On Sat, May 14, 2016 at 09:11:17PM +0200, Moritz Mühlenhoff wrote: > On Fri, May 13, 2016 at 02:10:48PM +0200, Guido Günther wrote: > > > No, I recommend to EOL src:qemu/qemu-kvm in wheezy (the bits relevant to > > > src:xen are > > > somewhat isolated and can be backported from the Xen Security > > > announcements) > > > Backporting jessie's qemu will end up in a similar situation as the > > > experiments > > > with libav. > > > > I don't think so. We already have the version of Jessie in Wheezy > > backports and I've looked into the dependency list. See > > > > https://lists.debian.org/debian-lts/2016/04/msg00002.html > > > > so this should be doable… > > But what would be the point of that? Upgrading a virtualisation server to > a new qemu effectively requires the same amount of testing/downtime of the > operation side as upgrading that system to jessie.
The server might not be a dedicated virtualization server but be running services that aren't easily upgraded as well. > > Except that you would now run your system on a mostly untested combination of > vintage > 2.6.32 Linux, old libvirt and modern qemu compared to the jessie stack which > is > extensively tested through a full freeze. And you'd still need to upgrade to > jessie later. The above proposal would update libvirt as well (and I've tested this setup quiet a bit - not comparable to a freeze though). That said I wrote earlier that I do agree that it would be better to drop QEMU/KVM if nobody is in favour up updating and supporting it. Cheers, -- Guido
