On Mon, Dec 07, 2015 at 10:27:15AM +0000, Daniel P. Berrange wrote:
> On Mon, Dec 07, 2015 at 09:35:42AM +0000, Daniel P. Berrange wrote:
> > On Mon, Dec 07, 2015 at 11:06:52AM +0800, Daniel Veillard wrote:
> > >
> > > So as planned I tagged rc2 in git and pushed signed tarball and rpms
> > > to the usual place:
> > >
> > >
ftp://libvirt.org/libvirt/
> > >
> > > but in my testing I'm hitting a serious issue. After upgrading to the
new
> > > set of rpms and restarting the service and virt-manager, while
virt-manager
> > > can connect and list the guests it fails to start a guest with the
following
> > > error message:
> > >
> > > Error starting domain:
> > > Failed to connect socket to '/var/run/libvirt/virtlogd-sock':
> > > No such file or directory
> > >
> > > That to me is a release blocker we need to have that in place upgrade
working,
> > > we can't just pull the socket even if we may have another better
mechanism.
> > > So unless there is a good reason for this due to my setup and not the
update
> > > I would delay 1.3.0 and put an rc3 instead on Wednesday.
> >
> > We register virtlogd with systemd for socket activation so that it should
> > automatically start when first needed.
>
> So it seems we merely get it enabled, but not started, unless you reboot.
> So we'll need to add some more logic to %post to ensure it is started
> if libvirtd is seen as running already
Right, I didn't rebooted, merely upgraded the rpms from rc2 and
restarted the libvirtd service, i.e. my usual testing routine
NB, there hasn't been any need to restart the libvirtd service
manually for a very long time now. The %post script always
restarts it.
Here is a patch that should fix virtlogd so that it is started
upon upgrades from libvirt < 1.3.0:
To test you'll obviously need to downgrade to some 1.2.x before then
installing the new 1.3.0 build
Regards,
Daniel
--
|: