Low max player count Call of Duty 2 server

[snoopdoge90]

With the Corona virus being out I just got a lot more time on my hand. Decided to finally finish my little mod and make a permanent COD2 server. However, only 12-13 players can play at the same time. When more players connect they have a ping of 999. Someone with COD2 server hosting experience has an idea whats going on?

I'm running a dedicated COD2 1.3 server and Voron's libcod builded against the latest source. It's running only 1 additional custom map and 1 small mod I wrote myself. The server is a virtual machine running on ESXi 6.7 and has 4 GB RAM allocated and plenty of storage left. CPU usage is only 10%, network usage around 2 mbps avarage, and almost no disk activity when 12-13 players are playing. The server is behind an enterprise firewall / router that served 100 concurrent connections with no hiccups in the past, and should easily do a multitude of that on paper. The internet connection is a 500/500 mbps connection and the COD2 server has its own IP address. The hardware firewall is not limiting connections. The software firewall on Ubuntu (UFW) is a standard configuration with a few alterations (allow outcoming, deny incoming except the COD server ports).

Is there some ESXi or Ubuntu (UFW) setting that I'm missing? Is it possible my ISP is limiting incoming connections per IP? Anyone knows a tool to test this without needing to find >12 clients on different network locations?

[snoopdoge90]

Well I ran a TCP HTTP concurrent connection test everything seems to be okay on concurrent connections. Still don't know a tool for testing concurrent UDP connections but I recon if TCP isn't the problem UDP would be fine either. I'm completely in the dark here, so I've just set the server power management to high performance mode. I can recall a few scenarios in which applications would perform bad on a balanced profile despite average overall system load (CPU, RAM, storage and network). However, running a COD2 server comes nowhere near the performance required by those applications but trying can't hurt (maybe only the electricity bill).

For my setup, that's configuring the HP server BIOS power settings to static high performance mode (can be done via iLO) and and setting the ESXi power management profile to high performance (can be done via vSphere or the Web UI on standalones). Guest VMs should inherit power management settings from the VM host in this case. Hope this works...

[IzNoGoD]

999 ping should not be happening in the situation you laid out. Please try running your server with developer 1 and see if there's any infinite loops detected in your script at those playercounts.

[snoopdoge90]

999 ping should not be happening in the situation you laid out. Please try running your server with developer 1 and see if there's any infinite loops detected in your script at those playercounts.

Nope log didn't show anything out of the usual if i recall correctly but I don't have the log anymore. It was at the beginning when developer was set to 1 and it was not running any mods. Maybe I overlooked something so I'll set developer 1 tonight to confirm if I'm correct on this statement.

Whats the logic behind endless loops causing 999 ping by the way? The server is now running with my own written mod tho, but no endless loops are made before the player spawns into the game (playerSpawn callback), as spectators immediately get 999 ping. Those loops run in separate threats and are always end on disconnect, killed_player (self) or intermission (level). It's a very basic modification.

The funny thing is when this player count is reached, the new player gets 999 ping, and takes one other connected player with him. All returns back to normal if the excess player disconnects. Back in the days like 2004 COD2 1.0 I tried to run a server on my crappy ADSL connection. The 999 ping thing also happened when too many players connected. But in this case, all players experienced lag when the excess player connected to the game. It could only handle like 9 players so that's when rented a server at I3D. Those were awesome times .

Edit: problem seems solved, no abnormalities in log, and disabling the power management settings in both ESXi and host bios resolved the issues on my VM. Despite host ready CPU was only at 10%, with minimal IO and network usage.