Manage volume returns with "No valid backend was found"

[Erlon R. Cruz]

After 'Cosmetic changes to scheduler'[1], if we try to manage a volume using the following procedure, we end up in a error due "No valid backend was found" while the backend is present:

$ cinder-manage host list
105-cinder-plugins nova
105-cinder-plugins@nfs nova
105-cinder-plugins@ceph nova
105-cinder-plugins@lvm nova
105-cinder-plugins@nfs1 nova
105-cinder-plugins@nfs2 nova
105-cinder-plugins@hnas-nfs nova

$ cinder manageable-list 105-cinder-plugins@lvm
+------------------------------------------------------------------+------+----------------+-----------------+-----------+------------+
| reference | size | safe_to_manage | reason_not_safe | cinder_id | extra_info |
+------------------------------------------------------------------+------+----------------+-----------------+-----------+------------+
| {u'source-name': u'volume-a8002db0-bc11-41e1-903c-29aae5e6257d'} | 3 | True | - | - | - |
+----------------------------------------------------

$ cinder manage --volume-type lvm 105-cinder-plugins@lvm volume-a8002db0-bc11-41e1-903c-29aae5e6257d
+--------------------------------+--------------------------------------+
| Property | Value |
+--------------------------------+--------------------------------------+
| attachments | [] |
| availability_zone | nova |
| bootable | false |
| consistencygroup_id | None |
| created_at | 2016-12-27T17:04:26.000000 |
| description | None |
| encrypted | False |
| id | dd37d2d6-5558-44e0-9ff9-7a86acfdb87f |
| metadata | {} |
| migration_status | None |
| multiattach | False |
| name | None |
| os-vol-host-attr:host | 105-cinder-plugins@lvm |
| os-vol-mig-status-attr:migstat | None |
| os-vol-mig-status-attr:name_id | None |
| os-vol-tenant-attr:tenant_id | bc752faf23834ca98641fe183eaa581c |
| replication_status | None |
| size | 0 |
| snapshot_id | None |
| source_volid | None |
| status | error |
| updated_at | 2016-12-27T17:04:26.000000 |
| user_id | 731db462453e4a42999f4b0cf6d73a35 |
| volume_type | lvm |
+--------------------------------+--------------------------------------+

In scheduler logs (added some extra prints):

Filter CapabilitiesFilter returned 1 host(s) from (pid=16623) get_filtered_objects /opt/stack/cinder/cinder/scheduler/base_filter.py:132
2016-12-27 15:17:45.846 DEBUG cinder.scheduler.filter_scheduler [req-6be5224f-d78f-4eef-9fa1-9a6fef124923 731db462453e4a42999f4b0cf6d73a35 bc752faf23834ca98641fe183eaa581c] Filtered [host '105-cinder-plugins@lvm#lvm': free_capacity_gb: 7.01, pools: None] from (pid=16623) _get_weighted_candidates /opt/stack/cinder/cinder/scheduler/filter_scheduler.py:345
2016-12-27 15:17:45.848 DEBUG cinder.scheduler.filter_scheduler [req-6be5224f-d78f-4eef-9fa1-9a6fef124923 731db462453e4a42999f4b0cf6d73a35 bc752faf23834ca98641fe183eaa581c] backend_passes_filters: backend_state.backend_id = 105-cinder-plugins@lvm#lvm,backend = 105-cinder-plugins@lvm from (pid=16623) backend_passes_filters /opt/stack/cinder/cinder/scheduler/filter_scheduler.py:139
2016-12-27 15:17:45.850 ERROR cinder.scheduler.manager [req-6be5224f-d78f-4eef-9fa1-9a6fef124923 731db462453e4a42999f4b0cf6d73a35 bc752faf23834ca98641fe183eaa581c] Failed to schedule_manage_existing: No valid backend was found. Cannot place volume d853ac6c-a806-4480-bdac-151f37f2a15a on 105-cinder-plugins@lvm

[1] https://review.openstack.org/#/c/346041/

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/421348

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/421438

[OpenStack Infra (hudson-openstack)]

Change abandoned by Erlon R. Cruz (<email address hidden>) on branch: master
Review: https://review.openstack.org/421348
Reason: Abandoning in favor of: https://review.openstack.org/#/c/421438/

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/421438
Committed: https://git.openstack.org/cgit/openstack/cinder/commit/?id=03ad26cdaf0a2e2fb0385ebadf85a4d42a785e8f
Submitter: Jenkins
Branch: master

commit 03ad26cdaf0a2e2fb0385ebadf85a4d42a785e8f
Author: Gorka Eguileor <email address hidden>
Date: Tue Jan 17 16:56:42 2017 +0100

    Fix volume manage

    After moving the manage operation to support Active/Active the scheduler
    no longer coud find the backend to place managed volumes.

    The issue comes from the fact that the user can request managing a
    volume just with the host, but if that backend is in a cluster the
    volume in the DB needs to be created with it, so in the API we create
    the volume using the service's host and cluster_name, but those are
    missing the pool information.

    The solution requires that the scheduler is capable of matching the
    backends even if we are searching for a backend without the pool as well
    as the scheduler updating the volume DB entry to include the pool.

    Co-Authored-By: Erlon R. Cruz <email address hidden>
    Closes-bug: #1652811
    Change-Id: I6110c0f93097f4e8c2b5fbc934924e3af1ec431b

[OpenStack Infra (hudson-openstack)]

This issue was fixed in the openstack/cinder 10.0.0.0b3 development milestone.