Leaked instance files when reverting migrations

[Lucian Petrut]

We're leaking instance files when reverting cold migrations/resizes and using multiple CSVs.

The manager tells us to skip deleting the instance files if the compute nodes use shared storage, assuming that the instance files will just stay in place while the instance moves back to the original node. When using multiple CSVs, the instance files may or may not be moved along with the instance (depending on a config option).

We can add a workaround at the driver level, ignoring the destroy_disks flag when destroying an instance having ‘resize_reverting’ task state. It’s safe to do so as we’re using a separate dir when backing up instance files (*_revert).

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/620908
Committed: https://git.openstack.org/cgit/openstack/compute-hyperv/commit/?id=976d316eb71cede6294f92b30e3011e4c618975c
Submitter: Zuul
Branch: master

commit 976d316eb71cede6294f92b30e3011e4c618975c
Author: Lucian Petrut <email address hidden>
Date: Thu Nov 29 13:51:14 2018 +0200

    Prevent reverting resizes from leaking files

    When reverting resizes, the manager will request instance files
    cleanup to be skipped if the hosts use shared storage, in which
    case we'd leak files if multiple CSVs are used.

    It's safe to cleanup the instance files when reverting resizes
    as long as we preserve the "*_revert" dir.

    Closes-Bug: #1805833

    Change-Id: I09428c2e25bed41f99b51c08a677d7152d9680ee

[OpenStack Infra (hudson-openstack)]

This issue was fixed in the openstack/compute-hyperv 9.0.0.0rc1 release candidate.