Max CPU Usage? Or kill processes with CPU too high

[Benheryet]

I have 4 cores on my server. When a GMOD server crashes, it doesnt just die and restart, it will just hang with 1000 ping and 25% cpu (using a whole core) untill it is manually restarted. Because it never stops responding (just has 1000 ping), TCAdmin doesnt auto stop server.

 

Is there anything in TCAdmin? or external program which can solve this?

[Steven Crothers]

Thats exactly why we don't host GMod. I wish you luck in finding something to fix that for you. Theres really nothing good out there.

[JasonF]

Do you have the Game Server Monitor plugin enabled to query game servers and restart them if they crash?

[Benheryet]

Do you have the Game Server Monitor plugin enabled to query game servers and restart them if they crash?

 

Yes, but when gmod servers crash they dont stop responding, just get stuck at 1000 ping and 25% cpu usage

[ECF]

So the server actually will return a query response in this state?

[Steven Crothers]

Right, the LUA locks up. Not the game engine. The engine will still respond to queries. It just can't process RCON, players, or anything else.

[Benheryet]

So the server actually will return a query response in this state?

 

Yes, but infact the server is stuck using 25% cpu and 1000 ping

 

It should not respond, but it does

[ECF]

Nothing we can do on this side about that. That is something that needs to be fixed in the game engine.

[Steven Crothers]

lol, fix gmod...

 

Let me quote Garry "Working as intended, make your LUA better"

 

How about a max cpu over x time kill? That could be done, and should be done now that you know a game requires it.

[Benheryet]

For all games it would be usefull for "if cpu is over x then stop/restart" not just gmod =]

[KingJ]

It would be good to see a "kill server if CPU is over 80% of a single processor core" function, there are games other than GMod that are based around server-side scripting and quite a few people have no idea how to script properly (A nested loop doing something complicated for every player on the server 10 times a second? Yeah that's fine, lets code it!)

 

It should be coded such that the calculation takes into account the singlethreaded nature of most game servers. In an environment where you may have dual, qual, oct and higher core servers, setting a simple percentage of total CPU will not work. You need to get the total number of processors and if TotalCPUOfAProcess * Cores > 80%, kill the server and inform user.

[Benheryet]

It would be good to see a "kill server if CPU is over 80% of a single processor core" function, there are games other than GMod that are based around server-side scripting and quite a few people have no idea how to script properly (A nested loop doing something complicated for every player on the server 10 times a second? Yeah that's fine, lets code it!)

 

It should be coded such that the calculation takes into account the singlethreaded nature of most game servers. In an environment where you may have dual, qual, oct and higher core servers, setting a simple percentage of total CPU will not work. You need to get the total number of processors and if TotalCPUOfAProcess * Cores > 80%, kill the server and inform user.

 

I agree

[Steven Crothers]

I don't think we'll see anything new feature-wise to TCAdmin for quite some time. Since all efforts seem to be going to TCA2 development, and not TCA1 bugfixing. Last I heard development was taking much longer then anticipated, so perhaps feature requests are better kept to ourselves until early to mid next year, when TCA2 comes out.

[ECF]

So basically you want us to fix what someone else has screwed up right?

[KingJ]

So basically you want us to fix what someone else has screwed up right?

 

Sorta, we want the ability to automatically terminate processes that are using way more processor than they really should to help protect the stability of the machine, server and prevent the skewing of the load-balancing system.

 

But I guess you could call it other people's screwups if you really want to

[Steven Crothers]

Or call it a feature, which kills off responding over cpu limit processes which use x cpu over h hours...

 

Seems like a fantastic idea to me, actually I can't think of a reason why it shouldn't be in there with or without the gmod problem. It should be in there simply because its a "good idea".

[ECF]

I undestand what you are asking. But it think the coders for the games should be held accountable for their issues not us.

 

Not that the feature would not be of a benefit to our product. Most likely it will be added in V2.

[Benheryet]

I undestand what you are asking. But it think the coders for the games should be held accountable for their issues not us.

 

Not that the feature would not be of a benefit to our product. Most likely it will be added in V2.

 

It would benifit TCAdmin greatly, the ability to make sure a certain server doesnt use over xxx CPU

 

This isnt just a random fix for a game, it can be used for load balencing and everything

 

Unless it's very hard to code, why wait for V2 (which could be years away as no date has been said )

[Steven Crothers]

It should be more then X cpu, it should be X cpu over H time.

 

That way we can set a more realistic usage. For example-

 

80% cpu over 30 minutes = kill server.

 

Should also be tied into service levels.

[ECF]

Why wait for Version 2? Because evertime we stop to add something to Version 1 that pushes Version 2 further off.

[Steven Crothers]

Playing the waiting game for v2 is going to get long and strenuous quickly though. You have already openly admitted that development is taking far longer then you want and you don't have a planned release date. Which tells me it will be a very long time before we see something new. I honestly don't expect to see anything until 4Q 09 or 1Q 2010.

[Benheryet]

Also this is a pretty big problem, as it means I am checking my server hourly, to make sure no GMOD (or other scripting games) have crashed

[ECF]

I will simply say it will not be added to V1. The reason why V2 is so far behind is all the reuest we have done for V1.

 

Time to move forward instead of running in place and get V2 coded out.

[Benheryet]

I will simply say it will not be added to V1. The reason why V2 is so far behind is all the reuest we have done for V1.

 

Time to move forward instead of running in place and get V2 coded out.

 

Can you give us a Year and quater for release then?