On Mon, Jun 11, 2012 at 10:13:13PM +0200, Bastian Scholz wrote: > Am 2012-06-11 17:17, schrieb Guido Günther: > >On Mon, Jun 11, 2012 at 02:42:35PM +0200, Bastian Scholz wrote: > >>Sorry, > >> > >>last mail was a little bit too fast. The PATCH for libvirt works > >>only, if "Cache Mode" is manually set to none. If default, it > >>fails with > >> > >>"error: Unsafe migration: Migration may lead to data corruption if > >>disks use cache != none" > > > >I think that's fine and the intended behaviour. Cache mode should be > >none to make this work reliably. > > I think it *is* the right behavior at the moment (See below). > My second mail was only as additional info, that it works only > with cache=none, and not with default caching mode, like in my > wrong attached xml. > > >That said the sheepdog driver doesn't > >handle the cache attribute at all (nbd does for example). Do you > >know if > >migration is always safe with sheepdog? > > I ask this on the sheepdog mailinglist a few days back. > > http://lists.wpkg.org/pipermail/sheepdog/2012-June/004549.html > > Short answer: Yes it should, but with cache enabled it is > not tested at the moment.
Thanks again for the info! So it seems the behviour we implemented upstream is correct! Cheers, -- Guido -- To UNSUBSCRIBE, email to [email protected] with a subject of "unsubscribe". Trouble? Contact [email protected]
