Re: [cobalt-users] CPU is overloaded - Help

["Brian Busche" <brian@xxxxxxxxxxxxx>]

Dan,

I get that error three or four times a day.  The culprit:  a CGI called
Hyperseek. It launches four or five instances of the CGI files it uses to
search it's database and these processes quickly eat up all of the available
CPU time, and then some.  Ten seconds later they are gone from memory and yo
can't catch them in the act unless you run 'ps aux' at just the right time.

You might not be running Hyperseek, but you may have a similar perl script
out there (funny that it runs at 9:15 every time, perhaps a cron job?).
How serious is this problem?  I have found that it is relatively harmless
unless you have some busy sites and other processes are fighting for CPU
attention at the same time.  Get three or four users hitting a big perl
script and you could lock up for an extended period like I do (five to ten
minutes without telnet, ftp, http access...ouch!)  If you suspect a
particular perl script like hyperseek, check the access log to see if it is
getting hit around 9:15.

Solution?  Adding memory does not cure the problem but may help in cases
were you are maxing out the mem also.  I have been considering installing
'mod_perl' to speed up CGI execution and get it out of the processor faster.
Problem is, I can't find anyone who has tried this on a Cobalt RAQ.  I do
not want to be the guinea pig who crashes his system.  It seems more complex
than simply installing a Perl module, it askes for the source code location
of Apache....Yikes!  Cobalt will not support you if you crash your server
doing this so proceed with extreme caution.  More info on mod_perl can be
found at http://perl.apache.org



Brian Busche
brian@xxxxxxxxxxxxx


----- Original Message -----
From: Dan <dan@xxxxxxxxxxxxx>
To: <cobalt-users@xxxxxxxxxxxxxxx>
Sent: Monday, February 28, 2000 8:28 AM
Subject: RE: [cobalt-users] CPU is overloaded - Help


>
> >
> > ps aux |sort -nr -k3 |head -10
> >
> > that will show you the top ten process sorted by CPU usage. :)
> >
> Trying the above after receiving the "needs more memory" email from the
> server and I see:
>
> root     26769  0.7  4.5  4960  1412  ?  S   09:15   0:02
> /usr/sbin/ahttpd -d /
> httpd    26899  0.3  3.3  5396  1036  ?  S   09:20   0:00 /usr/sbin/httpd
> httpd    26902  0.2  3.3  5396  1036  ?  S   09:20   0:00 /usr/sbin/httpd
> httpd    26900  0.2  3.3  5396  1036  ?  S   09:20   0:00 /usr/sbin/httpd
> httpd    26898  0.2  3.3  5396  1036  ?  S   09:20   0:00 /usr/sbin/httpd
> httpd    26897  0.2  3.3  5396  1036  ?  S   09:20   0:00 /usr/sbin/httpd
> httpd    26892  0.2  3.3  5396  1036  ?  S   09:20   0:00 /usr/sbin/httpd
> httpd    26891  0.2  3.3  5396  1036  ?  S   09:20   0:00 /usr/sbin/httpd
> admin    26836  0.2  3.3  2416  1036  p0 S   09:18   0:00 -bash
> root     26835  0.1  2.4  2628   756  ?  S N 09:18   0:00 in.telnetd
>
> top was showing:
>
>   9:20am  up 38 days,  4:04,  1 user,  load average: 0.31, 0.38, 0.49
> 56 processes: 55 sleeping, 1 running, 0 zombie, 0 stopped
> CPU states:  0.3% user,  2.7% system,  0.0% nice, 96.9% idle
> Mem:   31064K av,  24512K used,   6552K free,  22268K shrd,    420K buff
> Swap: 130748K av,  12208K used, 118540K free                  5492K cached
>
> Anyone have any idea what is going on that would cause the server to say
it
> needs more memory? This rarely happens, but when it does, it always seems
to
> be about the same time of the morning. 9 - 9:30AM ET. Got two messages,
one
> at 9:15 and another sent at 9:16.
> --
> Dan Kriwitsky
> 
>
>
> _______________________________________________
> cobalt-users mailing list
> cobalt-users@xxxxxxxxxxxxxxx
> http://list.cobalt.com/mailman/listinfo/cobalt-users
>