AvailabilityZoneFilter returns a wrong result because of the availablity_zone in the instances table has not be updated

[zengrui]

When you do a live-migration for an instance from one zone to another successfully or you change the zone of
the compute node which the instance launched on, but the availablity_zone of this instance in the instances table won't be updated.
Then when you do a cold_migrate or resize the instance, the AvailabilityZoneFilter will get a wrong zone of the instance, and the filter result must be wrong, and maybe you won't get the result you want for the cold_migrate or resize.

[Ritesh Paiboina (rsritesh)]

Could you please provide the steps you have used to reproduce this, it helps to understand better and also helps to reproduce the issue.

[zengrui (zengrui)]

If you only have two compute nodes, compute1 in zone1, and compute2 in zone2.
1. Create an instance in zone1;
2. Do a live_migration for the instance from compute1 to compute2, you can specify the destination node here;
3. The availablity_zone of the instance in the table instances is also zone1;
4. Then you can do a cold_migrate for the instance, you can migrate the instance from zone2 to zone1 successfully because of the instance's availablity_zone is wrong in the database;
5. And if you change the zone1's name or delete compute1 from zone1, then you may get an unexcepted result when you do a cold_migrate for the instance or resize the instance.

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/491226

[OpenStack Infra (hudson-openstack)]

Change abandoned by Matt Riedemann (<email address hidden>) on branch: master
Review: https://review.opendev.org/491226
Reason: As noted by Kevin, the AvailabilityZoneFilter uses the RequestSpec not the instance so this is a bit off anyway, but like I said I think this is already fixed.