UDK.exe CPU Usage drops to 0 then back up -massive lag spike



  • I have an odd one. Stripped down W7, no other programs running but UDK.exe, AMD 8120 8 core 3.1GHz, 32GB RAM, currently setup for TDM with only 16 players, its running perfectly smooth and fine then suddenly a MASSIVE lag spike in game, players just start running off in a straight line then 2 ~ 5 seconds later it catches up, I look at taskmgr and see the CPU usage plummeted to zero percent then back up again. I am on LAN with the server and sometimes my ping is rocketting from 10ms to nearly 100ms……ideas?



  • @ziggo0:

    I have an odd one. Stripped down W7, no other programs running but UDK.exe, AMD 8120, 32GB RAM, currently setup for TDM with only 16 players, its running perfectly smooth and fine then suddenly a MASSIVE lag spike in game, players just start running off in a straight line then 2 ~ 5 seconds later it catches up, I look at taskmgr and see the CPU usage plummeted to zero percent then back up again. I am on LAN with the server and sometimes my ping is rocketting from 10ms to nearly 100ms……ideas?

    CPU of the server or of your game client?



  • Oh, server sorry. I set affinity to a single core to observe this. All connected clients notice the massive lag spike as well.

    I suppose I should elaborate on my setup a bit more because it’s not straight W7. The AMD 8120 8 core/32GB RAM is an ESXi 5.1 box, I’ve tried few and tons of resources for the W7 VM but it did not matter. About to get a screenshot of what I’m talking about heh



  • Here is what it looks like in taskmgr http://i.imgur.com/tYXw16o.png

    Tried doing a linux/wine setup….still massive amounts of lag :(

    Edit: the linux based one seems to be fine after running for a few minutes, no lag spikes, interesting



  • This sounds similar to what’s happening on our windows 2008R2 hosted servers. Cpu load seems to plummet when the lag spikes occur on our set up as well.
    I made a post about it earlier http://www.chivalrythegame.com/forums/viewtopic.php?f=72&t=8857
    Only difference is that the Linux/Wine setup in our case doesn’t suffer the lag spikes.

    Additional server info we are also running everything on esxi 5.1.0, 799733 but using Intel CPU.



  • @Sgt.TAtari:

    This sounds similar to what’s happening on our windows 2008R2 hosted servers. Cpu load seems to plummet when the lag spikes occur on our set up as well.
    I made a post about it earlier http://www.chivalrythegame.com/forums/viewtopic.php?f=72&t=8857
    Only difference is that the Linux/Wine setup in our case doesn’t suffer the lag spikes.

    Additional server info we are also running everything on esxi 5.1.0, 799733 but using Intel CPU.

    Interesting, I was going to be giving 2008R2 a try later today to see if it’s simply W7 being problematic. How is your datastore setup for your 2008R2 VM? I’ve got a NAS running with NFS mounted over VMXNET3 back onto ESXi…works great but I’m wondering if there might be an issue there causing this lag spike (doubtful but worth a shot). I’m creating a local VM to test it now…

    I’m curious, did these lag spikes happen before the free weekend? Judging by the date on your thread they were - I’m wondering if the massive amount of ping requests are causing these CMW servers to have hickups



  • The datastore is a local 4drive RAID10 on a Adaptec HW Raid Controller. The esxi is also configured with a 40GB SSD Cache.

    And yes we had the lag spikes even before the free weekend, the amount of players doesn’t seem to affect the lag spikes at all. Sometimes it can run fine for hours with 20 players and all of a sudden it can just start getting massive lags even if its only hand full of players online.

    One thing I have noticed is if you don’t have the high udp port range (30000+) open to the server you might get a lag spike every time someone tries to join your server.



  • Interesting, so my local datastore idea is pretty much shot then. Watching iotop says not much writing is done by the server anyways. You mean have more ports open besides 7777/7778/27025?



  • @ziggo0:

    Interesting, so my local datastore idea is pretty much shot then. Watching iotop says not much writing is done by the server anyways. You mean have more ports open besides 7777/7778/27025?

    Yes if I look at the incoming udp state table to the chivalry servers I have a few in the low 50000 and a few around 63000. This only seems to be required by a few players and if the high range isn’t opened those players won’t be able to connect and you will see the UDK.exe try to allocate a lot of ports trying to find one they can use.



  • You using pfSense? Same here. I noticed in states there was a lot of high ports too (50000+) but didn’t know what to make of it. I noticed in Firewall logs it’s block many UDP connections left and right…perhaps in our case the pfSense firewall is causing lag? CPU usage of pfSense is never more than a few percent in my case.

    I up’d the players to 20 on my TDM server, there was lag for the first 5 minutes but it seems to be running great now. Odd?



  • Yes, I’m using pfsense as well but I’m running pfsense 2.1Beta 32bit. 1:1 NAT to the game server.
    Allow udp in and it might solve the lag at least if its caused by joiners.



  • I have the EXACT same issue, but running a 2008 r2 on a Hyper-v Server. Identical issue, cpu drop then massive lag spike. If someone found a solution I would aprecciate a lot



  • @ziggo0:

    I have an odd one. Stripped down W7, no other programs running but UDK.exe, AMD 8120 8 core 3.1GHz, 32GB RAM, currently setup for TDM with only 16 players, its running perfectly smooth and fine then suddenly a MASSIVE lag spike in game, players just start running off in a straight line then 2 ~ 5 seconds later it catches up, I look at taskmgr and see the CPU usage plummeted to zero percent then back up again. I am on LAN with the server and sometimes my ping is rocketting from 10ms to nearly 100ms……ideas?

    Hi, I know this is an old thread, but I have this EXACT same problem and did not see a solution here. My server runs perfectly with server lag around 5 to 12 ms (showdebug)… Everything runs perfectly well, but every now and then (perhaps every 5 to 10 minutes, sometimes less often and sometimes more often) there is a sudden FREEZE (lag spike). I know it is nothing wrong with my internet connection (100M up / 20M down, hosting 24 player server).

    I checked my CPU monitoring software after reading your post and lo and behold… UDK.exe drops to zero CPU during these spikes. At all other times CPU usage is rock steady at around 12%.

    Does anyone know how this could be tweaked or fixed? I’ve already set UDK.exe to high priority but it hasn’t made any difference.



  • Just as an aside, I am also curious whether it might have something to do with the free weekend, as another poster mentioned? The problems seems to get SLIGHTLY better (but not fixed) when I kick everyone off the server with a ping higher than 200 ms.

    Edit: On another note, is there any way to turn off server logging completely? I now use UDK.exe instead of UDKLogging.exe, but the server window is still constantly full of information being logged and I can’t help but wonder if it might be a factor…

    I get dozens of these messages every minute in the server box:

    [7341.53] Log: Command not recognized: SETSOUNDMODE Default
    [7341.76] Log: Command not recognized: SETSOUNDMODE Default
    [7341.43] Log: Command not recognized: SETSOUNDMODE Default
    [7341.40] Log: Command not recognized: SETSOUNDMODE Default
    [7365.10] Log: Command not recognized: SETSOUNDMODE Default

    ad infinitum….


Log in to reply