A possible reason for host crashes?

[Swuul]

Could the problem described in this post be connected to why the host seems to be (AFAICS that is) at the highest risk to crash while saving in CK MP? The save file in CK tend to grow big and saving tends to take quite a while. In my experience the crash rate is higher when the clients chat while the host is saving (the host is totally frozen when saving; clients see this as they get latency reports regarding the host), and if I understand right the problem, the host computer does get then severely out of sync.

Are there any tech-gurus around here who could tell if these are related? And if they are, is there any way to avoid the problem (except the method where the host tells the clients to do absolutely nothing, including chatting with each other, when he saves)?

 

[jpd]

I doubt it, unless of course all of the players are actually using NAT enabled routers. In the thread you refer to, that's the key trigger. It's not actually the game engine itself that needs the constant data feed, it's the translation tables in your NAT router that need to be refreshed and updated continuously.

If you pass the timeout trigger time for inactivity in your router, the router simply drops the translation entry from the table, and from then on communications from other PC's will no longer arrive on the PC behind said router. This is actually WAD behaviour for the router, to deal with crashing applications, PC's that have spontaneously been turned of, restarted or whatever.

If the router would not do this, the translation table would fill up very, very quickly (as there are only 65535 TCP ports, and of these a lot cannot be used), and the router must share these TCP ports for every PC behind it. So the translation table is maintained and purged aggressively.