Unexpected behavior when upload a snapshot-volume to vol-backed image

[Yikun Jiang]

Unexpected behavior when upload a snapshot-volume to vol-backed image

0. Configure the volume-backed image, according [1]
1. Create a volume from snapshot.
# cinder create --snapshot 2609e615-323f-4927-923f-028848d45940 1
# cinder list --fields snapshot_id
+--------------------------------------+--------------------------------------+
| ID | Snapshot_Id |
+--------------------------------------+--------------------------------------+
| ae4ce4dc-23b6-49a8-aeae-90c34b7e5db9 | 2609e615-323f-4927-923f-028848d45940 |
+--------------------------------------+--------------------------------------+
2. Delete snapshot
# cinder snapshot-delete 2609e615-323f-4927-923f-028848d45940
3. Failed to upload volume to image.
# cinder upload-to-image ae4ce4dc-23b6-49a8-aeae-90c34b7e5db9 test_image

Because:
Traceback (most recent call last):
  File "/usr/local/lib/python2.7/dist-packages/taskflow/engines/action_engine/executor.py", line 53, in _execute_task
    result = task.execute(**arguments)
  File "/opt/stack/cinder/cinder/volume/flows/manager/create_volume.py", line 1034, in execute
    **volume_spec)
  File "/opt/stack/cinder/cinder/volume/flows/manager/create_volume.py", line 448, in _create_from_snapshot
    snapshot = objects.Snapshot.get_by_id(context, snapshot_id)
  File "/opt/stack/cinder/cinder/objects/base.py", line 346, in get_by_id
    orm_obj = db.get_by_id(context, cls.model, id, *args, **kwargs)
  File "/opt/stack/cinder/cinder/db/api.py", line 1789, in get_by_id
    return IMPL.get_by_id(context, model, id, *args, **kwargs)
  File "/opt/stack/cinder/cinder/db/sqlalchemy/api.py", line 183, in wrapper
    return f(*args, **kwargs)
  File "/opt/stack/cinder/cinder/db/sqlalchemy/api.py", line 7074, in get_by_id
    return _GET_METHODS[model](context, id, *args, **kwargs)
  File "/opt/stack/cinder/cinder/db/sqlalchemy/api.py", line 183, in wrapper
    return f(*args, **kwargs)
  File "/opt/stack/cinder/cinder/db/sqlalchemy/api.py", line 3073, in snapshot_get
    return _snapshot_get(context, snapshot_id)
  File "/opt/stack/cinder/cinder/db/sqlalchemy/api.py", line 183, in wrapper
    return f(*args, **kwargs)
  File "/opt/stack/cinder/cinder/db/sqlalchemy/api.py", line 3066, in _snapshot_get
    raise exception.SnapshotNotFound(snapshot_id=snapshot_id)
SnapshotNotFound: Snapshot 2609e615-323f-4927-923f-028848d45940 could not be found.

The root reason of this failure is that the clone image is based on **snapshot** but not **snapshot-volume**.

Another impact of this bug is:
If we already make some data change on a snapshot-volume, then we export this snapshot-volume to image,
this image will lose all change data, beacause this image is based on **snapshot** but not **snapshot-volume**.

[1] https://docs.openstack.org/cinder/latest/admin/blockstorage-volume-backed-image.html

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/606235

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/606235
Committed: https://git.openstack.org/cgit/openstack/cinder/commit/?id=9d43d22bf3c887542e475a212f662e73ad44e65e
Submitter: Zuul
Branch: master

commit 9d43d22bf3c887542e475a212f662e73ad44e65e
Author: Yikun Jiang <email address hidden>
Date: Sat Sep 29 09:44:22 2018 +0800

    Fix unexpected behavior in _clone_image_volume

    Current, if we upload a snapshot-volume to image, then create
    a new volume from this image, this new volume is created from
    original snapshot rather than snapshot-volume.

    That means if we already make some data change on a
    snapshot-volume, then we export this snapshot-volume to image,
    this image will lose all change data.

    This patch try to fix this, we skip the "snapshot_id" and
    "source_volid" when we create tmp image volume.

    Closes-bug: 1795098

    Change-Id: Icf0dadc1ce4d008ff437ca5dcde1817d2e560e44

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: stable/rocky
Review: https://review.openstack.org/639755

[OpenStack Infra (hudson-openstack)]

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

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/639755
Committed: https://git.openstack.org/cgit/openstack/cinder/commit/?id=1a8021a281c9ac2bd7e3acd25ba14ed3c70d8029
Submitter: Zuul
Branch: stable/rocky

commit 1a8021a281c9ac2bd7e3acd25ba14ed3c70d8029
Author: Yikun Jiang <email address hidden>
Date: Sat Sep 29 09:44:22 2018 +0800

    Fix unexpected behavior in _clone_image_volume

    Current, if we upload a snapshot-volume to image, then create
    a new volume from this image, this new volume is created from
    original snapshot rather than snapshot-volume.

    That means if we already make some data change on a
    snapshot-volume, then we export this snapshot-volume to image,
    this image will lose all change data.

    This patch try to fix this, we skip the "snapshot_id" and
    "source_volid" when we create tmp image volume.

    Closes-bug: 1795098

    Change-Id: Icf0dadc1ce4d008ff437ca5dcde1817d2e560e44
    (cherry picked from commit 9d43d22bf3c887542e475a212f662e73ad44e65e)

[OpenStack Infra (hudson-openstack)]

This issue was fixed in the openstack/cinder 13.0.5 release.