Steps to reproduce:
Make sure your environment is configured to use the dns-integration extension.
Have an existing port created. Restart the server. It might be helpful first to insert a line into the beginning of the process_update_port method in neutron/neutron/plugins/ml2/extensions/dns_integration.py that prints out the DNS_DRIVER variable. It will be None the first time this method is called, afterwards it will correctly be an instance of your DNS driver object.
After restarting the server, run neutron port-update my-port with any arguments.
While the behavior is the same regardless of the argument, the bug only becomes a problem if the arguments are relevant to the DNS extension, such as dns_name or updating the IP address.
The command will claim to have completed successfully, but the DNS driver is not loaded until the end of the process, after it has been used. Certain functions will check to make sure the DNS driver has been loaded and will exit silently and prematurely because it hasn't been loaded yet. Any subsequent port-update commands will be fine because the driver is now loaded until the server gets restarted again.
DevStack all-in-one built from master
Perceived severity: medium
Confirmed. I am able to reproduce as indicated