[nycbug-talk] HUGE process size for httpd

[bruno]

On Wed, Jul 13, 2005 at 10:51:57AM -0500, max wrote:
> Hello,
> I noticed this when I ran top today:
> 92011 httpsd     4    0   175M     0K accept   0:25  0.00%  0.00% <httpd>
> 56850 httpsd     4    0   175M     0K accept   0:25  0.00%  0.00% <httpd>
> 
> The 175M is the total process size, RES size is 0k so I am assuming
the whole thing is swapped out. There are 10 procs like this, 5 for
httpd 5 for httpsd ( I separate those for different users/daemons). 
> This is on a machine that has httpd compiled with mm-1.3.1 as per
mod_ssl/apache guide (both httpd and httpsd are). Is that whats doing
it?  

Probably PERL.

> The apache server config with all the virtual hosts and everything
else is certainly nothing special or big, the only modules active is
php/pgsql/dav/mod_ssl and there is almost 0 traffic on that host right
now. 

> Would it make sense that all the modules loaded from apache can make
the process size so big? Or is this an mm thing? 

Its probably mod_perl.
You might want to remove mm, I'm not sure it is recommend anymore, but
just an idea you better check with google or someone who knows this
stuff better.

> The biggest question is, should I be worried about this? Ive never
seen a process so large while not actively running. 

You could check perl scripts for memory leaks..
But it might be normal.
Maybe recycle httpd children faster, to release memory but that is
probably not the correct way to fix this if it is not suppose to take
so much memory.

> Is there anything I can do to really find out what makes it so large?

You could enable apache server-status (detailed) and see what causes
it, what script or whatever. It could be someone uploading a file, and
so on. A PERL and/or programming expert would know more.
I hope this helps at least a bit.

bruno