Apache processes stuck in "Gracefully finishing" state

[Guillermo Pérez]

Binary package hint: apache

I have a big problem with apache 2.2.4-3ubuntu0.1 of latest gutsy.

Apache prefork uses a "internal dummy connection" to tell childs they should die. When apache emits that request to a child (too many idle workers, MaxRequestsPerChild reached...), the children correctly dies (is not longer a proc in the system), but the apache main process still thinks it's "Gracefully finishing".

Server status page shows processes on this state in a number increasing until Maxclients is reached and apache stops answering connections.

It can be reproduced on my system easily issuing a /etc/init.d/apache2 reload. With that I get a lot of connections in the 'G' state.

I'm also seeing problems with /etc/init.d/apache2 restart. It thinks its stoped too soon, and the start finds that the socket is still being used, logs are locked or other problems.

I have seen a maybe related issue on apache's bugzilla:
https://issues.apache.org/bugzilla/show_bug.cgi?id=42829
But I'm not sure.

It's on a xen virtual machine, maybe it's a problem related to this.

[Guillermo Pérez (bisho)]

The reload problem is reported on Ubuntu bug #174805