Comment 13 for bug 1712411

Reviewed: https://review.openstack.org/498861
Committed: https://git.openstack.org/cgit/openstack/nova/commit/?id=94b904abbad1c9655b6dec1a2e58d73bc913ed47
Submitter: Jenkins
Branch: master

commit 94b904abbad1c9655b6dec1a2e58d73bc913ed47
Author: Matt Riedemann <email address hidden>
Date: Tue Aug 29 12:28:43 2017 -0400

    Cleanup allocations on invalid dest node during live migration

    Starting in Pike, the scheduler creates an allocation on a
    chosen destination node during live migration. This happens
    before the destination node pre-checks occur, which could
    fail and trigger a retry to the scheduler. The allocations
    created in Placement against the failed destination node
    were not being cleaned up though.

    This change adds some cleanup code to the live migration task
    in conductor to clean the allocations for the failed destination
    node before retrying.

    The functional recreate test for the bug is updated to show
    that the bug is fixed now.

    Also updates the docstring in the SchedulerReportClient
    remove_provider_from_instance_allocation method so that we
    no longer have to enumerate all of the places that call it.

    Change-Id: I41e5e1fa9938b5e04f7e20f78ccd77eca658885f
    Closes-Bug: #1712411