> > So, qemu-ga doesn't understand that there can't be more than one
> > device with same MAC addr. So, I think we are left with the following
> > options:
>
> Actually that's wrong. You *can* have 2 completely different physical
> NICs with the same MAC address.
>
> > (i) Modify qemu-guest-agent to return addresses belonging to same
> > MAC address grouped under one interface only.
>
> You would not want to group based on MAC address - you explicitly just
> want to normalize by stripping the legacy aliases suffixes.
Actually, the order in which the qemu-agent returns the value isn't always:
ethX
ethX:0
ethX:1
ethY
ethY:0
ethY:2
It can be:
ethX
ethY:2
ethX:0
ethY:0
ethX:1
ethY
which, after stripping, will just be left with:
ethX
ethY
ethX
ethY
ethX
ethY
whereas, we would want:
ethX
ethX
ethX
ethY
ethY
ethY
Hence, grouping either by the stripped down interface name or
by the MAC address will be required. So that all IP addresses
related to a single NIC are not distributed in the output.
>
> > OR
> > (ii) Let the reply be as it is now. Strip the ":0", ":1" from the response
> > of guest agent (Is this really necessary?) . We'll have to parse the JSON
> > multiple times and fill the virDomainInterface structs by grouping them
> > according to the MAC addresses.
>
> I think we need to do (ii) regardless to cope with existing deployed
> QEMU agent versions.
>
> We should also recommend to QEMU developers to fix the agent to not expose
> these legacy device alias names.
>
>
> Daniel
> --
> |: http://berrange.com -o- http://www.flickr.com/photos/dberrange/ :|
> |: http://libvirt.org -o- http://virt-manager.org :|
> |: http://autobuild.org -o- http://search.cpan.org/~danberr/ :|
> |: http://entangle-photo.org -o- http://live.gnome.org/gtk-vnc :|
--
Nehal J Wani
UG3, BTech CS+MS(CL)
IIIT-Hyderabad
http://commandlinewani.blogspot.com
> > OR
> > (ii) Let the reply be as it is now. Strip the ":0", ":1" from the response
> > of guest agent (Is this really necessary?) . We'll have to parse the JSON
> > multiple times and fill the virDomainInterface structs by grouping them
> > according to the MAC addresses.
>
> I think we need to do (ii) regardless to cope with existing deployed
> QEMU agent versions.
>
> We should also recommend to QEMU developers to fix the agent to not expose
> these legacy device alias names.
>
>
> Daniel
> --
> |: http://berrange.com -o- http://www.flickr.com/photos/dberrange/ :|
> |: http://libvirt.org -o- http://virt-manager.org :|
> |: http://autobuild.org -o- http://search.cpan.org/~danberr/ :|
> |: http://entangle-photo.org -o- http://live.gnome.org/gtk-vnc :|
--
Nehal J Wani
UG3, BTech CS+MS(CL)
IIIT-Hyderabad
http://commandlinewani.blogspot.com
-- libvir-list mailing list libvir-list@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/libvir-list