Duplicate entry for image_location

[Stuart McLaren]

Latest devstack:

1) create an image

$ glance image-create --name s2
+------------------+--------------------------------------+
| Property | Value |
+------------------+--------------------------------------+
| checksum | None |
| container_format | None |
| created_at | 2013-07-25T16:55:46 |
| deleted | False |
| deleted_at | None |
| disk_format | None |
| id | 2e1883ee-1b2d-417d-bc81-4d5edf306e2b |
| is_public | False |
| min_disk | 0 |
| min_ram | 0 |
| name | s2 |
| owner | ecc4670241e842c991b8737802468e20 |
| protected | False |
| size | 0 |
| status | queued |
| updated_at | 2013-07-25T16:55:46 |
+------------------+--------------------------------------+

2) set the location parameter

$ glance image-update 2e1883ee-1b2d-417d-bc81-4d5edf306e2b --location 'swift+http://service%3Aglance:<pw>@10.6.101.232:5000/v2.0/glance/aaa' --disk-format raw --container-format bare
+------------------+--------------------------------------+
| Property | Value |
+------------------+--------------------------------------+
| checksum | None |
| container_format | bare |
| created_at | 2013-07-25T16:55:46 |
| deleted | False |
| deleted_at | None |
| disk_format | raw |
| id | 2e1883ee-1b2d-417d-bc81-4d5edf306e2b |
| is_public | False |
| min_disk | 0 |
| min_ram | 0 |
| name | s2 |
| owner | ecc4670241e842c991b8737802468e20 |
| protected | False |
| size | 0 |
| status | active |
| updated_at | 2013-07-25T16:56:19 |
+------------------+--------------------------------------+

there are two entries with the specified location in the database (once of which is deleted)

ubuntu@devstack24:~$ mysql -uroot -p <passwd> -h localhost glance -e 'select * from image_locations where image_id="2e1883ee-1b2d-417d-bc81-4d5edf306e2b"'
+----+--------------------------------------+--------------------------------------------------------------------------+---------------------+---------------------+---------------------+---------+-----------+
| id | image_id | value | created_at | updated_at | deleted_at | deleted | meta_data |
+----+--------------------------------------+--------------------------------------------------------------------------+---------------------+---------------------+---------------------+---------+-----------+
| 10 | 2e1883ee-1b2d-417d-bc81-4d5edf306e2b | swift+http://service%3Aglance:<pw>@10.6.101.232:5000/v2.0/glance/aaa | 2013-07-25 16:56:19 | 2013-07-25 16:56:19 | 2013-07-25 16:56:19 | 1 | €}q. |
| 11 | 2e1883ee-1b2d-417d-bc81-4d5edf306e2b | swift+http://service%3Aglance:<pw>@10.6.101.232:5000/v2.0/glance/aaa | 2013-07-25 16:56:19 | 2013-07-25 16:56:19 | NULL | 0 | €}q. |
+----+--------------------------------------+--------------------------------------------------------------------------+---------------------+---------------------+---------------------+---------+-----------+

[Feilong Wang (flwang)]

I can recreate this issue.

[yong sheng gong (gongysh)]

-> return func(self, req, *args, **kwargs)
  /opt/stack/glance/glance/api/v1/images.py(686)update()
-> purge_props)
  /opt/stack/glance/glance/registry/client/v1/api.py(194)update_image_metadata()
-> return c.update_image(image_id, image_meta, purge_props)
> /opt/stack/glance/glance/registry/client/v1/client.py(158)update_image()
-> encrypted_metadata = self.encrypt_metadata(image_metadata['image'])

  /opt/stack/glance/glance/api/v1/images.py(690)update()
-> image_data)
  /opt/stack/glance/glance/api/v1/images.py(535)_handle_source()
-> image_meta = self._activate(req, image_id, location)
  /opt/stack/glance/glance/api/v1/images.py(474)_activate()
-> image_meta)
  /opt/stack/glance/glance/registry/client/v1/api.py(194)update_image_metadata()
-> return c.update_image(image_id, image_meta, purge_props)
> /opt/stack/glance/glance/registry/client/v1/client.py(158)update_image()
-> encrypted_metadata = self.encrypt_metadata(image_metadata['image'])

the registry is called twice:
one is pure update, and then activate()

[Zhi Yan Liu (lzy-dev)]

Actually this is not a defect but an improvement. From DB you can see the original location be marked to "deleted" status. Since under current sqlalchemy db driver implementation [1], glance delete all "original" image's location entries first then to re-add "new" entries.

[1] https://github.com/openstack/glance/blob/master/glance/db/sqlalchemy/api.py#L859

[Tushar Bankar (tushar-bankar)]

If the glance image update api is called and a new image location url is specified, the old locations are soft deleted and new locations are added to db, but this should not happen if both old and new locations are same urls, hence we need to fix this issue

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/80494

[Tushar Bankar (tushar-bankar)]

The fix for this Bug is covered in BP (https://blueprints.launchpad.net/glance/+spec/image-location-status). Hence abandoning.

[Viktor Serhieiev (vsergeyev)]

Blueprint image-location-status was already implemented. Should we mark this bug as fixed?