 servers running on
the server but that really isn't an issue as my problems occur even when
there is only one or two servers running on the machine.
Lets run down the fixes I have tried. The server originally had a 2.4.18smp
stock kernel. That was downgraded to the 2.4.9 kernel with no real
improvement. Then upgraded to the 2.4.21 kernel straight from kernel.org so
no RH patches in there. This kernel was stripped down as far as it could
possibly go. While this did help some, the spikes are still there. I also
made sure every service that could be turned off was. So the box is as lean
as it could possibly be and still these darn spikes. I have experimented
with pingboost. pingboost0 has the spikes but the top they spike with no
people in them is about 50-60% but pings on server aren't really good with
no pingboost. With pingboost1-3 the spikes can reach 99-100% cpu usage with
no one in them but pingboost1 reduced the ping about 15-20ms. I have no clue
what else I could change in the software that would fix the problem any more
since it might as well be a stock linux install with the new kernel and
almost no services running.
I have been thinking for the past few days that maybe it's just a hardware
configuration/incompatibility issue but wasn't sure where to start on that
end as it's a remote server and a dedicated and not collocated. I saw Frank
mention something about NIC cards being an issue in the kernel thread and I
started thinking about it and the NIC card could very well be the problem.
Either the hardware, driver, or settings. The only thing really interacting
with a server while no one is in it is probably people pinging it through HL
or HLSW or the such when they are looking for servers.
So now how to investigate and fix the issue if any. Replacing the NIC with a
new model is VERY last on list. I would be interested to hear what NICs both
people who have the ZeroPlayer CPU spikage issue and the ones who don't have
installed. By the way /proc/pci shows it as:
Bus 0, device 2, function 0:
Ethernet controller: Intel Corp. 82540EM Gigabit Ethernet Controller
(rev 2).
IRQ 16.
Master Capable. Latency=32. Min Gnt=255.
Non-prefetchable 32 bit memory at 0xfe100000 [0xfe11ffff].
I/O at 0xecc0 [0xecff].
Drivers in Linux are unfamiliar with me. Any help on if it's even possible
to update the driver by itself or find if there is a more recent one would
be helpful. But I doubt it since the new 2.4.21 kernel was just installed.
That leaves configuration. How do I check to see what the configurations for
this nic are and any recommendations on what to change to make it run
efficiently?
Thanks
Jeremy
_______________________________________________
To unsubscribe, edit your list preferences, or view the list archives, please visit:
http://list.valvesoftware.com/mailman/listinfo/hlds_linux
