Page 1 of 2

Map change ucc-bin goes to 99%

Posted: 26-08-2004 21:14
by Imo
Has anyone seen it when there are quite a few people connected (> 10) the ucc-bin process goes to 99% and the map takes ages and ages to load?

Posted: 26-08-2004 21:47
by FL-
On the BuD server every single night...

Posted: 27-08-2004 00:11
by Imo
any solution? seems to be fine when less than 10 connected.

do you run multiple servers on that machine?

Posted: 27-08-2004 12:39
by Messy
Ask Imag ;p

Posted: 27-08-2004 13:27
by Imo
kthx - I think he may check back here some time. :)

Posted: 28-08-2004 07:27
by Imaginos
Unfortunately, I still see this problem and it drives me nuts.
I have no way to monitor the console or live cpu usage on my servers (gg recongamer), but I suspect that is what I am experiencing too.

Posted: 29-08-2004 14:50
by Imo
I can't seem to remember it happening on the 2.0 servers but I wasn't running that for too long.

Posted: 29-08-2004 16:16
by Imaginos
And yes, the box does run multiple servers. Wish I could do something about that, but I can't.

Posted: 03-09-2004 23:54
by Imaginos
I'm about to test patch UT2004 beta patch 3321 on the server. I hope to God that helps with the slooooooow mapchanges on my Windows-based server.

Posted: 12-09-2004 20:51
by Imaginos
Gah.. 3323 didn't help either. Still looking for an answer though.

Posted: 13-09-2004 09:44
by Imo
How close are your port numbers on the servers?

I had... 7777, 7877, 7977. I changed the 7977 one to 10000 and haven't notice the delay since.

Posted: 13-09-2004 11:02
by Imaginos
I'm still digging for an answer. Here's some of the latest leads I'm working on.
Yesterday 01:51 PM
Imaginos: Hey all, thought I would table this here.. I (and several others) have deathball servers that have been suffering from 2min+ mapchanges at times when the server is moderately full to completely full. I have no way to monitor the live cpu usage, but is there any way to log and spot the cause of the delays? I'm getting to my wits' end.

Yesterday 04:03 PM
ProAsm: This problem is a typical DeathBall problem when there are more than 14 players and its not map or mod dependant.
Our opposition used to have a DeathBall server and they had to limit it to 12 players else the server takes so long to change levels that the players leave.
It has also been found that when there are around 16 players on a DeathBall server the server lags so bad some players leave.

Yesterday 04:17 PM
Imaginos: I was afraid of that. Getting a lot of complaints about it and hearing from too many other admins as well.. I don't know what could cause it.. or how.. I would imagine that even if the mod code itself is shoddy, the game performance would suffer, certainly. But to slow down the mapchange too? Seems like it's not even related to the deathball coding itself, unless it's lazy in shutting some things down?

Yesterday 04:27 PM
ProAsm: Epic started a whole new level change and map detection method in UT2003 already and is mainly aimed at WebAdmin.
This change has now been completed in UT2004 with the last patch or the one before that is where all these long changes started as you can remember they are even in UT2004 under windows and thats why I had to add the extra time WaitForLevel=90 in UT2Vote to cater for this so no doubt if its this bad in UT2004, then in DeathBall it must be a nightmare.
I doubt if its to do with the coding in DeathBall, as I think all "custom" games are effected, just some more than the others.
Dont use MapLists as that will worsen the effect, also avoid maps that have custom utx packages and I know DeathBall has plenty of those.

Yesterday 05:04 PM
Imaginos: Sadly, deathball is nearly completly custom content. The developer team has been going out of the way to use absolutely NO epic content at all. I hope there's something that can be done to at least improve things in this situation..
I've just subscribed to the ut2004mods list to see if there's any info to see there. Hopefully something good will come from this.

Yesterday 05:45 PM
ProAsm: I see in the latest Beta Patch they have done something on these lines - lets hope its an improvement.

I wonder are any of you guys still on the AdminBeta List and if so could you PLEASE make a request as a matter of urgency for Epic to supply just ONE tiny map for VCTF.
This game exists in the XGame and many many people asked for the Game to be added to UT2Vote and if players vote for this game and there is no map for it, the server will crash.

Why on earth they release a Game without a map is anyone's guess and is a rather dumb thing to do.

I believe the patch will be officially released next week sometime.

Today 05:55 AM
Imaginos: Nuts. Another dead end four us. I've been running that beta 3323 patch since it came out. Maps that use Epic's textures have a better chance of loading quicker?

Posted: 15-09-2004 17:06
by Imo
Imaginos, I've not had this problem for a while now. What player number limit do you have on the servers?

When I had a 32 player limit with 27 people on there it did take a long time to load the map etc.. long enough for me to think the server died whilst loading - probably everone else too.

Not see the problem since changing that to a 12 player server. But also I changed a few other things.

Also, out of interest, how many servers are you running on the one machine and what are the ip/port details.

Posted: 15-09-2004 22:06
by Imaginos
Hi Imo, the problem is pretty consistant from 8 players and higher. I currently run a 10-man and a 12-man server, both of them on the default ports. I can try an alternate port and see what happens..

These are rented from a commercial host. I do not know how many (or if) there are other servers on the same box. I assume there are.

Posted: 16-09-2004 12:51
by 1234Bedlam
Its nothing to do with the hardware is it? i.e. single CPU vs dual CPUs?