iSCSI rescan at disconnect_volume, can orphan volumes

[Walt Boring]

When running nova live migration, doing an SCSI rescan inside of disconnect_volume can bring back removed volumes.

During nova live migration,
 for every volume attached to a VM, nova will first call os-brick to disconnect_volume.
Then nova will call Cinder to terminate_connection.

The problem is, when you do a rescan prior to having the Cinder driver terminate the connection, a volume that was removed locally, will show back up, and then be left behind after the migration is complete.

We should not be calling rescan during disconnect_volume time.

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/203265
Committed: https://git.openstack.org/cgit/openstack/os-brick/commit/?id=c0839fa55067e96bf577d16b7b9d825ae7b52aba
Submitter: Jenkins
Branch: master

commit c0839fa55067e96bf577d16b7b9d825ae7b52aba
Author: Walter A. Boring IV <email address hidden>
Date: Fri Jul 17 15:42:43 2015 -0700

    Remove the iSCSI rescan during disconnect

    Doing an iSCSI rescan at disconnect time is pretty
    dangerous to do. iSCSI rescan doesn't make volumes
    go away, it makes them show back up. So during Nova
    live migration it will make local scsi volumes show
    back up.

    Nova live migration first loops over volumes and calls
    disconnect_volume, then after that it loops over the volumes
    again and call's cinder to terminate_connection.

    If disconnect_volume does a rescan, then volumes there were
    removed during previous calls will show back up, because cinder's
    driver hasn't had a chance to terminate the connection yet
    from the array.

    Closes-Bug: #1483326
    Change-Id: I8708db16a4b9057b09abad9f956fa5a3bb4ba54c

[Eric Harney (eharney)]

Cinder stable/kilo: https://review.openstack.org/#/c/232847/