On (02/07/13 08:28), Stephen Gallagher wrote:
On 07/02/2013 03:54 AM, Lukas Slebodnik wrote:
> On (01/07/13 12:48), Stephen Gallagher wrote:
>> Patch 0001: Move pre and post scripts to sssd-common When we
>> switched to making 'sssd' a meta-package, we forgot to move the
>> upgrade scripts into the sssd-common package. This means that
>> updates won't work right without the 'sssd' package and that
>> removing the meta-package (i.e. to trim down requirements) will
>> remove SSSD from the services lists.
> Each scriptlet section contain command systemctl with argument
> daemon-reload or disable (or chkconfig (sysv)). Those actions are
> related to service file and sssd.service is in package
> sssd-common. post, preun and postun should not be run with "meta"
> package sssd and definitely must be run with sssd-common.
>
> But I have a question. Model situation: package sssd is installed
> and configured only with ipa. sssd is connected to ipa server.
> Someone removes sssd-ad subpackage, in this case, it does not make
> sense to restart sssd. Someone remove sssd-ipa. Should sssd be
> restarted?
>
> I am not sure if this model situation make sense, but users can do
> any mistake.
>
First off: just so we're clear, with these patches, SSSD is only
restarted when the sssd-common package is upgraded (and it's not
started automatically on initial install either). It is not started if
any other subpackage is added or removed. This is intentional, because
any change of that nature should obviously be associated with a change
to sssd.conf, which will necessitate a manual restart of SSSD in any case.
Additionally, because of the way that we list the dependencies, it's
not possible for sssd-[provider] to be updated without sssd-common
also being updated to match.
So I'd say that this is a non-issue. If someone is relying on a
provider and removes it, they've made a conscious decision to screw
something up. I don't think it makes any sense to code around that.
OK.
I should not be so paranoid.
ACK to both patches.
LS