Hi Michal!
Sorry, guys.... I was on vacation during this time.
Thanks to pointing all suggestions.
Basically, we are restarting libvirt to install properly the hook that
our package installs.
But it is very annoying to users.
Generally, they are using Virtual Machines when they receive a package
update and after the installation,
They lost the current connection to libvirt because the service was restarted.
So, think in 10K users using some Virtual Machines and losing
connection (virt-viewer, virt-manager, etc) while they are updating
the system. Unfortunately, our hooks setup some network
configurations. So, the hooks needs to be loaded. We can't wait for
the users shutdown/reboot his computer. :-(
But really, I don't want to introduce big changes (including
documentation) to fix only a load problem.
--
Julio Cesar Faracco
2016-09-19 5:09 GMT-03:00 Michal Privoznik <mprivozn(a)redhat.com>:
On 13.09.2016 02:25, Julio Faracco wrote:
> Hi guys,
>
> As I mentioned, we are working in a project that introduces a hook for QEMU.
> And when someone installs the RPM/DEB package with this hook,
> we need to restart the libvirtd service everytime. It is really
> annoying for users.
Well, RPM packages have %post macro where package maintainer can put
some scripts. Well, commands to run. For instance:
/bin/systemctl try-restart libvirtd.service >/dev/null 2>&1 || :
can be there. I bet DEB packages have similar option.
>
> In my opinion, we can include an #if and #else to compile with inotify support.
> But, I don't see any good reason to add.
Well, one of the reasons might be that other systems than Linux where
libvirt is built would fail after this patch. inotify is Linux specific
and for instance *BSD kernels don't have it. So we need to check for it
in configure phase.
Michal