ettv_b6 causing MAJOR lags on broadcasts

Forum for discussing ET TV

Moderators: Forum moderators, developers

User avatar
arni
Posts: 188
Joined: Sun Feb 20, 2005 2:32 pm

Post by arni »

This is for my linux recorder/replayer - maybe chosen can provide data for the linux master this happened on

RX packets:151205719 errors:0 dropped:0 overruns:0 frame:0
TX packets:69799602 errors:0 dropped:0 overruns:0 carrier:0
collisions:0 txqueuelen:1000

(other interfaces are only virtual)

017:~ # cat /proc/sys/net/core/rmem*
112640
112640

017:~ # cat /proc/sys/net/core/wmem*
112640
112640

017:~ # cat /proc/sys/net/core/netdev_max_backlog
300

(Yes, you might claim these values are not the ones you proposed, but i dont get any packetloss or dropped packets, so hey ;-) )
Image
stgraber
Posts: 3
Joined: Mon Aug 08, 2005 5:20 am
Location: Switzerland

Post by stgraber »

And for me (server with gameserver, record, hub) :
eth0 Link encap:Ethernet HWaddr 00:0F:EA:32:03:58
inet addr:194.242.113.29 Bcast:194.242.113.255 Mask:255.255.255.0
UP BROADCAST RUNNING MULTICAST MTU:1500 Metric:1
RX packets:278554940 errors:0 dropped:0 overruns:0 frame:0
TX packets:158002221 errors:0 dropped:0 overruns:0 carrier:0
collisions:0 txqueuelen:1000
RX bytes:3736797851 (3.4 GiB) TX bytes:4049664783 (3.7 GiB)
Interrupt:20 Base address:0x9000 Memory:f8000000-f8000038
stargate-server:~# cat /proc/sys/net/core/rmem*
107520
131071
stargate-server:~# cat /proc/sys/net/core/wmem*
107520
131071
stargate-server:~# cat /proc/sys/net/core/netdev_max_backlog
300
Perform #pr3dators-klan.et and #ettv.fr
poorboy08
Posts: 30
Joined: Mon Jun 06, 2005 6:06 am
Location: MD

Post by poorboy08 »

-=[VP]+arni+=- wrote:ok bani we seem to have localised the problem more:

Matchserver and ettv server were on the same network segment ...

Matchserver b4 & ETTV server b6 - Works perfectly

Matchserver b6 & ETTV server b6 - Causing these strange lags with frames with 50 or 00 at the end get dropped ...
I'm having these same problems with my windoze 2003 server running b6 and connecting to a b6 master.

http://66.82.15.70/ettvproblem00.bmp
http://66.82.15.70/ettvproblem01.bmp
http://66.82.15.70/ettvproblem02.bmp

Please excuse the large bitmap files, but you can see the lag spikes on the client and the *00, *50 on the server. If someone thinks it would be useful I could do a sniffer trace over the network to see how the packets look over the network!?
www.cxtreme.com
CXtreme*doc
Long Sigs Suck!
poorboy08
Posts: 30
Joined: Mon Jun 06, 2005 6:06 am
Location: MD

Post by poorboy08 »

anybody?
www.cxtreme.com
CXtreme*doc
Long Sigs Suck!
poorboy08
Posts: 30
Joined: Mon Jun 06, 2005 6:06 am
Location: MD

Post by poorboy08 »

TTT
www.cxtreme.com
CXtreme*doc
Long Sigs Suck!
chosen_
Posts: 11
Joined: Tue May 24, 2005 8:26 am

Post by chosen_ »

poorboy08, its exactly the same problem as we have, so take the ettv b4 for the master / game server please.

cya
chosen
User avatar
arni
Posts: 188
Joined: Sun Feb 20, 2005 2:32 pm

Post by arni »

bani, zynx: There must be something you changed from b4 to b5/6/7/8 because this problem still persists.

b4 matchserver on b4 broadcast: works perfectly
b4 matchserver on b6 broadcast: works perfectly
b4 matchserver on b7 broadcast: works perfectly
b4 matchserver on b8 broadcast: works perfectly

b6 matchserver on any broadcast: bad lags
b7 matchserver on any broadcast: bad lags
b8 matchserver on any broadcast: bad lags

I've tried it all - mtu, tos, netburst but nothing helps - as soon as more than ~11 people are on the server and the match begins (sometimes even before) matchservers starting from b6 (b5 wasnt tested) cause lags that make ettv no fun at all.

Simplest solution for now:

for all gamestv.org broadcasts we ask admins to install ettv b4 on their matchservers - with this version there were never any problems on broadcasts. Finally this bug is known to almost all ettv matchserver admins and almost all ettv enabled servers (at least in europe) run ettv b4.

If you dont know what it is - well we cant change that but sticking with b4 until a solution is found seems to be the best solution as it works together with all broadcaster versions from b4 to b8.
Image
User avatar
bani
Site Admin
Posts: 2780
Joined: Sun Jul 21, 2002 3:58 am
Contact:

Post by bani »

try mtu 1400 and netblast 0

if i could get ssh access on the broadcast machine and a slave machine that has problems, it would be useful.
User avatar
bani
Site Admin
Posts: 2780
Joined: Sun Jul 21, 2002 3:58 am
Contact:

Post by bani »

if you have 'bad lags' problem with > b4, please post the kernel version running on your master server. eg linux 2.4.2 or 2.6.11.10 etc.

this will help us verify the problem.
User avatar
arni
Posts: 188
Joined: Sun Feb 20, 2005 2:32 pm

Post by arni »

Linux 017 2.6.8-24.10-default #1 Wed Dec 22 11:54:27 UTC 2004 i686 athlon i386 GNU/Linux

--> Thats for my broadcaster

sorry again that i had to leave yesterday, was too tired
Image
User avatar
deej
Posts: 743
Joined: Fri Mar 19, 2004 12:44 am
Location: Belgium!
Contact:

Post by deej »

2.6.13.3-amd64.

Started to lag with over 60 viewers.
Our servers now run on 64 bit steroids. Point your ET to:
- Forgotten Ground StopWatch Server with occasional wolfrof 1
- Fraggle Rock ETPub Server - Mix up ET/UT & Duke Nukem
User avatar
arni
Posts: 188
Joined: Sun Feb 20, 2005 2:32 pm

Post by arni »

deej wrote:Started to lag with over 60 viewers.
erm, but this problem didnt concern number of viewers but data lost on the way from master to slave when it gets too much ...
Image
User avatar
bani
Site Admin
Posts: 2780
Joined: Sun Jul 21, 2002 3:58 am
Contact:

Post by bani »

-=[VP]+arni+=- wrote:Linux 017 2.6.8-24.10-default #1 Wed Dec 22 11:54:27 UTC 2004 i686 athlon i386 GNU/Linux

--> Thats for my broadcaster

sorry again that i had to leave yesterday, was too tired
need the master server not the slave.
User avatar
arni
Posts: 188
Joined: Sun Feb 20, 2005 2:32 pm

Post by arni »

ok gettings masters is not that easy, but here is one that i could get:

Linux pr3dators-klan.com 2.4.28intel #1 SMP Mon Jan 10 15:42:21 CET 2005 i686 GNU/Linux
Image
Post Reply