We have the super xeon, with 2x24 player CS pubs and a 32 player bf vietnam pub running. Every so often the cpu usage goes to 100 for the CS pubs. This occurs with nobody on the pubs and lasts anywhere from 5-10 secs. Just wondering if this is normal or is there something we can do about it. The last time it happened it lagged out the bfv server.

Hard to say exactly what is causing that but there are alot of things you should do on the machine before running any game servers.

What OS are you using?

redhat enterprise
kernel - 2.4.20-30.9smp

I'm not to great with Linux but the same applies to all gaming machines.

Make sure you shut down any service that you don't need! There are a million things running behind the scenes that you do not need.

Maybe someone else can help with how to go about doing that.

I don't do games but I think a random burst for 5 to 10 seconds should not be something to worry about.

Do you by any chance have some control panel on the server or running some cronjob in the background to do some updates or something during that time.

Sometimes server do backups and during file compression, file transfers, tying up databases or some administrative functions can spike the CPU as they can be quite extensive.

Sounds like you have something config'd wrong. "map/stats/mods/plugins"

CS should not just spike.

What does your server.cfg file look like what mods/maps/plugins are you running
Are you using CZ maps in a CS server?

There are many things to cause this. You need to spell out what you have for us to help.

sounds more like at map changes the cpu spikes up to load the map. If it doesnt cause lag to the other servers i wouldnt worry, thats normal for map changes

The spikes occur with nobody playing and no map changes and 1 of the servers has no external mods running, pingboost 1, cal.cfg (mp_maxrounds 0). So I'm wondering if this is a hyperthreading problem or do we just need to reinstall CS.

Pingboost set at anything other then "0" is a hog.

waht mods do you have and all the plugins runnign?

You are not giving me the needed infor to make a call on your problem.

Our 11 CS / CZ servers do not do this.
Use this tut to install all your cs servers.
http://www.hladmins.net/old/modules.php?op...rticles&secid=1

1 server is running amxmod with the following plugins added:
amx_atac.amx
amx_atac_cfg.amx
custom_map_configs.amx
dgcawpmap.amx
amx_blackjack.amx
advanced_tracers.amx
amx_ejl_cmdlistgenerator.amx
weaponarena.amx

The other server is a halflife deathmatch server with no plugins or mods.

Both servers show cpu spikes with no players on (cpu spikes occur even when the map isn't changing). The only thing I have noticed is that both hlds jobs always seem to run on the same cpu. Is this because each job was started under a different user?

Rolla and I run the xeon.

Several things. The HLDS is installed and correctly configured. I have been running CS servers for over 4 years. One cs match server running no plugins with no map changes spike to over 90% on the CPU randomly. This server does not change maps.

The half-life deathmatch server (12 player) runs the stock valve HLDS with no mods and spikes as well without map changes.

The CS public server (20) runs AMX with the plugins noted above and spikes as well independently of map changes.

Interestingly, the BFV server does not spike at all that I have witnessed.

We have no cpanel and only urchin is running in the background other than standard OS processes.

This is not a CS server issue. I have a pentium 2.4 box with linux redhat 9 and the CS servers run great with no spikes. I get better performance from the pentium 2.4 server than I do from the xeon server (in terms of cpu usage and random spikes). The dual xeon runs red hat enterpise 3.

The problem is either an OS problem or a hardware problem. It is not a software installation or configuration problem. I disagree about pingboost. I acknoledge that it does increase CPU usages but not much when set at 1. Now 3 is a different story. However, it does not cause random CPU usage spikes. I have 3 busy pub (24,20,14) with bots on 2 of them as well as 3 locked match server with cpu usages below 60% at peak hours and no spikes.

Any help would be greatly appreciated.

Cooley

Please specify: 100% of the CPU?

100% of the total CPU or 100% of the two Xeons or 100% of one of the four virtual procs?

I'm running CS on a Super Xeon with Debian, but I've upgraded the kernel to 2.6.4-smp. A full 16 player CS server runs about 30-40% of just 1 virtual processor. I have a wide range of AMX plugins though and its quite acceptable. CZERO is even worse, currently running 50% full, 16 slots no plugins.

The empty servers should not be spiking at all, they should be sitting there with 0%, with MAYBE small bursts to 5-10% for a map change. Definately nothing sustained over 10-20 seconds though.

I would suggest the problem is with RHE. There's multiple threads mentioning SM upgrading people to 2.6 kernels to resolve issues with Xeons and Hyperthreading.

Ferret,

Thanks for the help. We looked for threads about kernals and hyperthreading but must have missed them. The servers spike to anywhere from 70-99% on a single CPU with the others either 0 or at 30% if the the BFV is active. Rolla will correct me if I am wrong but this is what I observed. Looks like we need a kernal upgrade or we should do away the HT.

Thanks,
Cooley

Send in a support ticket to have SM upgrade the Kernel...this should do away with the spikes.

Many thanks. I have submitted the ticket.

Cooley