[Help] How to reduce ghost hits, desyncs, stuttering?



  • For the past 2 months, 2 of the 3 available servers in my area (South America) have been experiencing more and more teleport/warping issues, lots of ghost hits and a general decrease in performance and smoothness.

    I’m the owner of one of them (12 slots) and also friends with the staff of the other server (28 slots), so I was hoping maybe some more experienced people could help me/us figure out what is wrong with them.

    The issue I’m talking about is a lack of fluidity in character positions and some animations. Like judging distances is harder than in other svs, and I swear swing animations look choppier. Since I can’t access official svs, the only place I can compare performances with is the Brazilian Coldfire server, which is incredibly smooth. Even though I’m behind a VPN (I get shitty ping otherwise, >200), with 60 ms I can see a lot of frames in each swing and the model’s positions don’t jump around as much.

    I don’t know if the physical server has my own Chiv server virtualized (I think it’s unlikely because the log would not show my available RAM as 24 GB, for starters) and I can’t run much diagnostics on it. I only access it through TCAdmin. Another clue about it not being a VPS is that when one of the 2 server crashes (i.e., the process stops responding in TCadmin panel) it usually brings the other one down as well. I also don’t know which other gaming services they’re running at that physical computer.

    The hardware is an i7 and based on the log I can provide these other values:
    CPU Page size=4096, Processors=8
    3.410273 MHz
    Memory total: Physical=23.9GB (24GB approx) Pagefile=71.8GB Virtual=4.0GB

    It’s a 12 slots server. With 8-10 players and 6 bots (for stressing purposes, you know), showdebug 1 informs me that the server response is pretty stable, between 2 ms and 4 for most of the times. It sometimes goes as far up as 8 ms. The fluctuation is usually small. At one second it will be 2.4 ms then 3.1, then 2.something ms then 4.blah ms.

    Supposedly, the datacenter has a 10 MBPS guaranteed upload rate, and some pretty state of the art DDoS protections.

    I’ve read this article and tried different things with no real results, like increasing the tickrate among others. Btw, if I were to double my rates, the hosting would charge me twice as much :p

    At this moment, my rates are 30000 for MaxInternetClientRate and 30 for the tickrate.

    Is it a problem of bandwidth? Is hardware resources? Is it the aggressive DDoS protection or what?
    Is the server module getting worse and worse after each Chivalry patch?

    Have I set it up poorly? How should I go about network rates and tickrates?
    I’ve read a lot and haven’t found out just what the math should be like (MaxClientRate / Idk = What?)

    Anyway, this is all the info I could think of. Any help / advice is welcome.
    Please help me keep the Chivalry scene alive in the third world u_u :D



  • The config on my servers (COLDFIRE) (I had that username on this forum, forgot the password and the email LONG time ago, before AoC CR2) on theses settings are:

    [IpDrv.TcpNetDriver]
    AllowDownloads=True
    AllowPeerConnections=False
    AllowPeerVoice=False
    ConnectionTimeout=30.0
    InitialConnectTimeout=200.0
    AckTimeout=1.0
    KeepAliveTime=0.2
    MaxClientRate=50000
    MaxInternetClientRate=30000
    RelevantTimeout=10.0
    SpawnPrioritySeconds=1.0
    ServerTravelPause=4.0
    NetServerMaxTickRate=60.0
    LanServerMaxTickRate=35
    DownloadManagers=IpDrv.HTTPDownload
    DownloadManagers=Engine.ChannelDownload
    NetConnectionClassName=IpDrv.TcpipConnection

    I also have

    [Engine.Engine]
    bAllowRelevancyAutoAdjust=TRUE
    bNoComplexPawnRelevancy=TRUE

    Enabled, this is supposed to help with high cpu usage, but on the latest patches I think that it’s doing more harm than good.

    The server itself is a i5 3570 running Hyper-V, and here is where the settings gets a little weird. I was only able to run this server on Windows Server 2k3, on 2k8 and above it gives me random lag spikes where EVERYONE locks up and de cpu use goes to 0 on the server. I’m using the 32bits binaries.

    One problem that happens semi frequently is the lag spike that increases the pings for everyone and the server ms climbs, but since changing the server folder to an SSD I’m yet to see that happen (fingers crossed)

    The internet connection is a business fiber with 100/50 mbps running behind a pfsense firewall (also virtualized).


  • Global Moderator

    Did those ping spikes happen only on citadel and coldfront? At least mostly. I have 13 chivalry servers set up almost the same as yours (24 core intel Xeon :p) and have the same ping spike issues. Though I don’t get the lock up 0% usage problem. Servers don’t crash very often. In fact they probably crash more due to my stupidity than anything else. We also got a 1gbps connection. Which is technically a flat out lie as these servers are in Australia and Australia’s infrastructure cannot support it lol. Don’t even have fibre. Better off hosting in New Zealand.



  • Nope, the lag spikes (the pings remained stable but the server process was frozen so was everyone, including me, and I’m sitting just a switch away from the server) happened on every map.



  • Hi, sorry it took me a few days to reply.

    @onesizefitsall:

    The server itself is a i5 3570 running Hyper-V, and here is where the settings gets a little weird. I was only able to run this server on Windows Server 2k3, on 2k8 and above it gives me random lag spikes where EVERYONE locks up and de cpu use goes to 0 on the server. I’m using the 32bits binaries.

    One problem that happens semi frequently is the lag spike that increases the pings for everyone and the server ms climbs, but since changing the server folder to an SSD I’m yet to see that happen (fingers crossed)

    The internet connection is a business fiber with 100/50 mbps running behind a pfsense firewall (also virtualized).

    Thanks for the info on your current hardware and rates, man. That will be very helpful.
    I’ve noticed that the lagging seems to occur at random times and I’ve also found it was affecting teleports and weird shit with players with decent ping (about 70ms) more frequently. These players were mostly foreigners and you can see a vid here kindly recorded by one of our top players.

    I haven’t experienced ping spikes that would kick everyone, which is weird.

    Some of the stability issues I’ve come to realize are related to this bug. Our server runs only custom maps on its rotation, so it makes sense it would simply restart and the panel would become unmanageable. Or so I’m thinking atm.


Log in to reply
 

Looks like your connection to Torn Banner Forums was lost, please wait while we try to reconnect.