ETPro Server - Recursive Crash Error

Discussion for any ET/ETPro/BayonET bugs or cheats you find...

Moderators: Forum moderators, developers

Post Reply
MCGOOHQ-Michael
Posts: 7
Joined: Mon May 24, 2004 6:32 pm
Location: Canberra, Australia
Contact:

ETPro Server - Recursive Crash Error

Post by MCGOOHQ-Michael »

Hi guys,

We operate a game server rental company here in Australia. We have just taken on our first ET client, and have installed ETPro on the server at their request.

We are finding that the server (ET instance not the box) is having some serious reliability issues. I have read a few ET tutorials that suggest that the dedicated server is unreliable and to expect 1 crash per 24 hours - we are experiencing 4 - 5 per day.

This seems to be the most common issue when running the ET server:

-------- UNRECOVERABLE ERROR --------
This may be due to a bug in etpro
Information to be used in a bug report is being generated:
------------- CUT HERE --------------
Version: etpro 3.0.1
Platform: Linux
Signal: Segmentation violation (11)
Stack trace: 1 entries
/games/et/dark/etpro/qagame.mp.i386.so[0x5afd25d7]
------------- CUT HERE --------------
Trying to clean up...
---------- RECURSIVE CRASH ----------
Exiting hard.
Aborted

Just wondering if anyone can tell me if they are having the same issues, any possible solutions, or if this is just normal performance.

Not that it is relevant to this issue, but I will explain:
For all of our game server hosting we run Dual Opteron 64 bit servers with at least 4 gigs of ram. We are testing the new Fedora Core under 64 bit architecture.

The installer for ET crashes (as all quake3 installers do – they detect the 64 bit glibc first and give up rather than looking for the 32 bit) on the machines – this is normal. We run the installer with - -keep and manually copy the files into the correct directory structure.

We set up the server from a fresh install of ET Linux 2.55, patched to 2.56, and then installed ETPro. The ET server runs fine, players can join, everything is peachy – but we have the stability issue.
MCGOOHQ Internet Communications
http://www.mcgoohq.com.au
Ph: 1300 889 461
E-mail: sales@mcgoohq.com.au
User avatar
Badhabit
Posts: 312
Joined: Sat Oct 26, 2002 9:09 pm
Location: Boise, Idaho

Post by Badhabit »

I'm running 2 ETPro servers and a RTCW server off a redhat9 box and haven't seen it crash in 3+ months of running ETPro.
But it's not a 64bit system, so.... :?
{Zer0}'s House of Torment 67.19.67.118:27960
{Zer0}'s RTCW server 67.19.67.119:27960
Now Lets Go Kick Some ASS
And thats an Order.
Image
MCGOOHQ-Michael
Posts: 7
Joined: Mon May 24, 2004 6:32 pm
Location: Canberra, Australia
Contact:

Post by MCGOOHQ-Michael »

Being a 64 bit system has nothing to do with it - it is still fully 32 bit compatible and runs fine using the 32 bit architecture. Can anyone tell us anything about this error?
MCGOOHQ Internet Communications
http://www.mcgoohq.com.au
Ph: 1300 889 461
E-mail: sales@mcgoohq.com.au
User avatar
zinx
Posts: 267
Joined: Fri Jan 16, 2004 12:37 pm
Location: US
Contact:

Post by zinx »

You definitely should not be expecting even 1 crash per day, the server tends to stay up a while, even if it starts acting funny after it's been up that long.
Is there anything in the log before that when it crashes, or just chats/death/objective messages? (sadly, the stack trace is useless)
Zinx Verituse http://zinx.xmms.org/
MCGOOHQ-Michael
Posts: 7
Joined: Mon May 24, 2004 6:32 pm
Location: Canberra, Australia
Contact:

Post by MCGOOHQ-Michael »

No there is nothing before that in the log / console.

http://tjw.org/etded/
"STEP #3: Startup Scripts
As of the time of writing, the current relase of etded (2.55) is very unstable and crashes at least daily for me. Splash Damage is working on a fix, but in the mean time, you practically NEED some type of process to keep starting etded."

Any other ideas?
MCGOOHQ Internet Communications
http://www.mcgoohq.com.au
Ph: 1300 889 461
E-mail: sales@mcgoohq.com.au
User avatar
bani
Site Admin
Posts: 2780
Joined: Sun Jul 21, 2002 3:58 am
Contact:

Post by bani »

32bit emulation on 64bit linux is still a work in progress. there are bound to be bugs in it.

i would suggest trying to run the server in fully 32bit mode instead of 64bit mode with 32bit emulation, and see if it helps.

and unless you are running 64bit server binaries (which doesnt exist for q3, et, rtcw, etc), there's likely very little advantage to running in 64bit mode... in fact there may be a performance decrease due to the remapping and copying for 32bit->64bit syscalls.
MCGOOHQ-Michael
Posts: 7
Joined: Mon May 24, 2004 6:32 pm
Location: Canberra, Australia
Contact:

Post by MCGOOHQ-Michael »

You dont understand. The server has both libraries for 32 and 64 bit operation - there is no "emulation". It has nothing to do with this problem.
MCGOOHQ Internet Communications
http://www.mcgoohq.com.au
Ph: 1300 889 461
E-mail: sales@mcgoohq.com.au
User avatar
bani
Site Admin
Posts: 2780
Joined: Sun Jul 21, 2002 3:58 am
Contact:

Post by bani »

While it is certainly true that 32bit instructions runs 'natively' on 64bit linux, it is also true that there is an emulation layer which allows the 32bit api access to 64bit resources.

So while 32bit applications are not emulated, their access to the system (memory, i/o, etc) most certainly is. 32bit syscalls are translated to 64bit syscalls in arch/x86_64/ia32/ and some of this is incomplete.

So ia32 binary crashes on x86_64 are not entirely suprising. this is very new territory for x86 and linux and there are still bugs.

This is not even counting the kernel bugs which may be specific to x86_64!

Please, try a fully native ia32 install and try again (or as you seem to want to do, just ignore me and go about your business as usual and dismiss my suggestions totally out of hand).

And FWIW I have an Opteron system myself running FC2 in x86_64 mode.
MCGOOHQ-Michael
Posts: 7
Joined: Mon May 24, 2004 6:32 pm
Location: Canberra, Australia
Contact:

Post by MCGOOHQ-Michael »

I didn’t mean to come across hostile and I apologize for coming across that way.

We run other Quake3 based servers on our machines without problems – MOHAAS, SOF2 etc. Which is why I believe this issue is isolated to ET.
MCGOOHQ Internet Communications
http://www.mcgoohq.com.au
Ph: 1300 889 461
E-mail: sales@mcgoohq.com.au
User avatar
bani
Site Admin
Posts: 2780
Joined: Sun Jul 21, 2002 3:58 am
Contact:

Post by bani »

do you have ia32 system you can test on? it would be helpful to prove one way or another if its an ia32-x86_64 emulation issue or not.
User avatar
MrDeath
Posts: 3
Joined: Mon Apr 12, 2004 4:17 am

Post by MrDeath »

Version: etpro 3.0.1
Platform: Linux
Signal: Segmentation violation (11)
Stack trace: 1 entries
/home/et/server/etpro/qagame.mp.i386.so[0x4832cbf7]

plain 32bit linux server. wtf.

installed 3.0.1.b5, crash, back to 3.0.1 still crash. :(
We'll meet again.
MCGOOHQ-Michael
Posts: 7
Joined: Mon May 24, 2004 6:32 pm
Location: Canberra, Australia
Contact:

Post by MCGOOHQ-Michael »

We dont have any non-64 bit machines to test that on atm. However we will be purchasing some new P4 3.2 EEs next week. I will report wether we have the same problems with those as we are with the opterons.
MCGOOHQ Internet Communications
http://www.mcgoohq.com.au
Ph: 1300 889 461
E-mail: sales@mcgoohq.com.au
BOTA
Posts: 136
Joined: Wed Oct 23, 2002 3:53 am

Post by BOTA »

Maybe everyone with special machines should set up shell accounts for the crew to test the new code on.
MCGOOHQ-Michael
Posts: 7
Joined: Mon May 24, 2004 6:32 pm
Location: Canberra, Australia
Contact:

Post by MCGOOHQ-Michael »

we would be more than happy to offer this option to any of the devs that would like to test out the code on state-of-the-art hardware.
MCGOOHQ Internet Communications
http://www.mcgoohq.com.au
Ph: 1300 889 461
E-mail: sales@mcgoohq.com.au
Post Reply