Comment 22 for bug 1718512

Reviewed: https://review.openstack.org/560143
Committed: https://git.openstack.org/cgit/openstack/nova/commit/?id=757dbd17cf37aecea005dfdc954bf50bbddedd95
Submitter: Zuul
Branch: stable/queens

commit 757dbd17cf37aecea005dfdc954bf50bbddedd95
Author: Matt Riedemann <email address hidden>
Date: Fri Apr 6 20:28:53 2018 -0400

    Don't persist RequestSpec.retry

    During a resize, the RequestSpec.flavor is updated
    to the new flavor. If the resize failed on one host
    and was rescheduled, the RequestSpec.retry is updated
    for that failed host and mistakenly persisted, which
    can affect later move operations, like if an admin
    targets one of those previously failed hosts for a
    live migration or evacuate operation.

    This change fixes the problem by not ever persisting
    the RequestSpec.retry field to the database, since
    retries are per-request/operation and not something
    that needs to be persisted.

    Alternative to this, we could reset the retry field
    in the RequestSpec.reset_forced_destinations method
    but that would be slightly overloading the meaning
    of that method, and the approach taken in this patch
    is arguably cleaner since retries shouldn't ever be
    persisted. It should be noted, however, that one
    advantage to resetting the 'retry' field in the
    RequestSpec.reset_forced_destinations method would
    be to avoid this issue for any existing DB entries
    that have this problem.

    The related functional regression test is updated
    to show the bug is now fixed.

    NOTE(mriedem): This backport also includes change
    I61f745667f4c003d7e3ca6f2f9a99194930ac892 squashed
    into it in order to not re-introduce that bug.

    Change-Id: Iadbf8ec935565a6d4ccf6f36ef630ab6bf1bea5d
    Closes-Bug: #1718512
    (cherry picked from commit 6647f11dc1aba89f9b0e2703f236a43f31d88079)