as i'm sure you can all tell, the board is playing up more now than it ever used to.

 

for some reason, the server keeps stalling and using up all of its memory

 

i used to think it was busy board/file locking problem, but now i'm not so sure

 

anyone here able to help me out ?

Help like how? I have very very limited Unix experience.

Who was it who has a Unix box as an internet server for there house?

 

Stuart

Are you running ChiliSoft? We had stacks or probs with this.

 

What OS are you running?

 

Been chronicly slow today but just assumed it was Daves updates.

 

What else does the server run?

 

If you need s'ware let me know, I'll see what I can do.

 

Jezz.

What does top say is using all the memory?

 

I'll quiz my boss tomorrow, he's a genius.

 

Is there some better garbage collection prog that you can get? Not too sure but let us know and I'll have a look around tomorrow.

 

Jezz.

Craig you're running Free BSD aren't you with Apache???

 

You got enough space free in the /tmp partition? Sounds to me like the box is running out of swap a.k.a. virtual mem!

 

Log on and run sar, top or vmstat when it's happening. If you haven't got top you can run the uptime command to see what the system load is.

 

Not overly familiar with Free BSD but it is system V mainly from what I do know.

 

Very difficult to speculate without being able to see or get on and investigate! Have a fair bit of knowledge with Solaris Performance and Tuning. Oh and AIX and HP-UX! LOL

 

Drop me some mail if you want more help bud.

 

ok, maybe i should have been a bit more in depth

 

this is top from normal running.

 

when it starts dying, the perl processes are the ones that kill it, or seem to be.

They are every post and forum listing.

 

It might just be that each process is about 3 or 4M, but i dont think so.

it does a serious go slow, and i thought it might be a file lock conflict when too many people post on the same thread at once.

 

when it goes slow.... everything locks up

sometimes when it stalls, it reboots itself. but not very often.

 

there is enough swap, and not much of it is used. that might be the problem, that it is not swapping memory out properly.

 

there is enough disk space, there are enough inodes.

 

it is a P200, 64M with scsi.

the previous server was p75 16M and did not fall over once.. when i replaced with this one, it had 500+ days uptime.

 

i am beginning to think it is either a hardware problem or a simple performance tuning issue

 

last pid: 11264; load averages: 2.96, 2.93, 3.79 up 0+06:53:50 20:54:28

45 processes: 2 running, 43 sleeping

CPU states: 93.7% user, 0.0% nice, 5.2% system, 1.1% interrupt, 0.0% idle

Mem: 16M Active, 9132K Inact, 17M Wired, 2724K Cache, 14M Buf, 16M Free

Swap: 256M Total, 7120K Used, 249M Free, 2% Inuse

 

PID USERNAME PRI NICE SIZE RES STATE TIME WCPU CPU COMMAND

11260 nobody -6 0 4340K 3920K pipdwt 0:02 21.40% 7.76% perl

11256 nobody -6 0 4500K 4076K pipdwt 0:02 15.57% 7.03% perl

11264 nobody 50 0 3956K 3520K RUN 0:02 48.18% 4.59% perl

 

Stabs in the dark here.

 

Do you have a limit of number of instances of perl that can run?

 

The reason I asked about chilisoft before was the we use it and what was happening expired instances of the daemon weren't dying, we had a 200 instance limit which once they were all used meant that no more .asp pages could be read. 1 daemon per .asp page served.

 

So because they weren't dying and relinquishing resources etc a script was written to kill old, expired intances.

 

Maybe sommat like that perhaps?

 

Have you upgraded PERL, which version do you run I'll check for known leaks/issues tomorrow.

 

Jezz.

Stabs in the dark here.

 

Do you have a limit of number of instances of perl that can run?

 

The reason I asked about chilisoft before was the we use it and what was happening expired instances of the daemon weren't dying, we had a 200 instance limit which once they were all used meant that no more .asp pages could be read. 1 daemon per .asp page served.

 

So because they weren't dying and relinquishing resources etc a script was written to kill old, expired intances.

 

Maybe sommat like that perhaps?

 

Have you upgraded PERL, which version do you run I'll check for known leaks/issues tomorrow.

 

Jezz.

That box is getting hammered! 0% idle!

 

What I would reccomend is making sure you have all the latest software packages. If I remember correctly on FreeBSD you just run a make in the src tree somewhere and it automatically pulls down the latest packages (It's been about 2 years since I touched a BSD box).

 

It could also be the fact that you have slow disks, and once it runs out of physical memory it starts swapping and then ends up waiting ages for disk i/o (especially with large processes, we had a similar problem here with one of our av gateways). Do a vmstat and look at the disk stats.

 

There are quite a lot of things it could be - easiest thing to do would be get a faster box :-)

Ha ! Typical lazy programmers response , buy more hardware. Issue is that it is already running on faster hardware before and now it doesn't work right. Solution is to identify the root cause and fix it ...

Come on Unix people, you know who you are. Last time I did proper Unix work was on Ultrix 1.2 !!!!!!! (Gulp)

Well if that top output is from the box while it's performing normally I would agree with JPH!!!

 

You've also got a Perl process hogging nearly half the CPU...asking why a unix box is going slow is like asking how long a piece of string is! LMAO

 

As I suggested you need to either start running sar(system activity report) on a 24x7 basis outputting to a log file so that when the "go slow" starts you can get an idea of where the bottleneck is.

 

Run vmstat also to see what swapping is going on, how many processes are on the CPU and how many are waiting/queing etc.

 

Run iostat -d to see what the disks are doing, oh if only you had a Sun box! I could load Virtual Adrian on and solve all your problems in no time.

 

What else?

 

Your load average Craig for normal running is a tad high mate, 3-4 is not good on a constant basis.

 

Just spoke to my mate Matt, he's looked at the top output and says you need to get Perl inside Apache and the performance will be a thousand times better.

 

You need to use mod_perl apparently here's the link mate...

 

Perl at Apache...mod_perl

 

He(Matt) was saying that everytime a user is posting or whatever, Perl is having to run to interpret the script, compile it etc.

 

If you bring Perl into Apache this is no longer the case, once a script is run and compiled it doesn't keep having to recompile everytime that script is run there after.

 

Only thing he warns is that your scripts have to be clean and tidy. Be careful of global variables etc.

 

Right, that'll be 1 hour at my going rate so I'll do you a discount for cash mate...50 squid!!! ROFLMFAO

 

Tim,

What's Virtual Adrian ? Possibly something to do with Adrian Cockcroft the infamous Sun system tuning expert ? If so he used to live around Hatfield somewhere and I met him a few times because he used to play in a band with a mate of mine. Can't remember whether he's in Cambridge or California right now though. Off topic be relavent, becuause he's also into fast cars ...

Timmy's mate is right. You are invoking the command line Perl program which spawns and hogs all kinds of resources. A load of 3-4 is very high if that is a typical load. I'm used to seeing loads of 0.95 etc.

Get mod_perl installed and once the perl code has been called once, a compiled version is held in memory and you'll probably see an improvement of about 100 times.

 

Andy

come on you guys...

 

the perl process itself is not really the problem, and besides, i am not going to rewrite someone elses BB to work with mod_perl. it is not just a case of install and go...

 

the perl processes are using up loads of resources because they are hanging around and running for too long

 

the reason for that is because it looks like the scsi disks are too slow.. they are performing really badly, worse than the ide disks the previous server was using....

 

anyone want to swap my 2 quantum fireball st's for something else ?

 

consider a shared web host with ~250 or more sites on it (not unusual, often more) the host company doesnt ask everyone to write mod_perl scripts.... and the server doesnt run more than 256M of memory i bet either....

 

at least AndyP is on the same wavelength lol

Craig,

Waht's the disk I/O like, heavy ? Maybe they are badly fragmented. What about a de-frag and some caching s/w ?

What the FUCK are you lot on about LMAO!

 

Craig, what type of SCSI are you using though? I wouldn't say the IDE has any advantage over SCSI for christs sake...otherwise all Unix platforms would have been using them from the year dot instead!

 

You need proof mate, run iostat -d if you have it and have a look at the seek times etc. Not saying that it isn't the disks but hey, proof is much more realistic than speculation bud.

 

Andy, yep the very same Adrian Cockcroft who is the Sun performance guru and with his track record with Sun I'm not surprised he's into fast motors...LOL

 

 

 

Craig,

 

Have you any more details on your hardware architecture? IE AMD/Intel?

 

also you may want to look into

 

RLimitMEM

RLimitCPU

 

in your apache httpd.conf they will kill off processes if they use to many resources,

 

Also this board software isn't the best in the world, it does cause load problems when to many people use it at the same time I have found,