Le jeudi 13 janvier 2005 à 10:05 +0000, Joe Orton a écrit :
On Wed, Jan 12, 2005 at 04:33:13PM +0100, Féliciano Matias wrote:
> Is this a bug ?
> $ rpm -q --scripts bind
> postuninstall scriptlet (using /bin/sh):
> if [ "$1" -ge 1 ]; then
> /etc/rc.d/init.d/named condrestart >/dev/null 2>&1 || :
> fi
I'd say it depends on the daemon, for some it could be necessary. But
for any network-facing daemon I would say that it's surprising and
undesirable behaviour if a package update breaks active client
connections etc. Making configuration changes take effect from a %post
is unfriendly; especially if the config is not marked noreplace.
When I update my system, the propose is to use the new version (security
fix, ...). Perhaps it's better to schedule an update at midnight than
when there is a rush of connection. It's up to the administrator to
decide when and how to update his system.
If daemons are not restarted when I update a system then when this
append ?
Do you ask me to always reboot my computer after "yum update" ? Too bad.
Suppose I update httpd and httpd is not restarted. One week later the
server is restarted (power failure), the new httpd does not start and
"bad luck" I am in holidays...
What do you think about this scenario ?
btw, "service httpd reload" (or condreload) is not enough alter an
update ?