[B3.1][DedicatedServer] Memory leak

[Xymph]

Just for sake of posterity.. back when we had this same problem in TMCanyon before ultimately abandoning it, we were running CentOS v5 64b, but after experiencing this issue tried CentOS v6 32b, Gentoo 32b, Ubuntu (whatever the current version was.. 10..?) 64b, before ultimately settling on CentOS v6 64b just for simplicity of configuration and interest of familiarity. No matter the distrib, no matter the arch.. same problem every time.

If I'm not mistaken, aren't all of the aforementioned (besides Gentoo) Debian based to begin with..?

No, CentOS falls under the Red Hat family, not Debian/Ubuntu: http://en.wikipedia.org/wiki/List_of_Li ... tributions

[Bueddl]

Just for sake of posterity.. back when we had this same problem in TMCanyon before ultimately abandoning it, we were running CentOS v5 64b, but after experiencing this issue tried CentOS v6 32b, Gentoo 32b, Ubuntu (whatever the current version was.. 10..?) 64b, before ultimately settling on CentOS v6 64b just for simplicity of configuration and interest of familiarity. No matter the distrib, no matter the arch.. same problem every time.

If I'm not mistaken, aren't all of the aforementioned (besides Gentoo) Debian based to begin with..? Also if I'm not mistaken, judging by the various botched start-up scripts Nadeo has released in the past for the ded server and ManiaLive -- I'd say they are coding for/on Debian themselves.

I'd really like to know how/if Rots is not suffering from this.. Anybody on here with decent enough *nix skills to give him a command to output the mem usage of a pid into a log file..? Never keep in practice long enough to remember syntax, hehe.

The server software seems to be a gcc port of the windows one, which is developed with MS Visual C++ (as the game client, too).

And yes, CentOS is an RHEL derived OS, not a Debian one.
Ubuntu is for sure Debian derived, but it is really strange, that the bug should not occur there.

Commands that can be use to check the memory consumption:

Get the vsz by a certain pid:

Code: Select all

ps -eo pid,vsz | grep YOUR_PID | awk '{ print $2 }'

To get a simple pid, vsz mapping for all ManiaPlanetServer processes

Code: Select all

ps aux | grep ManiaPlanetServer | grep -v grep | awk '{ printf("%u %u\n", $2, $5); }'

a sample logger would be like that:

memlog.sh

Code: Select all

#!/bin/bash

PID=$1
logfile=mylog.$PID.txt
interval=1

while [ "1" ];
do
        ps -eo pid,vsz | grep $PID | awk '{ print $2 }' >> $logfile
        sleep $interval
done

started like that:

Code: Select all

./memlog.sh YOUR_PID

I prefer to use start scripts for such things, lets say you have both in the same directory, e.g. /home/user/scripts

start-memlog:

Code: Select all

sudo start-stop-daemon --start --background --chdir /home/user/scripts --exec /home/user/scripts/memlog.sh -- YOUR_PID

This command can be executed directly over the command line.

Hope this helps.

Kind regards,
Bueddl

[Rots]

I'm already monitoring process data and i see nothing out of the rule

Anyway... its a fresh server instalation (<2 weeks) maybe memory problems show up later

[Bueddl]

I'm already monitoring process data and i see nothing out of the rule

Anyway... its a fresh server instalation (<2 weeks) maybe memory problems show up later

Has nothing to do with that -

Why should that, it's not windows

[Rots]

I know xD i'm newbie on this i said to myself... " who knows, maybe magically server turns bad in some weeks xD "

[Bueddl]

I know xD i'm newbie on this i said to myself... " who knows, maybe magically server turns bad in some weeks xD "

again, thats windows you are talking about :D:D

[nocturne]

No, CentOS falls under the Red Hat family, not Debian/Ubuntu: http://en.wikipedia.org/wiki/List_of_Li ... tributions

Ah, duh.. can't think straight at all lately, damn cold, hehe

So, seems to affect Debian-based distribs, RPM-based distribs, and Gentoo based distribs; in various flavors/versions in i386 and x86_64. I also tried SciLinux v5 x86_64 for awhile with the same issue, though of course it's just CentOS v5 with a few additions.

[Bueddl]

No, CentOS falls under the Red Hat family, not Debian/Ubuntu: http://en.wikipedia.org/wiki/List_of_Li ... tributions

Ah, duh.. can't think straight at all lately, damn cold, hehe

So, seems to affect Debian-based distribs, RPM-based distribs, and Gentoo based distribs; in various flavors/versions in i386 and x86_64. I also tried SciLinux v5 x86_64 for awhile with the same issue, though of course it's just CentOS v5 with a few additions.

No, thats wrong! It seems not to affect Ubuntu x86_64, which is a Debian based distro.

Maybe there is something special with Ubuntu....However, the DedicatedServer was compiled using gcc on an Ubuntu machine!

[nocturne]

Well, as I stated above, I had the same problem with Ubuntu 10 x86_64 in TMCanyon.. Maybe I'll set up a quick vmserver with the newer version and test it myself. Lacking any real info that it does indeed work fine (besides one user's quick statement), I wouldn't rule it out.

Heck, if I was positive it worked.. I'd of already migrated to Ubuntu/Debian. Years of experience on RHEL wouldn't mean anything, versus being able to actually run the dedicated server as it should.