So far I have 2 sites on my dedicated server (both are fairly high traffic). And the server load average is around 1.50 constantly...

Just wondering if this is normal for high traffic sites.

I am checking 'top' and have this line:

%CPU: 96.2
%MEM: 1.5
TIME: 3658m
COMMAND: httpd

Does httpd actually run for that long and take up 96.2% CPU?

well, while that isn't an unacceptable load average, it would be easier to determine some things by giving your hardware stats. For example, it a single or dual processor, how much ram, etc.

Also take into account how much idle CPU is available... if that is really 96.2% CPU usage giving you only 3.8% idle, you may have some problems.

AMD Duron
1GHz
512 MB RAM
40 GB Hard Drive

I've got 2 other dedicated servers and I've never seen any process run for over 3600 minutes... would you recommend i restart httpd?

Yeah, it isn't too often you would see a standard process like that run for that long. A quick restart of the httpd may fix it if it is a runaway process.

I'd go with rebooting it too. I've had problems in the past where after a few days of massive traffic Apache would be using 1-3 loads. Rebooting the httpd made it drop to 0.00-0.20 again.

ok then... i'll restart httpd... guess it's not quite normal then :D

Thanks

I'd look through your logs and see what happened during the past 3658 minutes (thats a lot of logs though) that may have caused it, chances are you are more likely to find it as an older entry than a newer one, if anything at all. Though you may find nothing.

Chalk it up as a glitch and hope it doesnt happen again. If it does, you will need to work on reproducing it somehow.

Heh, it didn't do anything... I ran: /etc/rc.d/init.d/httpd restart and it's still up there...

k, I ran a stop and start and that solved the problem :p

httpd restart should do it also :P

Definately check you log files. CPU will always spike when a process starts just like when you fire upa program on your desktop. Its when it is consistantely up there you want to check it out.

aNc

What I've found in the past is that a PHP script has usually hung, causing Apache (httpd) to consume excessive amounts of CPU / memory. With Solaris, you can find the working directory of that httpd PID using pwdx, which helps narrow down where the process is operating or do a truss on the PID to see what exactly the PID is doing. A simply apachectl restart (or graceful, which is a bit nicer) will kill off those nasty PID's. Try to find out where they're coming from though, as chances are they'll be back!

Try running the "sar" command (If you use Redhat, it should be installed by default).

That'll give you a good idea of what your usage is like... it also will store (default on redhat) usage for each day in /var/log/sa/ (run sar dayfilename to generate the stats)

Just ran into this again:

nobody 28409 96.9 1.5 11712 7688 ? R Nov23 629:30 /usr/local/apache/bin/httpd -DSSL

96.9% cpu

Before ...

root@zeus [~]# uptime
4:15am up 3 days, 12:58, 1 user, load average: 2.18, 2.19, 1.75

root@zeus [~]# httpd restart
/usr/sbin/httpd restart: httpd restarted

Now ...

root@zeus [~]# uptime
4:22am up 3 days, 13:05, 1 user, load average: 0.00, 0.58, 1.13

Not sure why it's doing that but that's the only reason I get a load issue (from the Apache thing). I just find it hard to believe it's a PHP script causing this.

What you should do is find the process (PID) number in top and zap the hung process. Usually:

kill *** (stars being the process number)

If it's still there then you need to send the kill command in first via:

kill -9 *** (being the PID)

Then restart Apache if you want though there shouldn't be a need.

I kill'd -9'd it before about 7 days agoi, went to watch TV for a bit then noticed that apache had totally died. I figured I probably shouldn't of killed it and just figured restarting it was a safer bet.

I'll remember that for next time thanks. ;)

Restarting Apache won't always clear hung processes as they cling to port 80. Just as restarting MySQL won't kill off MySQL processes some times. A direct kill will do in the bad process but if it hasn't let apache do anything else, you leave it dead in the water, hence the need for a restart.

...have you checked the error log?

Apache is generally pretty good at reporting errors.

NexDog, oh okay.

clockwore, no. I don't know how ... new to using Linux just got my own servers about 1 or 2 weeks ago and don't know too much yet. Still learning a great deal about it.

Try to just run:

tail -f /usr/local/apache/logs/error_log

And watch over it, or, alternatively, open it up in an editor (or a simple 'more') and look at the timestamps for when your problem was occuring.

I'm just used to Plesk's structure but the error_log in Linux is pretty much in the same place:

/var/log/httpd

and do:

less error_log

to scroll to the bottom, hit shift and g.

Of course your domain will have it's own error_log too, visible in shell and via FTP too probably.

Originally posted by NexDog
I'm just used to Plesk's structure but the error_log in Linux is pretty much in the same place:

/var/log/httpd

and do:

less error_log

to scroll to the bottom, hit shift and g.


Of course your domain will have it's own error_log too, visible in shell and via FTP too probably.

Well, he's got it installed in /usr/local (see his previous posts), could very likely be a CPanel box.