Migrate volume raises ObjectActionError if volume is created using image

[Ankit Agrawal]

Steps to reproduce:

1. Create volume using image.
$ cinder create 1 --name <vol_name> --image <image_name or id>

2. Check "os-vol-host-attr:host" host on which volume is created.
$ cinder show <volume_id>

3. Try to migrate on different host. If volume is created on host-1, try to migrate it on host-2.
$ cinder migrate <volume_id> <host-2>

4. Return following response to the user.
Request to migrate volume <Volume: b5e28051-fdb2-412b-9363-0ac2aafa0390> has been accepted.

But raises following error in c-vol logs at source node. In this case the volume's "Migration Status" remains "starting" and we are not able to delete that volume any more.

2016-01-27 22:55:12.640 ERROR oslo_messaging.rpc.dispatcher [req-21c36b52-abf2-4fe6-af94-a19531ecfb42 809bd0cd851247efa98192f523fde3f7 a954d549046a49ff86e725bb18c1a09d] Exception during mes
sage handling: Object action save failed because: glance_metadata changed
2016-01-27 22:55:12.640 TRACE oslo_messaging.rpc.dispatcher Traceback (most recent call last):
2016-01-27 22:55:12.640 TRACE oslo_messaging.rpc.dispatcher File "/usr/local/lib/python2.7/dist-packages/oslo_messaging/rpc/dispatcher.py", line 143, in _dispatch_and_reply
2016-01-27 22:55:12.640 TRACE oslo_messaging.rpc.dispatcher executor_callback))
2016-01-27 22:55:12.640 TRACE oslo_messaging.rpc.dispatcher File "/usr/local/lib/python2.7/dist-packages/oslo_messaging/rpc/dispatcher.py", line 189, in _dispatch
2016-01-27 22:55:12.640 TRACE oslo_messaging.rpc.dispatcher executor_callback)
2016-01-27 22:55:12.640 TRACE oslo_messaging.rpc.dispatcher File "/usr/local/lib/python2.7/dist-packages/oslo_messaging/rpc/dispatcher.py", line 130, in _do_dispatch
2016-01-27 22:55:12.640 TRACE oslo_messaging.rpc.dispatcher result = func(ctxt, **new_args)
2016-01-27 22:55:12.640 TRACE oslo_messaging.rpc.dispatcher File "/usr/local/lib/python2.7/dist-packages/osprofiler/profiler.py", line 109, in wrapper
2016-01-27 22:55:12.640 TRACE oslo_messaging.rpc.dispatcher return f(*args, **kwargs)
2016-01-27 22:55:12.640 TRACE oslo_messaging.rpc.dispatcher File "/opt/stack/cinder/cinder/volume/manager.py", line 1858, in migrate_volume
2016-01-27 22:55:12.640 TRACE oslo_messaging.rpc.dispatcher volume.save()
2016-01-27 22:55:12.640 TRACE oslo_messaging.rpc.dispatcher File "/usr/local/lib/python2.7/dist-packages/oslo_versionedobjects/base.py", line 221, in wrapper
2016-01-27 22:55:12.640 TRACE oslo_messaging.rpc.dispatcher return fn(self, *args, **kwargs)
2016-01-27 22:55:12.640 TRACE oslo_messaging.rpc.dispatcher File "/opt/stack/cinder/cinder/objects/volume.py", line 296, in save
2016-01-27 22:55:12.640 TRACE oslo_messaging.rpc.dispatcher action='save', reason=_('glance_metadata changed'))
2016-01-27 22:55:12.640 TRACE oslo_messaging.rpc.dispatcher ObjectActionError: Object action save failed because: glance_metadata changed
2016-01-27 22:55:12.640 TRACE oslo_messaging.rpc.dispatcher

Affects: lvm and nfs drivers.

[Mitsuhiro Tanino (mitsuhiro-tanino)]

>>In this case the volume's "Migration Status" remains "starting" and we are not able to delete that volume any more.

Just a question. Can you change the status of volume which was failed to migrate volume using "cinder reset-state --reset-migration-status"?

[Steve Currie (steve-currie3463)]

I wanted to add a comment on this bug hope you don't mind but it might be somehow related and I recently started seeing this error after stacking from the Mitaka master this week
" Exception during message handling: Object action save failed because: glance_metadata changed"

 I see the same error in c-vol.log when attempting to change an image volume type from 1 backend to another - this used to work
for me
I simply create an image vol using the default Cirros image and I then select to change volume type, selecting on demand and point to another backend - (using cinder block storage)
At this point the volume is stuck in a retyping Status
thank you

[Ankit Agrawal (ankitagrawal)]

@Mitsuhiro Tanino, Yes I am able to reset migration status using "cinder reset-state --reset-migration-status" but the original issue related to "ObjectActionError" still remains there. Thanks!

[Mitsuhiro Tanino (mitsuhiro-tanino)]

Ankit,

Thank you for your confirmation. Please keep to fix the problem.

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/276747

[OpenStack Infra (hudson-openstack)]

Change abandoned by Szymon Borkowski (<email address hidden>) on branch: master
Review: https://review.openstack.org/276747
Reason: It seems that that issue is already fixed on current master.

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/276747
Committed: https://git.openstack.org/cgit/openstack/cinder/commit/?id=6ead3b86b201b1aed8930464a48df35f1492674a
Submitter: Jenkins
Branch: master

commit 6ead3b86b201b1aed8930464a48df35f1492674a
Author: Szymon Borkowski <email address hidden>
Date: Fri Feb 5 15:05:39 2016 +0100

    Fix for glance_metadata during volume migration

    This patch allows to transfer glance_metadata together with volume
    during the migration process, when the volume is created from the
    image.

    Co-Authored-By: Michal Dulko <email address hidden>
    Change-Id: Ifee5f2c53865076f43556d9dd6b6cbef77494bc8
    Closes-Bug: 1538926

[Thierry Carrez (ttx)]

This issue was fixed in the openstack/cinder 8.0.0.0rc1 release candidate.