Linux Server Bugs?

p00n1s

<div class="IPBDescription">Just what the title implies ;)</div> I consider myself a fairly advanced server op, I've run a total of 5 servers of the last 2 years and 2 currently. One is a dedicated CS server, the other is CS right now but I'd like it to be NS. They are both running on *nix boxes (although I have experience running a Windoze server as well) - one is on Mandrake 9.0 (CS only) and the other that I'm here to discuss is on FreeBSD 4.6 Stable (Both CS/NS). For some reason though, the FreeBSD box won't run NS properly. Just 2 people sitting in the ready room is enough to trigger terrible lag spikes with pings spiking into the 500s. However, running CS on the same server with the same hlds installation returns no problems with pings consistently under 150ms even with the number of players reaching 20. I'm sure its not the connection (5xT1s linking to a redundant OC-x Qwest firewalled backbone) and I highly doubt its the box (although its not dedicated, average monthly server loads for the other users is ~2% and its a P4 1.7ghz w/ 768 RAM). What bothers me is the fact that it only does this under NS and not under CS - leading me, of course, to believe its a mod-oriented bug and not the hlds files. To test this I tar'ed up all the files and used the exact same hlds/cs/ns files on my other box - where I had an identical experience (no problems on CS, horrible NS performance). I tried re-downloading/installing first the mod files, and then the entire hlds server files AND the mod files together - both times returning no satisfactory results. In addition, after trying new server files and mod files I tar'ed it all up again and I tried it on my linux box once more with no improvement. <!--emo&???--><img src='http://www.unknownworlds.com/forums/html/emoticons/confused.gif' border='0' valign='absmiddle' alt='confused.gif'><!--endemo-->

I've heard this is a bug with the Linux server but haven't heard too much about it on these forums - I don't know where to draw the line between rumors/excuses and a real bug. I have seen Linux servers out there though which makes me think it IS possible to get this working. <!--emo&:D--><img src='http://www.unknownworlds.com/forums/html/emoticons/biggrin.gif' border='0' valign='absmiddle' alt='biggrin.gif'><!--endemo-->

I'm quite frustrated <!--emo&:angry:--><img src='http://www.unknownworlds.com/forums/html/emoticons/mad.gif' border='0' valign='absmiddle' alt='mad.gif'><!--endemo-->

I love NS, and while I'm here posting, must say its an excellent contribution to the HL world and definitely a breakthrough in the entire gaming industry. And living in NC makes me glad to see the UNC PS used too <!--emo&;)--><img src='http://www.unknownworlds.com/forums/html/emoticons/wink.gif' border='0' valign='absmiddle' alt='wink.gif'><!--endemo--> (Duke fan though <!--emo&???--><img src='http://www.unknownworlds.com/forums/html/emoticons/confused.gif' border='0' valign='absmiddle' alt='confused.gif'><!--endemo--> ) I REALLY want to get this working <!--emo&:D--><img src='http://www.unknownworlds.com/forums/html/emoticons/biggrin.gif' border='0' valign='absmiddle' alt='biggrin.gif'><!--endemo--> Advice will be greatly appreciated.

p00n1s

Re-reading my post I believe that I didn't make my main point very clear. I'm aware of the CPU performance issues - especially on *nix machines but my box seems to be getting it substantially worse then most other linux/unix servers. I've been on servers where the pings are bad and its laggy at ~16 players - but I'm talking lag where you can barely walk across the room when there are <4 people on and no structures built. This seems to be a lot worse for me then what others are describing the performance issues to be - and I seem to have a better a connection and box then a lot of them. I was wondering if there are any actual bugs rather then just poor performance - especially on the Unix/BSD side of things. (My problem was definitely not as bad on the Mandrake machine, although still unplayable). Perhaps it was just a fault configuration on my part? <!--emo&???--><img src='http://www.unknownworlds.com/forums/html/emoticons/confused.gif' border='0' valign='absmiddle' alt='confused.gif'><!--endemo-->

verbose

There is currently a Linux high-CPU-usage problem, but it's not severe to the degree that you're experiencing. How's the CPU usage when NS is running (avoid the bast and hera maps while testing, as they're more CPU-intensive)? Any odd messages in your syslog? Do you run any additional plugins (metamod's, etc.)? When you say "average monthly server loads for the other users is ~2%", do you mean the load average as shown by <b>w</b> and other utilities is 0.02 (it's not a percentage)?

verbose

<!--QuoteBegin--p00n1s+Jan 6 2003, 07:21 AM--></span><table border='0' align='center' width='95%' cellpadding='3' cellspacing='1'><tr><td><b>QUOTE</b> (p00n1s @ Jan 6 2003, 07:21 AM)</td></tr><tr><td id='QUOTE'><!--QuoteEBegin-->(My problem was definitely not as bad on the Mandrake machine, although still unplayable). Perhaps it was just a fault configuration on my part? <!--emo&???--><img src='http://www.unknownworlds.com/forums/html/emoticons/confused.gif' border='0' valign='absmiddle' alt='confused.gif'><!--endemo--><!--QuoteEnd--></td></tr></table><span class='postcolor'><!--QuoteEEnd-->
If the problem exists on both your FreeBSD and your Linux server, then we just need to find out what's common. Can you post your hlds server config, and any other changes you make, no matter how insignificant you deem them to be? What else do these servers run? Output of <b>netstat -i</b> and <b>netstat -m</b>(FreeBSD) while NS is running?

p00n1s

Nothing odd on the linux box, the FreeBSD box is a shell so I don't have too much access to it but I haven't noticed anything out of the ordinary.

By 2%, I mean just that. <!--emo&:p--><img src='http://www.unknownworlds.com/forums/html/emoticons/tounge.gif' border='0' valign='absmiddle' alt='tounge.gif'><!--endemo-->

As far as daemons they both run SSHDs and FTPDs (obviously if ones a shell <!--emo&:p--><img src='http://www.unknownworlds.com/forums/html/emoticons/tounge.gif' border='0' valign='absmiddle' alt='tounge.gif'><!--endemo-->), Apache with various modules, MySQLd. Nothing you wouldn't see in an ordinary installation of an average linux flavor...

I'm considering about switching the tasks around - making my FreeBSD shell CS only and then making the Mandrake box the NS shell. I would switch the Mandrake box to Windows until Linux performance gets better (the Mandrake box is actually mine so I can change it as I need). However, the Mandrake box is on cable modem, a very good cable conncection (it runs a 12 person CS server fine), but nonetheless a cable connection. Do you think I'm better off running a smaller cable server or persuing this bug?

Vadakill

I run a 12 player NS server on a Athlon 1.9Ghz 256Mb RedHat 7.3 Linux box. When the server is full running ns_bast or ns_hera (infamously resource hogs) I run at about 60-70% CPU usage with no lag spikes I notice; when running ns_eclipse I get about 40-50% CPU usage.

I've heard that FreeBSD boxes are Unix based with Linux "emulation", could this be part of the problem? When you tarred your files up and moved them to the other box (mandrake box?) and you still see the same lag spikes?

verbose

<!--QuoteBegin--Vadakill+Jan 6 2003, 10:16 AM--></span><table border='0' align='center' width='95%' cellpadding='3' cellspacing='1'><tr><td><b>QUOTE</b> (Vadakill @ Jan 6 2003, 10:16 AM)</td></tr><tr><td id='QUOTE'><!--QuoteEBegin-->I've heard that FreeBSD boxes are Unix based with Linux "emulation", could this be part of the problem?<!--QuoteEnd--></td></tr></table><span class='postcolor'><!--QuoteEEnd-->
Not likely. It's not emulation, it's an alternate system call vector. In essence, a basic Linux kernel on top of a FreeBSD world. Instead of trying to (poorly) explain it, I'll drop down a link to the official docs:

<a href='http://www.freebsd.org/doc/en_US.ISO8859-1/books/handbook/x29470.html' target='_blank'>FreeBSD Handbook 22.7</a>

Technical mumbo-jumbo aside, I've run a CS server using FreeBSD's Linux binary compatibility, and I've run NS using it. Works great.

verbose

<!--QuoteBegin--p00n1s+Jan 6 2003, 09:10 AM--></span><table border='0' align='center' width='95%' cellpadding='3' cellspacing='1'><tr><td><b>QUOTE</b> (p00n1s @ Jan 6 2003, 09:10 AM)</td></tr><tr><td id='QUOTE'><!--QuoteEBegin-->However, the Mandrake box is on cable modem, a very good cable conncection (it runs a 12 person CS server fine), but nonetheless a cable connection<!--QuoteEnd--></td></tr></table><span class='postcolor'><!--QuoteEEnd-->
I believe NS uses less bandwidth per player than CS, but I don't know how much less.

<!--QuoteBegin--></span><table border='0' align='center' width='95%' cellpadding='3' cellspacing='1'><tr><td><b>QUOTE</b> </td></tr><tr><td id='QUOTE'><!--QuoteEBegin-->Do you think I'm better off running a smaller cable server or persuing this bug?<!--QuoteEnd--></td></tr></table><span class='postcolor'><!--QuoteEEnd-->
Depends on if this is battle you want to fight, or if you'd prefer to just flip things around and have a running server.

One possibility: The -STABLE branch occasionally has some weirdness. Maybe convince the owner of said FreeBSD machine to upgrade to 4.7-RELEASE? Since 4.7 came out, security problems in kadmind and BIND were reported, so if the box uses Kerberos or BIND, upgrading to the RELENG_4_7 cvs tag (alias 4.7-RELEASE-p2) is advised. Just a shot in the dark, and not likely the source of your problem, but you never know...