07-18-2006, 10:21 AM
(This post was last modified: 07-19-2006, 10:32 PM by simonlange.)
Well, the topic says it.
But let me begin in the past.
Once we had a Dual AthlonMP 2200+ with 1.5GB RAM (ECC Registered)
Linux Distribution was Fedora Core1
These processes did run on the machine:
hlds (40ports cs1.6) with clanmod, steambans, vac
srcds (40ports css) with maniadmin, steambans, vac
srcds (40ports css) with maniadmin, steambans, vac
srcds (12ports css) cvarblock, vac
hlds (12ports cs1.6) vac
...for the gamepart.
also these:
exim4
dbmail lmtp daemon
spamassassin
mysqld
spfd
lighttpd
proftpd
webmin
...and some selfmade daemons which did watch out for intrusion attemps and such.
Result: NO LAGS. Top Performance. No Problems so far. And even that although srcds or hlds DOES NOT participate of symetric processor architectures.
then we moved to another ISP (better peering) and a new hardware.
Now we using:
a single processor Opteron148 with 2GB RAM.
Distribution is Fedora Core 3
We have running these processes:
srcds (20 ports css) maniadmin, steambans, vac
srcds (12 ports css) cvarblock, vac
srcds (12 ports css) cvarblock, vac
...for the gamepart.
also these:
proftpd
but now we HAVE lags. terrible long lags. and we cannot find out why.
when a lag accurs we have these information:
400mb used of physical ram (2gb existing)
0mb of 1024mb swap used
cpu usually under 10%
at lagtime srcds consumes 100% no matter how many ppl are on the server, what map or what rates the users are using.
while lag, the icmp ping is stable and we can eliminate the existance of dos attacks.
all warservers (old and new hardware) did run on tickrate 100.
we think that valve has some serious problems with 64bit system since its the only thing we did change. smp systems are not used by valve. valve only consumes cputime of the first cpu in a smp system.
any ideas are welcome.
best regards
Simon
But let me begin in the past.
Once we had a Dual AthlonMP 2200+ with 1.5GB RAM (ECC Registered)
Linux Distribution was Fedora Core1
These processes did run on the machine:
hlds (40ports cs1.6) with clanmod, steambans, vac
srcds (40ports css) with maniadmin, steambans, vac
srcds (40ports css) with maniadmin, steambans, vac
srcds (12ports css) cvarblock, vac
hlds (12ports cs1.6) vac
...for the gamepart.
also these:
exim4
dbmail lmtp daemon
spamassassin
mysqld
spfd
lighttpd
proftpd
webmin
...and some selfmade daemons which did watch out for intrusion attemps and such.
Result: NO LAGS. Top Performance. No Problems so far. And even that although srcds or hlds DOES NOT participate of symetric processor architectures.
then we moved to another ISP (better peering) and a new hardware.
Now we using:
a single processor Opteron148 with 2GB RAM.
Distribution is Fedora Core 3
We have running these processes:
srcds (20 ports css) maniadmin, steambans, vac
srcds (12 ports css) cvarblock, vac
srcds (12 ports css) cvarblock, vac
...for the gamepart.
also these:
proftpd
but now we HAVE lags. terrible long lags. and we cannot find out why.
when a lag accurs we have these information:
400mb used of physical ram (2gb existing)
0mb of 1024mb swap used
cpu usually under 10%
at lagtime srcds consumes 100% no matter how many ppl are on the server, what map or what rates the users are using.
while lag, the icmp ping is stable and we can eliminate the existance of dos attacks.
all warservers (old and new hardware) did run on tickrate 100.
we think that valve has some serious problems with 64bit system since its the only thing we did change. smp systems are not used by valve. valve only consumes cputime of the first cpu in a smp system.
any ideas are welcome.
best regards
Simon
theCENTER netWork Head Administrator
[url=\"http://esports.the-center.at/hlstats/index.php?mode=playerinfo&player=1662\" ][/url]
[url=\"http://esports.the-center.at/hlstats/index.php?mode=playerinfo&player=1662\" ][/url]