Ferric is rusting, do you have a de-ruster?
#1
The B} servers ( /connect ferric.tk [1492] [1927] [26283] ) have been executing mass kicks recently without reason it seems. That is the problem, the other problem is that I'm not sure why it does this.

I'll pop up here with full server specs here in a second, but just by memory it has something like 256 mb of ram, and it runs Ubuntu 10.10. afaik the servers are vanilla (almost dead sure).

I seem to recall a TyD server having similar symptoms.

I am going to turn on verbose logging here in a second to see if that gives me anything.

This was also happening on a previous VPS (reason the host was switched) and from memory the previous VPS was running CentOS. I really don't know where to go from here.
Thanks given by:
#2
dude, govpsgo sucks, get a new vps, like amazon.com
or virpus.com
Thanks given by:
#3
Yes, thank you for not actually understanding the problem.
Thanks given by:
#4
I like jg´s league:
Thanks given by:
#5
Do you mean simultaneous disconnect(normal)'s, or does the server actually start kicking people?
Thanks given by:
#6
Former I think.
It explodes and throws everyone out instantly-ish.
Thanks given by:
#7
I have this problem on a VPS running CentOS. At the end of some rounds it kicks everyone off the server. It doesn't happen if the round is ended with a vote instead of reaching the time limit.
Thanks given by:
#8
Seems like a virus or something? Or a type of worm?
Thanks given by:
#9
DAMN OXIDATION.
Thanks given by:
#10
Funny you should mention this because the German TyD servers runs on CentOS.

(11 Apr 12, 03:17AM)Ronald_Reagan Wrote: The B} servers ( /connect ferric.tk [1492] [1927] [26283] ) have been executing mass kicks recently without reason it seems. That is the problem, the other problem is that I'm not sure why it does this.

I'll pop up here with full server specs here in a second, but just by memory it has something like 256 mb of ram, and it runs Ubuntu 10.10. afaik the servers are vanilla (almost dead sure).

I seem to recall a TyD server having similar symptoms.

I am going to turn on verbose logging here in a second to see if that gives me anything.

This was also happening on a previous VPS (reason the host was switched) and from memory the previous VPS was running CentOS. I really don't know where to go from here.

Thanks given by:
#11
i think its an ac bug too, a timer error, my servers have kicked everone before, incl me for this reason.
Thanks given by:
#12
It is a bug, it didn't happen pre 1.1. If it did, not with that frequency anyway. Although I haven't tried compiling the server to see if that makes any difference..
Thanks given by:
#13
Yes, it will randomly boot everyone from the server. iirc the reason given is "normal" which is entirely unhelpful. Everywhere it just looks like you disconnected the server with /disconnect.

Usually it kicks everyone, not always at the same time, sometimes you can reconnect fast enough to see one remaning person get kicked.

We also have 2GB of RAM, not sure where my other number came from.

From what I can see, it is more likely with more than two people in the server. It also doesn't seem to have a trigger.
Thanks given by:
#14
it doesnt happen on custom compiled servers like mine
Thanks given by:
#15
(12 Apr 12, 02:28AM)Jg99 Wrote: it doesnt happen on custom compiled servers like mine
Well it's happening for his server. Thanks for trying though.
Thanks given by:
#16
I am not sure if this helps, but I have noticed that server starts to get very laggy before the kick. The server will struggle for a minute or two and then *plonk*.
Thanks given by:
#17
buffer overflow?
https://en.wikipedia.org/wiki/Buffer_overflow

I remember masterkaen telling me why bloods servers crashed after end of some games..
or maybe he said something else, not sure.
Thanks given by:
#18
Well, this is anywhere during the game, but I'll read up on that.

Isn't buffer overflow a reason the server can give?
Thanks given by:
#19
(13 Apr 12, 12:19AM)Ronald_Reagan Wrote: Isn't buffer overflow a reason the server can give?

not sure, i never had that problem for my servers.
but buffer overflow was just a guess.
Thanks given by:
#20
Hey, that's pretty crazy. I just migrated from a CentOS box and I was having the same issue there. I don't know how but it seems to correlate to the OS for some reason. I don't know much about CentOS specifically, but maybe someone knows of some oddity that would cause it there.

This issue doesn't apply to me anymore, but I'd like to hear if you have any success with a server compiled in your environment rather than the standard binary.
Thanks given by:
#21
Some of MyS' servers have the same problem, the server crashes after some/most rounds. I've heard that it's something to do with Internet spikes but this might not be true, I'm just trying to help but not sure of what I'm saying.
Thanks given by:
#22
From what I can see, the only thing in common is CentOS, except for the current ferric servers are running on Ubuntu.

Do we know what was the original OS that it was compiled on?
Thanks given by:
#23
Snipa mentioned to me that it might be a glibc-related issue if the problem happens only with the pre-compiled binary. I don't know what exactly it would be, but it seems like an easy thing to test.
Thanks given by:
#24
(11 Apr 12, 03:17AM)Ronald_Reagan Wrote: afaik the servers are vanilla (almost dead sure).

I thought the Ferric servers were supposed to be fixed?

Also, are you sure your VPS provider isn't overselling the physical machines? I guess there'd have to be pretty severe contention to actually have players lose connection.
Thanks given by:
#25
These are the pub servers that do this. I'll see what binaries the private servers are running on.
Thanks given by:
#26
I have a couple logs of this happening. First log is pretty much unedited, it shows all interactions a minute before hand. It has some interesting (duplicate connection) disconnects. The server was not full when this happened.

http://pastebin.com/g6qaabCv

Second log is just another occurrence, I took the previous scoreboard to show that the pings are all about the same. Not much else is interesting about this log. Maybe I am missing something. This is a full server.

http://pastebin.com/b34irpRj
Thanks given by:
#27
Some really weird stuff there (apart from the mass disconnects ofc).
First, those "duplicate connection" disconnects - that shouldn't happen at all, at least not unintentionally.
Then, what's this?
Quote:Apr 15 14:40:19: 2 unarmed RVSF 0 202 9 8 1 144 normal xx.xxx.xx.xxx
Apr 15 14:41:13: 4 |AOD|FireFox. CLA 2 270 12 3 0 130 normal xx.xxx.xx.xxx
and
Quote:Apr 16 14:42:23: 0 TheEveMe CLA 0 94 9 18 0 147 normal xxx.xx.xxx.xx
Apr 16 14:42:53: 1 unarmed RVSF 0 44 5 9 1 163 normal xxx.xx.xxx.xx
That's the game status output, which is one loop through all the clients here. That can't just take almost a second between iterations, so what's going on here?
Thanks given by:
#28
My server does massive disconnects when I scroll up and down the ¿system propmt? for a while..
EDIT: Windows 7 , XD
Thanks given by:
#29
Windows? Yeah, I think it actually suspends the process while you're doing that :S
On some kind of *nix, this shouldn't happen though.
Thanks given by:
#30
(20 Apr 12, 12:21PM)tempest Wrote: Some really weird stuff there (apart from the mass disconnects ofc).
First, those "duplicate connection" disconnects - that shouldn't happen at all, at least not unintentionally.
Then, what's this?
Quote:Apr 15 14:40:19: 2 unarmed RVSF 0 202 9 8 1 144 normal xx.xxx.xx.xxx
Apr 15 14:41:13: 4 |AOD|FireFox. CLA 2 270 12 3 0 130 normal xx.xxx.xx.xxx
and
Quote:Apr 16 14:42:23: 0 TheEveMe CLA 0 94 9 18 0 147 normal xxx.xx.xxx.xx
Apr 16 14:42:53: 1 unarmed RVSF 0 44 5 9 1 163 normal xxx.xx.xxx.xx
That's the game status output, which is one loop through all the clients here. That can't just take almost a second between iterations, so what's going on here?

Assuming this is 'loggamestatus', all that function does is play with memory. I only see few reasons: poor implementation of loopv/concatenation, server running out of ram and switching to swapping, ac server contains a memory leak that results in swapping, or just some bug that causes high cpu usage in some cases.

I'm pretty sure server owners will rule swapping out. To the bug hunt!
Thanks given by: