I just spent all morning working on this and talking to several admins. It is not due to the rate 30000 bug, as a workaround was found and does not solve it.
net_graph 5 shows a ton of net lag - when you're not viewing a lot of action (even if a lot is occuring elsewhere) there's no issue, but any spike in bandwidth usage is 'faded' on the netgrpah - not being sent completely. I have plenty of bandwidth. My server has plenty of bandwidth. This occurs for everyone, and started with the patch. http://www.infernalsoul.net/Junk/Dump/whoknows.png I switched off sourcetv last night, upped my servers to realtime scheduling, upgraded the kernel. They're running great. "var" rarely goes over 2ms due to my nanosleep() hook (connect and see!) - there is something limiting bandwidth. It happens to me, and every other admin i've talked to. I've disabled my performance plugin that hooks things. I've disabled sourcemod. No effect. SourceTV is off. The combination of: - "rate" being bugged and limiting to 30000 if sv_minrate isn't 0 - net_splitpacket_maxrate and net_splitrate being introduced - sv_client_cmdrate_difference working differently now (i suddenly was being forced to 40 cmdrate without changing any server settings after the patch due to this) Suggests valve changed a lot in the networking side of things. Well something is causing my update rate to drop to 30, 25, 45, every time there's even moderate action. The server FPS stays stable. var stays stable. CPU usage stays stable. I've tried: - setting rate/maxrate to ninty bajillion - messing with every cmd/updaterate cvar - turning up net_splitrate and net_splitpacket_maxrate - messing with every interp setting i could Any ideas would be welcome. the slew of related cvars suggests there might be something on the configuration end that might improve things. - Neph _______________________________________________ To unsubscribe, edit your list preferences, or view the list archives, please visit: http://list.valvesoftware.com/mailman/listinfo/hlds
