[horde] Apache running Horde segfaults

[Daniel A. Ramaley]

On Tuesday 04 April 2006 12:27, Ryan Novosielski wrote:
>Where did you get all of your Apache stuff? RHEL RPM's?
>
>> mod_perl is blamed for many segmentation fault problems. I checked my
>> installation and we don't have it installed
>
>Are you certain? It was installed by default on my system. Unless you
>are running from non-RPM's, there is probably a perl.conf in your
>/etc/httpd/conf.d directory that's being sourced (unless you've
> already removed it, but from the sound of it, you said you've never
> had it).

When setting up the server i performed a minimal installation of Red 
Hat. Then i installed the database and Apache with:
    # up2date postgresql postgresql-contrib postgresql-server
    # up2date httpd
I did have to build PHP from source in order to get mcrypt support. I 
found non-Red Hat packages for the mcrypt libraries. Then i used Red 
Hat's SRPMs to build PHP.

The mod_perl package is not installed, nor did i install mod_perl from 
source.

>> Could you please check what your MaxRequestsPerChild is set to?
>> I might try setting MaxRequestsPerChild to something extremely low
>> and see if that fixes the problem.
>
>I just lowered mine to 1024 after talking to you. It has segfaulted
> once an hour later. Typically I get a LOT of faults, though, so that
> might be an improvement. I'll have to watch awhile longer to really
> be able to tell.

I tried drastically lowering mine, to 64. I think that was a mistake; 
the segfaults have been coming in at a rate of several per minute since 
doing that. However, the server is still responsive. Go figure. I do 
wonder how bad the segmentation faults are as long as the server is 
still performing properly, though. It does seem that some types of 
regular errors in the logs are normal for a Horde server, such as EOF 
errors in PostgreSQL's log and missing file errors in Apache's log.

>Incidentally, watch out for restarts -- the cache directory of those
>accelerators will really screw things up if there are orphaned files
> in there.

I hadn't thought of that; next time i restart Apache i'll clean out the 
cache directory to see if it improves anything. If it does then i'll 
have to wonder why the cache would have become corrupted in the first 
place.

------------------------------------------------------------------------
Dan Ramaley                            Dial Center 118, Drake University
Network Programmer/Analyst             2407 Carpenter Ave
+1 515 271-4540                        Des Moines IA 50311 USA