apache restart crashes and locks httpd

[ghazlewood]

ASL Version 2.2.6: CentOS 5 (SUPPORTED)
httpd-2.2.3-31.el5.centos.4
mod_security-2.5.12-3.el5.art
CentOS release 5.4 (Final)
PSA 9.2.1 CentOS 5 92090422.13

Not sure if this is a bug in asl, mod_sec or apache itself but it's causing real headaches here so any help or pointers would be appreciated.

A graceful restart of apache (by Plesk) seems to produce this in the /var/log/httpd/error_log

[Tue Jun 08 16:49:35 2010] [notice] Graceful restart requested, doing restart
[Tue Jun 08 16:49:35 2010] [notice] seg fault or similar nasty error detected in the parent process

Apache is then dead and needs killing with a big stick to get it to restart.

Cheers

George

[mikeshinn]

That looks like an apache bug.

[ghazlewood]

Thanks Mike, have you seen it before? Any ideas? This is a pretty standard install with no monkey business!

[BruceLee]

try to disable eaccelerator and test if it's still happening.

[ghazlewood]

Thanks for the tip BruceLee, I've just been reading the other posts on eaccelerator and PHP 5.2.13 and can see that the likely culprit is eaccelerator as the symptoms are identical - I also have the segfaulting problem

[BruceLee]

I have no segfaults anymore, because I have rebuilded the rpm from Scotts srpm.
During my investigation I found a fix for the segfault error I had and send it to Scott.
Unfortunately he didn't build it in. So I had to do it. It's included in the srpm buut commented out.
This should help you:
http://www.atomicorp.com/forums/viewtop ... f=1&t=4024

[ghazlewood]

BruceLee thanks for the tip I'll follow your instructions when I get a chance, I've removed eaccelerator and a graceful restart works ok now. I tried the php-pecl-apc rpm but it looks like that is a little borked as well...