Bug#497307: apache2.2-common: /etc/logrotate.d/apache2 should reload, not restart

[Stefan Fritsch <sf@sfritsch.de>]

On Friday 05 September 2008, Mark Hedges wrote:
> > It has been decided that this won't be changed in etch
> > (because reload had different problems at the time).
> > However, the problem and the fix is already documented in
> > README.Debian.
>
> Indeed it is.  Which is what I figured out to do.
>
> Am I going to run into these unspecified problems with
> reload?  It's just a PHP site with a few perl CGI's and no
> mod_perl handlers.

Uhh... Possibly.

Some apache child processes may hang on reload. After many reloads, 
you may run into the max number of apache childs or exhaust your 
server's memory.

The problem affects prefork and itk MPMs, but only if you use more 
than one listening socket. If you only listen on port 80, you should 
be safe.

If you are affected, a possible workaround may be to kill all apache 
child processes with "pkill -u www-data apache2" some time after the 
reload. But I haven't tried this.


Cheers,
Stefan