Hi everybody,
I have a proposal related to the way init scripts are handled on upgrades;
when you upgrade a package that has an script in /etc/init.d/ the typical
procedure is stop the service on the 'preinst' and start it again on the
'postinst' or restart it on the 'postinst'.
The problem I see with this is that if the service wasn't running because
the admin disabled it on the current runlevel (removing the start symlink in
the '/etc/rcRUNLEVEL.d/' or the current RUNLEVEL in '/etc/runlevel.conf' if
using 'file-rc') the program is started automatically on the postinstall,
leaving a service running when it should be stopped.
My proposal is to have some standard way to control if a service must be
(re)started on upgrades, depending on the current runlevel, in the same
spirit of update-rc.d, maybe something like:
control-init.d SCRIPT ACTION
where SCRIPT is the name of the '/etc/init.d' script and ACTION is one of
{start|stop|restart|reload|force-reload}
My first aproach is that 'control-init.d' presumes that the desired state of
a service is the one configured with 'file-rc' or the rc.d symlinks:
- It only applies the 'start', 'restart' or 'force-reload' commands if the
service is supposed to be active in the current runlevel.
- The 'stop' and 'reload' commands probably should be executed always (if
the service is not running they will not do any harm and if it was
started manually and an upgrade needs to stop it, I think it's OK to
leave it stopped).
Any comments, ideas, opinions? Should I post a 'beta' version of
'control-init.d'?
Kind regards,
Sergio.
Attachment:
pgpndU9sfYDMc.pgp
Description: PGP signature