Please have a look at Dedicated server exhibits extreme CPU load, because I think it is what happens, but not in general as I initially thought. It only happens after a while and it might be related to Mills Metropark, because last night we had a session at Outer Loop A and I noticed high CPU load after some coming and going, when only three of us were left all sitting in the garage and having a chat on Teamspeak. Other possible causes from my point of view could be the coming and going of drivers and changing the track at runtime. Maybe a combination, but if I remember correctly, the high CPU usage was first noticed on a quite heavily frequented server that was set to 30 hours test day that had no track change, but I am honestly not sure of that. Let me try and reconstruct last nights events as best I can from memory. The server session had some uptime already, at least a day, CPU usage was at about 6-7% on an Athlon II X4 @ 2.3 GHz. The vMod running points to the Renault Clio Cup Vehicles, Mills Metropark 1.3 (all layouts including Outer Loop A, B and C) and Limerock Park taken from the Skip Barber Racing School mod. The setup for the sessions is: 120 min practice, 30 min qualifying, 15 min warmup, 20 min race. The first time I connected there were 2 other drivers already online. The event was "Limerock Park -- No Chicanes". The server filled up to about 10-15 drivers at some point during practice. Some drivers had left and about 8-10 were still left when we skipped "Straight to race" from qualifying. The race lasted the whole 20 min. During the race some 5-10 more drivers joined, pending an open session, some drivers left again while the race was still going. Whether those were active or pending drivers, I don´t know. After the race was finished a friend in Teamspeak became impatient and did not want to wait for the event switch timeout to tick down. So I logged in on the root server and clicked "Go To Next Race". I must have double clicked or coincidentally click at the same time the server was starting to switch anyways, because instead of Mills Outer Loop A, which was next in line the event after that was loaded. Not wanting to click "Go To Next Race" all the time to do a full loop to the correct track, I used "/callvote event Mills Outer Loop A". I misstyped it so the wrong layout was loaded and I did another "/callvote event", which then switched to the desired layout. The server was not frequented as much as before, I think we never had more than 8 drivers at the same time, with only little coming and going. After the practice time was up was when I noticed the high CPU usage of almost 40% with only us three idle drivers left on the server. There was another server idling on the same machine, its CPU usage was less than 6%. I exited the server exhibiting the high CPU usage, the window disappeared but the program kept running, using exactly 25% CPU, so I guess one thread was stuck in an infinite loop. Having terminated that instance and restarted the server us three joined again and switched the event tot Mills Outer Loop A again. CPU usage was less than 9% now. I think the most likely cause for this is the coming and going of drivers, since that is what both incidents have in common. Best regards, Marcus
I have seen this issue happen again several times now. I guess, it has something to do with the skin download feature or the HTTP server. Because of this, I disabled the auto download of skins on the server: Code: Download Custom Skins="0" // whether to automatically download skins from when clients join server with custom skins For the time being I left the HTTP server enabled, because we had a vMod running. Clients were still trying to updload skins apparently. Their little square grey icons occasionally blinked red with that red line below their names. It took several minutes for this to stop and that icon to turn black. Or was this my client attempting to download their skins? I have seen a lot of "404" messages in quick succession on the console of our root server, indicating that a skin file could not be found. So, I suspect my client knew about the other drivers having custom skins and tried to download them, which could not work, since the server did not download the files due to my config change. And I think, if my client does this, the others do that too. BTW, custom skins that were already on the server could be downloaded and seen by the clients. This time it happened on a server that ran the Clios at Limerock park. So the only constants I can see so far are the Clio mod, coming and going of clients and an enabled HTTP server. So, I disabled the HTTP server entirely now, but since then, the server was pretty much vacant, so I cannot draw any conclusions. It seems unrelated to pure runtime, since the last time it happened the server was freshly restarted, if I remember correctly. I don´t know, if this helps any further in tracking down the issue, but I wanted to share my observations and thoughts. Cheers, Marcus
