Hey guys, you guys have been very helpful to me in the past, so I thought I'd bring it back here and see if anyone could help me or not.

Lately, I've been having trouble with a linux box I use a web-development test environment. When I start it up(or reboot it), everything works fine, and normal speed, performs properly, etc etc...

After some time (I typically check on it on a daily basis), it starts to slow down incredibly...

I'm not really using a gui on it or anything, the only time I use a gui on it at ALL is when I run vnc, which starts gnome, and then (I'm guessing) stops gnome when I'm done running vnc. Typically however, I start a vncserver session and just leave it on, I never terminate it or anything, and I dont use / start more then one server session at a time.

In any case, a day or so later, it just bogs down completely. The smb shares on it don't connect quickly (if they connect at all), even when I ssh / putty in, it takes a long time for me to be able to log in or do anything. Evne when i run 'top', it still takes a while for it to pull the process list up.

My question is basically, how do I figure out what's wrong with this thing, what should I do to go figuring out what is causing the problem?

when you run top, what does it say is eating up all the cpu? That might narrow the problem down a bit. That might help us answer more effectivly.

...drkstr

hey drkstr, thanks for the reply. Under the cpu column, the processes ALL have 0.0 for the cpu%, the process wwith the highest memory usage is httpd (apache) with a grand total of 1.5.

It gets weirder... the TOTAL cpu usage never runs higher than .14% us.

Mem: 507,372k total, 452,032k used, 55340k, free, 124000k, buffers
Swap: 2,096,472 total, 52k used, 2,096,420k free, 118,792 cached


Needless to say, I'm very confused...

EDIT:: In fact, here's a screenshot of said top screen, a coupla minutes later, so the numbers are slightly different me thinks.

System Processes

Few things come to mine.
1. Maybe a heat issue. Check to see if fans are running. See how warm the air is exiting the case. May not be enough air flow through case. Also check power supply fan as well.

2. Another heat up issue can be a underpowered power supply or one going bad.

3. Maybe the processor is a speedstep type of one and some cpu speed control is slowing the proeceesor down during less usage time.

Brian1

Definitely not a heat issue, nothing is even remotely warm, the cpu heatsink is barely warm to the touch. The power supply is brand new, working properly, seems to be moving air quite briskly.

The computer is a server, and thus, not relegated to the speedstep changes like laptops, but it DOEST have a speed setting, which I've turned to maximum (not overclocked, just not slow).

Thanks for all your suggestions, got any more? I'm still stumped!!

You have Gnome, so it may be famd. I had big trouble with fam when I had Gnome, it just went berserk after some NFS transfers. No matter how I (re)configured it the trouble persisted. Your screenshot shows CPU usage 100% but fam seems not to be listed. Anyway, try stopping the fam daemon. Hope this helps.

*blinks* what is the fam daemon and how do I stop it?
Also, where does it show that cpu usage is at 100%?
Thanks, apparently I'm reading something wrong or something's going on.

EDIT:: Oh wow, I just saw that in the graphic....but nothing within the cpu% us column shows anything using it....wow...

So how exactly do I stop this fam daemon?

also, the computer updates (daily) the local database, I've noticed that always seems to take a little while. the whole updatedb thing...could that have anything to do with it?

It depends on distro. It would be '/etc/init.d/famd stop' for Gentoo.

Edit: I'm sure it has no direct relation to your database update. FAM is file alteration monitor (if i recall correctly) and is responsible for refreshing file listings in GUI.

famd is not just going berserk in Gnome, it happens in KDE as well. BUT you definitly see it show up in top.

It might also be that some process can slow down a lot of activity when it is a critical process, waiting for something.

For example, if you do a lot of disk swapping, your computer is stuck, but still you see 0.1% processor load.

Something similar could happen if you are connected to this box over the network, and you have a network problem. For every byte which has to be sent or received, the network stack waits a very long time. Since the process goes to sleep meanwhile you won't see any processor load either. Network problems can even be hardware problems.

jlinkels

Hi,
Over time top is nice but won't really tell ya what is pounding the memory. 512 Meg really is not that much in my book for memory, that is what I am using for a web server and email server combined and my mem usage is around 500 meg. Not sure if you built the kernel from scratch or are using the kernel from the distro that would be the first thing I would suggest. Only run in the kernel what you need. I would start there. I believe Linux has some issue's with releasing memory I am working on your kind of issue also. Good Luck,
Rick

Given that the cpu shows 100% in userspace, but no process shows up as using cpu and mem usage is fine (low swap usage), I'd guess it could be a rootkit that hides it's process info.
Try http://www.rootkit.nl/projects/rootkit_hunter.html & http://www.chkrootkit.org/ just in case.
A sleeping network issue should not hammer the cpu, but look at eg netstat to be sure.

Start terminating processes until the system seems more responsive.

The symptoms you describe are related to some form of a memory leak.

Systems that run great at reboot or power-on and then slowdown over a period of time are usually experiencing some form of a memory issue.

Based on what you've said I would say the memory is physically fine, however, i'd agree with others that some process is causing the problem.

I do not know of any other more in-depth memory monitors out there, but my start would be to find one and compile it and then reboot and fire the program up, and watch to see what starts to suck up memory and never release it.

With Windows Server, XP and W2K a guy can use the perf mon to create a counter and track usuages such as *above* over a period of time. Very helpful in situations like yours.

Anyone out there who could point him (and me because im curious) in the direction of a good memory/cpu/process monitor... maybe even one that would allow logging over time for review later?

Scott

hmmm...that makes sense, I would attribute it to the same thing within windows if it were doing that, but I've pared down all the processes and such being used on this box for that reason.

I'm sure I could optimize more, and yes, that utility for analyzing / detecting memory use would be fantastic.

Anyone have any suggestions?