About the patch itself, I would like to test it this week.  Don't let
that stop you from pushing it if you get the required approvals
however.  I scanned the patch and it looked fine.  More below ...

On Tue, May 12, 2026 at 05:47:48PM +0100, Daniel P. Berrangé via Devel wrote:
> On Tue, May 12, 2026 at 06:23:30PM +0200, Martin Kletzander via Devel wrote:
> > diff --git a/tests/vmx2xmldata/esx-in-the-wild-10.xml 
> > b/tests/vmx2xmldata/esx-in-the-wild-10.xml
> > index 1b1fdf06623f..3a8bb20140a1 100644
> > --- a/tests/vmx2xmldata/esx-in-the-wild-10.xml
> > +++ b/tests/vmx2xmldata/esx-in-the-wild-10.xml
> > @@ -1,6 +1,6 @@
> >  <domain type='vmware'>
> >    <name>w2019biosvmware</name>
> > -  <uuid>421a6177-5aa9-abb7-5924-fc376c18a1b4</uuid>
> > +  <uuid>501af9f2-6d29-1c76-19a9-b208ede5f374</uuid>
> 
> Hmmm, seeing this indicates that this change is effectively a semantic
> incompatibility across the libvirt upgrade path.
> 
> Any management application which is using the libvirt reported UUID
> for tracking / correlating will effectively see all their VMs
> disappear and an entirely new set appear with different UUIDs.

Are there management applications which use libvirt with VMware?  Red
Hat has a couple of programs using libvirt to access VMware, but we
only use it for converting VMs off VMware, and we don't care if the
UUIDs change.

> This is the kind of upgrade incompatibility that libvirt promises
> not to impose on applications.
> 
> What options do we have for mitigation ?
> 
> Given that we don't have /etc/libvirt config files for stateless
> drivers, I feel like we need to at minimum have a URI parameter
> to allow the toggle of old/new UUID representations.
> 
> A strict view would be for the old behaviour to be the default,
> but on the other hand the old behaviour violates our API semantics
> by not actually being unique. So that could lean me towards accepting
> the compat break, in order to get better API compliance, as long as
> we have the URI param to opt-in to the original behaviour

Definitely opt-in.  Having to add a parameter to the URI forever
doesn't sound great.

Rich.

-- 
Richard Jones, Virtualization Group, Red Hat http://people.redhat.com/~rjones
Read my programming and virtualization blog: http://rwmj.wordpress.com
virt-top is 'top' for virtual machines.  Tiny program with many
powerful monitoring features, net stats, disk stats, logging, etc.
http://people.redhat.com/~rjones/virt-top

Reply via email to