ManiaPlanet date=2012-09-27_13_38 Svn=46718 GameVersion=3.2.0
maniaplanet://#join=fng_sms_01@SMStorm
Ok all. I got some of my crew and some other random players in our TrackMania 2 Canyon Time Attack and ShootMania Royal servers tonight.
Here are my findings... (both servers were running at the same time, but only one was populated at any given time.)
TrackMania 2: Canyon (Time Attack)
===============================
I tested first without any server controllers running, to ensure that wasn't a part of the equation.
I later hooked up my TournaMania app to diagnose the callbacks and response of my server controls within the app to see if I could learn anything else.
I need to clarify that all of us were able to race fine in regards to our own player... we just couldn't see opponents well, and noticed the following symptoms:
- Players had to make multiple attempts before being able to successfully connect to the TM2 server
- After finishing a lap, scores take a long time to update (5-10 seconds, maybe more.)
- When Time Attack map time runs out, game still plays several seconds before going to podium.
- Major chat lag.
- Opponents appear in a spot frozen, then disappear and reappear frozen farther along the track.
- Spectating resulted in the same as mentioned in the previous item.
- CPU maintains 50% usage until track switch (and ~330MB memory used)
- During map switch, CPU usage drops to 0-10% and memory usage ~225MB
- XmlRpc works fine while map is switching (must correlate with the CPU usage going down) -- My TournaMania app reported scores immediately (no chat lag)
- Chat lag returns as soon as the new map starts.
ShootMania: Storm (Royal)
===========================
- While people were joining the server (2 people joining at the same time) the 2 already in the server were experiencing major lag and the same symptoms as opponents in TrackMania 2 listed above.
- When myself and the other person joining the server finally got in, we saw the top-down view of the map, but no gameplay was updating on it. Two full rounds were played and all we saw was the same still show of the top-down (clouds were still moving overhead.) During those two full rounds, the other 2 players continued to play, but it was so jerky/laggy that they couldn't do much. One of them was stuck in a tree for 1/2 a round.
- Timer still appeared for the two players still waiting to start the action (still in spectator view, but not listed as spectators); and the timer was counting down from 36:00 minutes for the auto-offzone activation; and continued to count down until one of the active players got to the pole and activated (then immediately the timer goes to zero) - spectator mode still doesn't update or show gameplay.
- Server at 50% cpu all the time, INCLUDING DURING ROUND CHANGES (not map change, but transitioning to a new round within the same map.)
--- Reset Server ---
After this we decided to stop and restart the server.
4 players doing alright now... CPU is NOT at 50%
Jumped to 7 players - CPU still good.
back down to 5 players 1 map later... CPU still good (no LAG) - Memory at ~170MB
We played for another 45 minutes or so on a couple more maps. We had players jumping in and out from all over the world, maxxed at 6-7 players. We didn't see any problems after this.
=================================================
I'm thinking there is a problem with the XmlRpc server and it is causing a spike that remains constant CPU usage at 50% in both servers. The memory usage is double when the problem is exhibited. I wonder if there might be too much XmlRpc chatter going on when the server is staying at 50% usage and 300+MB memory usage.
I say this because my TournaMania app normally connects and authenticates in less than one second to the XmlRpc server, but tonight when I tied it in with the TrackMania 2 server, it took around 30-40 seconds (I thought it was going to time out.)
It finally connected, but sending server commands was taking similar amounts of time while the maps were being raced. I started a tournament with the app, and when it was executing its code and responding to the EndMatch data, the updates in chat and ManiaLink UI were immediate. I watched the CPU drop to 0-10 percent, but not all the time (looked like the stable server normally does.)
Once the new map started, wait 10-15 seconds and the chat and opponents lag was back again.
The reasons I bring this up are:
- I want to make sure everyone understands we saw the problem go away during transition from one map to another in TrackMania 2: Canyon (Time Attack). We DID NOT see the problem go away during map transition on ShootMania: Storm (Royal).
- I believe Server Controllers like XAseco, MPAseco, Fox Control, etc. will augment the problem, because they're adding to the seemingly XmlRpc "chatter flood".
If I get time in the next week I will try to enable some XmlRpc callback debugging in my app and see if I can confirm more chatter/callback activity when connected to the 2012-09-27 server versus the stable 2012-09-06 version. (That is, unless someone else can confirm before I get to it.)
=================================================
The logs in the "log" folder don't seem to say anything out-of-the-ordinary from normal operations. Do you have other special logs kept elsewhere I can track down and connect you with? I ran the ManiaPlanetServer.exe without using the NoLogs version.
=================================================
System Specs for the server:
----------------------------------
Intel Core 2 CPU (non Core 2 Duo): 1.8GHz
2GB Ram
Windows XP Professional SP3
This box runs both servers just fine with the 2012-09-06 [B1.3] version of the server as well as a Mumble server. We can normally host players in both ManiaPlanet servers and not experience the issues we saw tonight.
I had to revert the TM2 server back to the [B1.3] 2012-09-06 version, but I'll leave the now-working ShootMania server up for anyone else that wishes to connect and give it a shot.
It is in the USA | Utah zone and it is called
"[B1.3c] Team FNG (LAG TEST)"
Join at the following URL:
maniaplanet://#join=fng_sms_01@SMStorm
=================================================
I hope this helps narrow down where the problem is some. I'm happy to provide more info if you need and if I can find it.
Here's to hoping the culprit can be tracked down soon!
Thanks for your continuing hard work Nadeo! You all rock.
