Comment 3 for bug 1838541

Reviewed: https://review.opendev.org/673873
Committed: https://git.openstack.org/cgit/openstack/nova/commit/?id=ad4ef5af8bf7dc34caef21b0062ef9cc504bc216
Submitter: Zuul
Branch: master

commit ad4ef5af8bf7dc34caef21b0062ef9cc504bc216
Author: Matt Riedemann <email address hidden>
Date: Wed Jul 31 12:41:00 2019 -0400

    rt: soften warning case in _remove_deleted_instances_allocations

    During an instance_claim during initial server create or unshelve,
    the instance.host/node values will be set before the resource
    tracker has the instance in the tracked_instances dict. If
    _remove_deleted_instances_allocations is called with the instance
    before it's being tracked, a warning like this is logged:

      Jul 31 13:12:57.455904 ubuntu-bionic-rax-iad-0009534722
      nova-compute[31951]: WARNING nova.compute.resource_tracker
      [None req-d6f2ae97-d8f7-46f6-8974-b42aeb58302d None None]
      Instance 227c23cd-aeb2-4b5a-b001-21bd920a5e77 is not being actively
      managed by this compute host but has allocations referencing this
      compute host: {u'resources': {u'MEMORY_MB': 64, u'VCPU': 1,
      u'DISK_GB': 1}}. Skipping heal of allocation because we do not know
      what to do.

    This shows up quite frequently in CI runs (see the bug report for
    a logstash query) which means it should not be a warning.

    This change checks the instance task_state and if set then we only
    log a debug message rather than the warning since we can assume we
    are racing and the task will correct itself upon completion.

    Change-Id: I6db8bea6761b68c39e6332d4698d1f8312863758
    Closes-Bug: #1838541