Re: Support cloning of VMs - part 2

On Fri, Jul 11, 2025 at 01:49:28PM +0100, Mark Cave-Ayland wrote:

>>>> This probably takes us back to needing to have another element at the >>>> top level in the XML. >>>> >>>> <uuid>....</uuid> >>>> <hwuuid>...</hwuuid> >>>> >>>> where <uuid> is mandatory as today, and <hwuuid> can optionally be used >>>> to override what is exposed to the guest, defaulting to <uuid> if >>>> omitted. >>> >>> Thanks for the update, I'll take a look at implementing something along >>> these lines. Would it be slightly easier to add the guest-visible UUID as an >>> attribute to the existing uuid element e.g. >>> >>> <uuid guest-uuid="....">....</uuid> >>> >>> since then it should already be accessible at the places where we already >>> have an existing reference to the uuid element? >> >> We already have the 'genid' UUID as a separate element, so I figure >> we just carry on that practice with 'hwuuid'. > > Thanks for the hints! I've been looking at the various drivers within > libvirt, and from what I can see not all virtualisation platforms will allow > a separate UUID to be provided in this way. > > Would the best way forward be to add a new capability to represent this > ability, and then follow up with a QEMU implementation? I think its probably overkill to add capability. Just define the XML and wire up QEMU.