Best settings to setup a lagless server ???

Spookstah · 2003-09-13 09:58:16 UTC

I like to know what settings other server admins did apply to their server(s) and why they did set this particular setting/value.

Settings like:

g_antilag 0/ - why did you set it off/n and how did it change your server(s) behaviour.
sv_maxrate 0-25000 - what setting(s) did you test and why did you set it to the setting you have now.
set pmove_fixed 0/1 - why did you set it off/n and how did it change your server(s) behaviour.
etc. (other non cpu/bandwitdh settings are also welcome)

Im asking this because i get complains about lag in our servers, while we have enough raw cpu power (left) and plenty of bandwidth on the fastest inet link of our country and i wanna make sure it isnt my server setup thats causing it.

Our settings are atm:

set sv_maxRate “10000”
set g_antilag “1”
set pmove_fixed “0”
set pmove_msec “8”
set sv_floodProtect “1”

Lekdevil.NL · 2003-09-13 10:37:21 UTC

I’ve taken a look at the LP II server. It appears that the server interrupts its update stream to the clients every 10 seconds, exactly. This indicates that there could be a rogue process running on your server that causes a CPU spike every 10 seconds (causing the ET server to lag out of a moment). It’s also possible that your IDE subsystem is not configured properly, leading to process interruptions whenever the disk gets accessed.

You can spot the update interrupts by connecting to the server and watching the lagometer. You’ll see a big yellow triangle appear on the upper graph every 10 seconds. That shows you the moment when the client isn’t receiving updates from the server.

In any case, treat it as a system issue. There is nothing fundamentally wrong with your ET config (although it wouldn’t hurt to increase your maxrate to 12000 or so).

HTH.

Spookstah · 2003-09-13 18:45:29 UTC

I did check what you said and indeed around every 10 secs there is a lag spike, i only cant think of anything that causes this.

Its a unix server btw so im sure i dont have to worry about IDE settings and stuff like that, also i dont notice any cpu spikes or other odd things.

CPU states: 41.7% user, 0.0% nice, 29.2% system, 4.2% interrupt, 25.0% idle
Mem: 329M Active, 433M Inact, 183M Wired, 53M Cache, 112M Buf, 6596K Free
Swap: 1024M Total, 136K Used, 1024M Free

PID USERNAME PRI NICE SIZE RES STATE TIME WCPU CPU COMMAND
91227 etserver 2 0 152M 95220K select 22.0H 30.03% 30.03% etded.x86
42452 etserver2 48 0 150M 92788K RUN 925:10 29.59% 29.59% etded.x86
88037 goserver 2 0 23176K 5600K poll 781:03 3.66% 3.66% teamspeak
78204 etserver2 10 0 3120K 1656K nanslp 128:02 0.93% 0.93% wolfbot
28205 etserver 10 0 3508K 2076K nanslp 200:57 0.63% 0.63% wolfbot
88038 goserver 2 0 23176K 5600K select 224:15 0.20% 0.20% teamspeak
88036 goserver 2 0 23176K 5600K poll 184:18 0.20% 0.20% teamspeak

As you can see the server has 25% cpu power left and it never changes a lot maybe a few %, but nothing that shows spikes in 1 of the processes.
Ill ask our OS admin to give it a look maybe he can come up with something.

Lekdevil.NL · 2003-09-13 18:52:14 UTC

Things you might want to look for is an ide driver running in PIO mode rather than in (Ultra) DMA mode. If the server writes to disk every 10 seconds, that could possibly affect things. Or some other hardware drivers using incorrect IRQ settings. It must something like that, i.e. a system process that doesn’t consume that much CPU cycles but somehow lags the entire system.

HTH.

Spookstah · 2003-09-14 11:46:09 UTC

Still cant get in contact with my OS admin (also known as my brother ) but i did check the server this morning without any players and there isnt a like spike every 10 secs on an empty server so it probly has something to do with players with bad setup (bandwidth setting?) or with 2 many info that needs to be written into the logs or matbe punkbuster is doing a cvar check every 10 secs?

Spookstah · 2003-09-16 21:21:09 UTC

Its indeed punkbuster which is checken cvar`s every 10 seconds or max set back to 30 secs are those guys nuts or so, a check on join and each 10 minutes sounds good enough to me.

Anyone else having these lag spikes on there pb enabled server?

_fun_scott · 2003-09-17 15:31:50 UTC

set sv_maxrate to 13000 (read the et readme it says it needs 13000 to run smoth. set pb_sv_autoss 0, <<auot snapshot. pb_sv_maxdlrate 1 <<max dl speed for pb updates 1k/s, pb_sv_maxcondls 1 << set maximum pb updates at one time, pb_sv_sleep 100 (default is 60 i think) <<tiem between checks. also redirect any dls to a difrent server.

_fun_scott · 2003-09-17 17:08:00 UTC

the reason why you might be getting spikes every ten seconds is due to this cmd pb_sv_cvarfreq 6 (reduce to 2) and also thsi cmd at the same time pb_sv_maxcondls 0 … i beleave this will check client cvars every 10 seconds and all clients at the same time… to stop this as i said before pb_sv_cvarfreq 2 and pb_sv_maxcondls 3

Spookstah · 2003-09-17 20:03:13 UTC

The problem is, i already have set them as low as possible and i get the lag spike every 30 secs.

pb_sv_maxcondls
The Lost Paradise PB Server: pb_sv_MaxConDls = 1 (1 to 6)
pb_sv_cvarfreq
The Lost Paradise PB Server: pb_sv_CvarFreq = 2 (2 to 10)

I did raise bw setting today to see if it helps, but i think it wont solve the PB lag spike maybe it can reduce the effect a little tho.

_fun_scott · 2003-09-17 21:50:49 UTC

you definatly got pb_sv_autoss 0… this is the biggest lag producer form pb. how many cvar checks do you have? i found having only 3 -4 cvars to check reduces lag a fair bit

Spookstah · 2003-09-17 21:58:09 UTC

ss = off

This is our cvar list atm:

pb_sv_cvarlist
The Lost Paradise PB Server: 1: “rate” IN 2500 25000
The Lost Paradise PB Server: 2: “snaps” IN 20 40
The Lost Paradise PB Server: 3: “cl_maxpackets” IN 15 100
The Lost Paradise PB Server: 4: “cl_timenudge” IN -50 0
The Lost Paradise PB Server: 5: “cl_freelook” IN 1 1
The Lost Paradise PB Server: 6: “m_pitch” OUT -0.015 0.015
The Lost Paradise PB Server: 7: “m_yaw” IN -0.022 0.022
The Lost Paradise PB Server: 8: “cg_bobup” IN 0 0.005
The Lost Paradise PB Server: 9: “cg_errordecay” IN 100 100
The Lost Paradise PB Server: 10: “com_maxfps” OUT 0.00001 40
The Lost Paradise PB Server: 11: “r_flares” IN 0 1
The Lost Paradise PB Server: 12: “r_picmip” IN 0 3
The Lost Paradise PB Server: 13: “r_picmip2” IN 0 3
The Lost Paradise PB Server: 14: “r_gamma” IN 0 3
The Lost Paradise PB Server: 15: “r_intensity” IN 0 1.5
The Lost Paradise PB Server: 16: “r_overbrightbits” IN 0 3
The Lost Paradise PB Server: 17: “r_mapoverbrightbits” IN 0 3
The Lost Paradise PB Server: 18: “r_uifullscreen” IN 0 0
The Lost Paradise PB Server: 19: “r_ext_nv_fog_dist” IN 0 0
The Lost Paradise PB Server: 20: “r_overbrightbits” IN 0 4
The Lost Paradise PB Server: 21: “r_mapoverbrightbits” IN 0 4
The Lost Paradise PB Server: 22: “cg_shadows” IN 0 1
The Lost Paradise PB Server: 23: “r_rmse” IN 0 75
The Lost Paradise PB Server: 24: “cg_fov” IN 80 125
The Lost Paradise PB Server: 25: “r_allowextensions” IN 1 1
The Lost Paradise PB Server: 26: “r_softwareGL” IN 0 0
The Lost Paradise PB Server: 27: “r_wolffog” IN 1 1
The Lost Paradise PB Server: 28: “r_shownormals” IN 0 0
The Lost Paradise PB Server: 29: “r_showtris” IN 0 0
The Lost Paradise PB Server: 30: “r_lightmap” IN 0 0

_fun_scott · 2003-09-18 16:42:37 UTC

i suggest running a test pb cvar check with only 3 - 4 checks. to see if that solves your problem… as we have a 100mb conn too with cpu usage at 55% and we dont have a problem now when we reduced the checks… are you running in smp 1 mode? if so turn it off… causes mroe trouble than its worth.