Instance delete should fail in neutron.delete_port() fails

[Phil Day]

Currently neutronv2 deallocate_for_instance() will put the instance into an error state if neutron.list_ports() fails, but continue with the delete if the subsequent neutron.delete_port() fails - leaving ports behind in Neutron in a way that is not obvious to the user.

It would be cleaner if Nova only continued with the delete if a PortNotFound is raised, but otherwise fails, which would result in the Instance remaining in a Error state with an entry in the instance_fault table so that the user can address the issue or at least re-try the delete.

[Gary Kotton (garyk)]

Hi,
From the code it looks like this isssue is already dealt with - please see - https://github.com/openstack/nova/blob/master/nova/network/neutronv2/api.py#L403.
This was resolved in https://github.com/openstack/nova/commit/b12da559b6a40fba4e4431d74372b5e350047525
Thanks
Gary