Servers dying?

quspiv

Im getting constant "server not responding msg" and dc's. No issues in other online games, so i guess it's not my ISP.

suicidalgodot

...sometimes it's only the char server (I presume - and assume tht your character file is stored on a separate server from the login and the game server) being overly challenged.

As to why I get to that presumptive conclusion: I found that this frequent discon issue often magically disappears when I change to a different character, and the probability for this wonder-cure seems to increase when the char I change to has been created at a different time than the bugging one, and/or has not been played intensely in recent time.

What also sometimes remedy this: A client relaunch with "Force verify" ticked on in the launcher "options" menu. No opinion on as to why this latter thing is the case, as in principle this should, according to the filenames being cued during the process, only verify the integrity of the game engine files. But it sometimes does the trick...

kvet

I get SNRs mostly in the days after maint when some on-demand patch is running. Also, the patch server, if it happens to be overloaded or otherwise running slowly for any reason, can play havoc with your game. Just try having Live up while actively patching Preview. Things sorta work, but no assets that you haven't already loaded and cached will load. The gameclient.exe sometimes hangs and crashes at that point, but not always. If you try to change instances, you'd get an SNR error and an eventual timeout. If you stop the patching process, though, it all clears up instantly.

I don't have any way to prove it, but I think memory leaks caused by an incomplete garbage cleanup in instances is a serious contender for the root cause here. When you go to *new* instances (an instance recycles itself only once there's no players in it), you're much less likely to get lag and SNR and the inverse is true for "old" instances that haven't recycled in a while. Sorta like keeping windows running for too long - sometimes you really just need to reboot.

It would probably behoove us if they could somehow create a TTL for instances - after X amount of time has passed, a given instance simply stops allowing new connections no matter what and once it bleeds down to 0, it resets and can accept new connections. One problem we have in this case, I betcha, is bots. A bot will hold an instance open indefinitely (until maintenance, anyway) and prevent it from naturally recycling...