(9:29:36 AM) danpb: mriedem: this is still nova's fault
(9:29:57 AM) danpb: mriedem: the API contract for virDomainDetachDevice says you must pass the full XML for the device shown in the current guest XML
(9:30:13 AM) danpb: mriedem: nova doesn't do this - it re-creates the minimal <interface> XML from scratch
(9:30:31 AM) danpb: mriedem: as a result it is missing the PCI device ID libvirt would normally use to uniquely identify the device
(9:30:40 AM) danpb: mriedem: so libvirt has no option but to try and find it using mac address
(9:30:54 AM) mriedem: ok, do we have any existing code to build that interface xml from the vif?
(9:31:01 AM) mriedem: seems like a simple fix
(9:31:24 AM) mriedem: it's a latent bug, so not an rc issue, just wanted your ack on it
(9:31:33 AM) danpb: yeah, its been broken forever
(9:29:36 AM) danpb: mriedem: this is still nova's fault Device says you must pass the full XML for the device shown in the current guest XML
(9:29:57 AM) danpb: mriedem: the API contract for virDomainDetach
(9:30:13 AM) danpb: mriedem: nova doesn't do this - it re-creates the minimal <interface> XML from scratch
(9:30:31 AM) danpb: mriedem: as a result it is missing the PCI device ID libvirt would normally use to uniquely identify the device
(9:30:40 AM) danpb: mriedem: so libvirt has no option but to try and find it using mac address
(9:30:54 AM) mriedem: ok, do we have any existing code to build that interface xml from the vif?
(9:31:01 AM) mriedem: seems like a simple fix
(9:31:24 AM) mriedem: it's a latent bug, so not an rc issue, just wanted your ack on it
(9:31:33 AM) danpb: yeah, its been broken forever