We have several multi-homed VMs, running reverse proxies and VPN gateway.
Whenever these are live migrated, we have a problem, where the reverse proxies lose access to the domain (ADSI calls fail, even though SSPI succeeds).
In these cases, the LAN adapter has become unidentified.
In order to have it re-classified properly as Domain, we either have to
a) reboot the image
b) disable the external NIC, then disable the LAN NIC, re-enable the LAN NIC, then re-enable the external NIC. If the external NIC is enabled, it prevents successful discovery of the LAN NIC.
I've seen other posts for years about similar issues with NLA and multi-homed computers. MS, you really need to allow us to force adapter usage.
This to me screams that there's some kind of broadcast discovery going out all adapter, and choosing the wrong source IP. There was a long-standing bug relating to broadcast on multi-homed windows hosts, where a broadcast UDP packet would go out all adapters but use the source address of only 1 of the adapters. This prevents replies from all the networks joined to the other adapters.
Maybe it's somthing else, but it means live migration is not possible, since it breaks our reverse proxies and requires manual intervention. Quick Migration does not do this, but failover uses Live migration.
Cheers
Adrien de Croy