(re)starting of a daemon after package update

Joe Orton jorton at redhat.com
Fri Jun 22 13:01:14 UTC 2012


On Wed, Jun 20, 2012 at 05:59:09PM +0200, Reindl Harald wrote:
> i saw a dist-upgrade on the fist-testmachine restart httpd
> which failed for some reason and AFTER some time you could
> start httpd again

Leaving version X of /usr/sbin/httpd running after version Y of 
/usr/sbin/httpd is installed, plus associated loadable modules, is 
simply not a supportable system state.  The software is not designed to 
cope with this situation.

Notably, a SIGHUP of the httpd binary will attempt to use loadable 
modules from the new httpd with the old httpd binary.  That is liable to 
kill or segfault the running daemon at best.  Prior to the 
%posttrans-restart, we had numerous bug reports from people who updated, 
forget to restart the daemon, and then found their web server got killed 
during the night at the weekend when logrotate kicked in.  That's not a 
good default.

It's not unreasonable to request that the %posttrans-restart be made 
configurable, though I'm not sure exactly the best way.  It would be 
nice if this could be co-ordinated in some fashion, so an update of any 
httpd module package could (optionally) trigger an httpd restart in 
%posttrans.

Regards, Joe


More information about the devel mailing list