Hi,

I'd like to report a bug I found when doing cinder-migrate test.

I really would appreciate it if somebody inform me if it is solved at the latest tag or not solved yet.

I checked the latest tag 24.2.0 (https://github.com/openstack/cinder/releases/tag/24.2.0) but failed to find a bug patch looking relevant to my issue.

I tested the volume migration of nvme volume (nvme-tcp volume with LVM) with "openstack volume migrate --host <servername@diskname#typename> <UUID>" command.

The first migration worked well but the second migration failed always with "Skipping remove_export. No NVMe subsystem for volume: <uuid>" error.

Below is what I found

If the original volume UUID is '23c5155e' that exists in server-0002:

= The first migration: server-0002 => server-0001 =

- Create a new LVM volume with a different name UUID 'ed7a9c7b' => new nvme subsystem name is 'ed7a9c7b' (server-0001)

- Set the UUID of new volume 'A' and set new attribute "os-vol-mig-status-attr:name_id | ed7a9c7b"

- Remove/unmap the original volume on server-0002: unmap nvme subsystem name is '23c5155e'

= The second migration server-0002 => server-0001 =

- Create a new LVM volume with a different name UUID 'af60b1c7' on server-0001

- Try to unmap the nvme subsystem of 'af60b1c7' of server-0001

==> it failed because there is no subsystem 'af60b1c7'. It should've unmap the subsystem "ed7a9c7b".

I guess there is a problem at unmapping a subsystem of already migrated volume.

It happens 100% whenever I migrated a volume twice.