It would be awesome if Valve would fix the memory consumption of the
various engines. Given how long this has been an issue though it is
unlikely.
Are you running GoTV Winnie? I know when I used to run SourceTV on my
TF2 servers the memory consumption was even worse. I imagine it's the
same with GoTV.
On 11/8/2012 10:26 AM, Kyle Sanderson wrote:
Memory consumption aside, it looks like you're hitting the horribly
implemented threading in the various engines (Which is why your graphs
indicate you're crashing when the FPS drops). If it's any consolation,
I have the same problem. The only work around is to attempt to
optimize your system as much as possible. Or you can totally disable
threading, which would probably stop the crashing at the expense of
performance.
http://i.imgur.com/uK76I.png
As far as memory usage, it's probably unrelated to the I/O you're seeing.
Thanks,
Kyle.
On Thu, Nov 8, 2012 at 10:12 AM, Winnie the Pooh
<[email protected] <mailto:[email protected]>> wrote:
Can anyone else shed any light on this issue? Or help figure out
what is going on?
You can see what I've put together over time here:
http://64bitvps.com/csgo/ticket/srcds-memory-consumption-increases-every-map-change-and-eventually-affects-server-fps/
https://forums.alliedmods.net/showthread.php?t=199175
Basically,
* It has been seen for sure on Windows Server 2012, 2008 R2, and
2003. (x86 and x64)
* It is NOT a plugin issue. This has been tested and happens on
servers without an /addons/ directory with just vanilla configs
* It happens with both sv_pure 0 and sv_pure 1
* Memory increases each map change (this may be by design)
* /pak01_051.vpk gets read abnormally like a handle isn't closed
after ~550mb of RAM use
* It is based on population - highly populated servers, frequent
map changing sees this more clearly
* It is not an issue of file corruption or setup - the servers
have been deleted and recreated from scratch
* When FPS drops, CPU utilization almost doubles and I/O is high
(reading /pak01_051.vpk)
o I/O is not the limiter - has happened on a RAMdisk
o CPU Affinity or priority doesn't seem to affect it
* Running -debug -dev doesn't tell much in server console except
that client cmd rates are off during the CPU spike
On more populated servers I have yet to ever hit an uptime of more
than 20 hours! ( http://i45.tinypic.com/2n0qyw.png )
Yes, yes, yes I know I can fire up a Linux VM and call it a
day(maybe. some report it happening on centOS too) but that isn't
really the solution here.
I've created a full memory .dmp during the time of the FPS drop
and will see if I can find anything in it after I get home. If
anyone with more WinDbg experience wants to take a look at this,
let me know.
If you have any information or have this same issue, please chime in.
--
-Winnie the Pooh
TimetoKill.net
---------------------------------------------------------------------
Rest of the conversation truncated. Look in your inbox.
_______________________________________________
Csgo_servers mailing list
[email protected]
<mailto:[email protected]>
https://list.valvesoftware.com/cgi-bin/mailman/listinfo/csgo_servers
_______________________________________________
Csgo_servers mailing list
[email protected]
https://list.valvesoftware.com/cgi-bin/mailman/listinfo/csgo_servers
_______________________________________________
Csgo_servers mailing list
[email protected]
https://list.valvesoftware.com/cgi-bin/mailman/listinfo/csgo_servers
