Today, I find some of our Windows Server VMs cannot be accessed. After some inspection, I find that there is expected MAC address change on those VMs. There are inconsistencies between "ps aux | grep instance-***" output on such compute nodes where affected VMs reside, with "virsh dumpxml instance-***" on the same compute nodes. The  "virsh dumpxml instance-***" output is consistent with TAP interface MAC address on "ip link" output and the MAC address of port list on OpenStack dashboard/CLI.

Following are one of those outputs. This weirdness is back to normal after I do a live migration to another compute node.

virsh dumpxml instance-000003e5

...

    <interface type='bridge'>
      <mac address='fa:16:3e:ef:4e:3b'/>
      <source bridge='br-int'/>

...

ps aux | grep instance-000003e5

...

netdev=hostnet0,id=net0,mac=fa:16:3e:8e:92:f0

...

What can change the MAC address on  "ps aux | grep instance-***" output? This weirdness happens on Ussuri with Ubuntu 20.04 and OVN. Is this a bug? What kind of bug? Is there any other information should I look for to find the root cause?

Best regards,