Cinder A/A error when cloning a volume or snapshot

[Alan Bishop]

When cloning a volume or creating one from a snapshot, the scheduler uses the 'resource_backend' of the parent volume/snapshot. When cinder is running active-active, the resource backend should reference the cluster name, not the host. That is not happening, and so the scheduler fails the request because it cannot a suitable backend.

The following logs aren't too revealing, but they show an example of how the scheduler fails when trying to clone a volume associated with a cluster named "tripleo."

2018-12-12 22:00:58.351 6 DEBUG cinder.scheduler.host_manager [req-fe59c33a-c4af-4ab6-b0f7-71392ab1ee1f 5066ba8b37174f2cb3b2621a097283cb ae7f527abbed478d9345d6fa8d7de9b9 - default default] Weighed [WeighedHost [host: node-0@tripleo_ceph#tripleo_ceph, weight: 0.0]] get_weighed_backends /usr/lib/python2.7/site-packages/cinder/scheduler/host_manager.py:500

2018-12-12 22:00:58.354 6 WARNING cinder.scheduler.filter_scheduler [req-fe59c33a-c4af-4ab6-b0f7-71392ab1ee1f 5066ba8b37174f2cb3b2621a097283cb ae7f527abbed478d9345d6fa8d7de9b9 - default default] No weighed backend found for volume with properties: {'name': u'tripleo', 'qos_specs_id': None, 'deleted': False, 'created_at': '2018-12-04T20:20:33.000000', 'updated_at': None, 'extra_specs': {}, 'is_public': True, 'deleted_at': None, 'id': u'e05f1612-2a5d-42ab-b5e2-573e2d61f18e', 'projects': [], 'description': None}

2018-12-12 22:00:58.357 6 INFO cinder.message.api [req-fe59c33a-c4af-4ab6-b0f7-71392ab1ee1f 5066ba8b37174f2cb3b2621a097283cb ae7f527abbed478d9345d6fa8d7de9b9 - default default] Creating message record for request_id = req-fe59c33a-c4af-4ab6-b0f7-71392ab1ee1f

2018-12-12 22:00:58.397 6 ERROR cinder.scheduler.flows.create_volume [req-fe59c33a-c4af-4ab6-b0f7-71392ab1ee1f 5066ba8b37174f2cb3b2621a097283cb ae7f527abbed478d9345d6fa8d7de9b9 - default default] Failed to run task cinder.scheduler.flows.create_volume.ScheduleCreateVolumeTask;volume:create: No valid backend was found. No weighed backends available: NoValidBackend: No valid backend was found. No weighed backends available

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/625052

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/625052
Committed: https://git.openstack.org/cgit/openstack/cinder/commit/?id=541168b86ec5b42db34524b3946b701ed8500737
Submitter: Zuul
Branch: master

commit 541168b86ec5b42db34524b3946b701ed8500737
Author: Alan Bishop <email address hidden>
Date: Thu Dec 13 08:42:59 2018 -0500

    Fix A/A 'resource_backend' when scheduling volumes

    Fix an issue with the 'resource_backend' included in the scheduler spec
    for creating a volume associated with another volume, snapshot, or
    group/cg. When running A/A, the 'resource_backend' must reference the
    cluster, not the host.

    Enhance the unit tests that cover this area. This includes fixing the
    'expected_spec' so it copies a dictionary rather than referencing it,
    so that external changes to the dictionary don't inadvertently update
    the unit test's expected results.

    Closes-Bug: #1808343
    Change-Id: I7d414844d094945b55a094a8426687595f22de28

[OpenStack Infra (hudson-openstack)]

This issue was fixed in the openstack/cinder 14.0.0.0rc1 release candidate.