rpgxEF server difficulties - Asteroid Field/Vanderbilt [CLOSED]

Started by Veritas, July 13, 2013, 04:46:40 pm

Previous topic - Next topic

Veritas

Hi all,

I'm posting this here and cross-posting it to UberGames and EscapedTurkey to get some input from the developers and other server runners.

As many of you know, we've recently had some pretty serious issues with certain maps on the RPGXEF server, particularly during SRPs (particularly particularly during Asteroid Field and on the vanderbilt_v3 map). The problems started some months ago, but seem now to have escalated (with the growing popularity of the SRP, in my opinion) to the point where we sometimes can't actually run AF as we had planned to. I'm posting this here in the hopes that others will let me know if they've experienced similar issues, and so that the developers and those with mapmaking knowledge can help us figure out where to go from here.

Essentially, the issue is thus: We run an RPGXEF server (version 8.4.4, though we could upgrade if it were stable/fixed the problems we have) on EscapedTurkey with 512 MB of RAM available. The server seems to work well on most maps (has anyone had issues with it?) and at any player number <=10 (<= 15 on some maps). However, under the right conditions - big maps like Iowa or Vanderbilt, and lots of players (10 or more usually), the following happens:

When the map is changed, some players get in and some don't. Who gets disconnected seems essentially random, and it's usually "a reliable command was cycled out," "singal loss" or some other silly error like that. Sometimes, everyone gets in, sometimes, nobody does.

Anyone who attempts to reconnect after this is stuck at "awaiting gamestate" perpetually and never actually manages to join the server - occasionally somebody gets through but again, it seems essentially random as to who.

The only remedy to this is to restart the server, load a reasonably non-taxing map (like Poseidon) and try again once everyone has reconnected - and sometimes even this doesn't work and the map is completely inaccessible.

What seems to be happening, as far as I can tell, is that rpgxEF is loading the map but then never actually triggering the players' clients to do the same. I'm not sure what's going on network-side or server-side to fix this, and the logs really aren't any help.

I figured I'd open it up to all of you, then: Have you had similar issues? Did anything work to fix it in your case?

Any and all suggestions are appreciated.

John Adams

I have it a lot. At first I was wondering if it was because I had 6 GB of ram allocated to RPGXEF.

Veritas

Probably not, John. But you really don't need more than 1024MB allocated to RPGXEF at any given time; it's highly unlikely to use more than that.

GSIO01

Actually you can't allocate more then 1024mb to it as far as I remember, you can set the cvar to more but the game will always only allocate 1024mb as maximum.

As for the issues you describe I must take some time and think about it. I can't tell you if a newer rpgxEF version will solve this it would require testing.

Grizz

For the record, we got Vanderbilt V2 to work with 10 players. Reloaded the map. Reconnections. Button mashing, All good.

GSIO01

So it comes down to what is different between those versions of the map to narrow down the cause of the problems.

Martin Thompson

Probably has to do with the size of the map and the manner of building. The map is verry taxing on system and has no optimisations at al (like area portals etc.). Poseidon B however does have those optimisations and therefore runs like a baby (though Vanderbilt is also a bit bigger then Possy-B). The addition of the promende is probaby (like Grizz already speculated) the killer there. Large open area with no map optimisations at all. The AMD bug also pop up on vanderbilt for me no matter the fix, when looking at certain directions the promenade is drawn (you can see that with the r_showtris or r_drawtris command). Dauntless had the same problem there (the entire map almost always gets drawn) but again, it was a smaller map. This is however a client side issue, but a large map in combination with many players might just be too taxing for the engine in its current state.

Anyways, GSI, i dont know if its possible for you, but maybe you can trace down the bits in the code where the "A reliable command was cycled out" errors pop up and see where they are comming from (but you probably thought of that yourself already :P ).


Veritas

I've reached out to TF, the map developer, but have not heard anything from him yet. Hopefully he can comment on this further.

Griffwald

I wonder if there's hope for v3 being edited and working somehow.

Veritas

That would be ideal; if ~TF~ isn't able to do it, we certainly have mappers who could look into it, if he's willing to let us.

Martin Thompson

Quote from: Veritas on July 14, 2013, 11:35:15 am
That would be ideal; if ~TF~ isn't able to do it, we certainly have mappers who could look into it, if he's willing to let us.


*cough*


Veritas

I've heard from ~TF~... he confirmed my suspicions that it's a severe size issue. He is working on it but wasn't able to say if he'd be able to make enough improvements to counteract the map size.

I urged him strongly to sign up here and join the discussion, so I hope he does that. I also let him know that people have expressed interest in helping him work on the issues.

SFC3

If we can do a total collab we can all try to make the ultimate Vanderbilt "Pioneer Station" map.

Veritas

I've just heard again from ~TF~; he's working on the map and thinks that if we used RPG-X 2.3 and he redid the map to make better use of LUA scripting we might have much better results... this would require a server/RPGXCE update and I have no idea whether it would be stable. What do you guys think?

John Adams

Go ahead with it, I've been testing RPG-X 2.3 and it seems stable enough.