What is the server doing when idle? - high CPU usage

F1Fan07

F1Fan07

Member
What causes the dedicated server EXE to be using ~25% CPU when idle (this is with the zero player idle flag set to either 1 or 0)? Note that the definition of CPU is a bit strange for me because rF2 is running in a cloud computing virtual machine (Amazon AWS) that delivers CPU resources when needed but Windows is showing the CPU taking a significant hit from rF2.

I've done two installs (current build and the one before.. 110?) and see the same behavoir. This same setup hosts 2-3 rF1 servers with no abnormal behavoir.
 
No one?

Let me ask the question in a different way then... is there an option or flag in multiplayer.ini that I can adjust to stop or minimize the idle CPU usage or is rF2 just freaking out in this virtual server environment?
 
If you haven't tried yet try a clean install (save Packages somewhere, uninstall current rF2, delete any left over files, do full reinstall) any of the latest build.
 
F1Fan07 said:
What causes the dedicated server EXE to be using ~25% CPU when idle (this is with the zero player idle flag set to either 1 or 0)? Note that the definition of CPU is a bit strange for me because rF2 is running in a cloud computing virtual machine (Amazon AWS) that delivers CPU resources when needed but Windows is showing the CPU taking a significant hit from rF2.

I've done two installs (current build and the one before.. 110?) and see the same behavoir. This same setup hosts 2-3 rF1 servers
with no abnormal behavoir.
Click to expand...

Howdy,

I think you've answered your own question... 25% of what?

I looked into Amazon AWS. They talk about the performance of their servers being scalable... As a matter of fact most of the reading I did at their site left my head in a cloud... sorry, I couldn't resist. :)

Even though they say that it scales to meet the requirements of the program running, is there lag time? I'd find it hard to believe that it's in real time.

Although they make the claim of being able to run applications from their cloud servers; I wouldn't assume that, that meant games as well.

As far as what you could do with rF1 vs rF2... maybe you could run 2 or 3 rF1 servers from that "same" cloud. rF2 is quite the different animal... it requires, substantially, more resources.

And being a cloud setup, you probably can't set the affinity of rf2_dedicated.exe anyways. (Durge's suggestion of running rF2 dedicated on one core.) What core??? LOL!

Hope this helps,

Bryan
 
Last edited by a moderator:
I too have noticed this increase in performance by the servers.
happens that the server begin to raise the request of the processor.
A server just started demands by 05/10% power.
An example now an empty server that takes 30%, and one with 10 people taking the 09%, in my opinion be a problem.
Now I'll try to get it going with a core, I tried with two, and remain the same.

Edit: I tried to just one core and the situation remain the same.
In my opinion it does not work well, and go to conflicting increasing the demand for power.


Babs
 
Last edited by a moderator:
Yeah, that's the root of my original question. When there is no active session, CPU usage should be the same as rF1 in that state (0-5%). There is some code running unnecessarily that is consuming CPU cycles.
 
Bryan Birtwell said:
And being a cloud setup, you probably can't set the affinity of rf2_dedicated.exe anyways. (Durge's suggestion of running rF2 dedicated on one core.) What core??? LOL!
Click to expand...


What causes the dedicated server EXE to be using ~25% CPU
Click to expand...


:confused: I am obviously missing something.......


View attachment 5305
View attachment 5306

I never said it would work or that it would reduce CPU usage.

From what I am seeing you can run it one core and it shows no usage. !
Maybe I need new glasses these are a bit THICK. lol ;)
 
Last edited by a moderator:
F1Fan07 said:
Yeah, that's the root of my original question. When there is no active session, CPU usage should be the same as rF1 in that state (0-5%). There is some code running unnecessarily that is consuming CPU cycles.
Click to expand...

The same machine is hosting multiple servers? (Trying to track the issue)
 
If I am reading my graphs correct, 3 cores are disabled and the empty dedicated server is running with no usage, not 30%

All I can do is hazard a guess ?

Windows lies to me OR 3rd party servers may have something to do with it as I don't use one.
 
last update on the situation.
The server does not immediately take over power this happens after a certain period of time.
I build up to 125, I was 5 server on the same machine, and they never gave performance problems.
Usually a server takes a 5/10% maximum, this evening we had a championship race and I had to cancel, because the server (for the race I had turned off all the others) after an hour and went to full power, and we all received the connection lost, and now I can not even connect with the remote to stop it.
I think it's 100%
I'm hoping for a quick fix, because I fear to again turn on the server.

Babs
 
Here server has been up for 2 hours but only a few drivers this morning.

So I gather it won't be a problem for a 8 car room running off the dedicated server ?
I have not seen it.............. then again I have not been looking.

I have had matchmaker fail or server crash but that has happened with even 1 car
more then any other build , Build125 was slightly better.

Build45 I used to put room up all day and night at Spa never a Reload ( to keep rubber) and barely a crash, so I don't know.

View attachment 5307
 
Last edited by a moderator:
F1Fan07 said:
Yeah, that's the root of my original question. When there is no active session, CPU usage should be the same as rF1 in that state (0-5%). There is some code running unnecessarily that is consuming CPU cycles.
Click to expand...

I see mine idling at about 10% of core use. But then that's not some kind of magic dynamic core, just the one that is assigned to VMware to use out of the three on the CPU.

Do you have any way of checking how much CPU you have at any given time? I'm wondering if rf1 is idling at 0% (ie under 0.5%) of a lot whereas rf2 is idling at 30% of not very much at all.

Or maybe it's something like atmospherics being constantly calculated even when nobody is there. Like, realroad is still real even if nobody is driving on it. Track temps are still calculated from weather. That sort of thing.

graphic:

snapshot1.png
 
Last edited by a moderator:
Spadge said:
I see mine idling at about 10% of core use. But then that's not some kind of magic dynamic core, just the one that is assigned to VMware to use out of the three on the CPU.

Do you have any way of checking how much CPU you have at any given time? I'm wondering if rf1 is idling at 0% (ie under 0.5%) of a lot whereas rf2 is idling at 30% of not very much at all.

Or maybe it's something like atmospherics being constantly calculated even when nobody is there. Like, realroad is still real even if nobody is driving on it. Track temps are still calculated from weather. That sort of thing.

graphic:

View attachment 5308
Click to expand...


I blame Orinoco, he's probably fallen asleep and left the telly on. :D
 
Hi
I confirm resources have increased in me 4% lower and amounted to 50% and the time is random.

Server configuration: Intel E5200 Windows XP 3G ram
 
Guys, i think the issue was not noticed in the internal tests, not in the ISI server and not in the Euro server hosted by an internal tester.

Both the servers are running one dedicated server instance per machine.

Having confirmation that the issue happens only on machines running multiple dedi instances might help narrowing the pattern of the issue.
 
That's me Max is talking about I guess ;)

I'm hosting rf2 on a KVM-based CentOS 6 Linux box (windows xp guest), and cpu usage with rf2 dedi more or less idle is around 1% on the guest or 0.05 for the CentOS server itself (it's also hosting a minecraft server, so not all of that load may be caused by rf2).

You can check the server in the lobby, it's called "hoover kvm build #134" ATM.

Cheers, Uwe
 
You must log in or register to reply here.
Back
Top