Re: libvirt-guests configurability regression
On Thu, Nov 10, 2022 at 08:57:25AM +0000, Daniel P. Berrangé wrote:
On Wed, Nov 09, 2022 at 09:17:08PM +0100, Olaf Hering wrote:
> Wed, 9 Nov 2022 09:04:12 -0800 Andrea Bolognani <abologna(a)redhat.com>:
> > Olaf, can you please remind me why the files we dropped were
> > problematic but these ones apparently aren't?
>
> These are equally problematic because they are owned by the admin as well.
>
> They are essentially empty, their content is a duplication from doc/ (I hope).
> They should not be part of libvirt.git.
We are not going to remove these files from git or from the RPM, it would
be incredibly user hostile. I didn't mind removing the sysconfig files
because they were legacy configs from initscripts era which are obsolete
with systemd and so should not really be used except for existing
deployments.
Should we reconsider the fate of libvirt-guests's sysconfig file
then?
The other sysconfig files basically only influenced the way each
service is spawned (VIRT*_ARGS, mostly used for --timeout) but in the
case of libvirt-guests the sysconfig file is used to configure the
behavior in ways that don't really fall under the purview of systemd.
Maybe a proper /etc/libvirt/libvirt-guests.conf would make sense, for
consistency with other services? But then, you don't want to be
parsing libvirt's config format from a shell script. Could we rewrite
libvirt-guests in C, possibly improving on the hairier parts of it in
the process? Perhaps even take the opportunity to reconsider the way
some of its settings interact with each other.
I might be getting a bit ahead of myself here :)
--
Andrea Bolognani / Red Hat / Virtualization