backup delete fail

[FengJiankui]

an backup attribute has_dependent_backups is False, but exception is raised when delete the backup , and the execption as follows：

Failed to delete backup with name or ID '40b9d629-bc36-4094-913b-b6c32903c19c': Invalid backup: Incremental backups exist for this backup. (HTTP 400) (Request-ID: req-31bda700-7b9a-4f60-b6b8-3811cce25632)
1 of 1 backups failed to delete.

the backup detail as follows:

+-----------------------+--------------------------------------------+
| Property | Value |
+-----------------------+--------------------------------------------+
| availability_zone | nova1 |
| container | 40/b9/40b9d629-bc36-4094-913b-b6c32903c19c |
| created_at | 2020-12-10T07:51:07.000000 |
| data_timestamp | 2020-12-10T07:51:07.000000 |
| description | None |
| fail_reason | None |
| has_dependent_backups | False |
| id | 40b9d629-bc36-4094-913b-b6c32903c19c |
| is_incremental | False |
| name | test1 |
| object_count | 22 |
| size | 1 |
| snapshot_id | None |
| status | available |
| updated_at | 2020-12-10T07:51:21.000000 |
| volume_id | 50ae12d3-1608-44cd-9a3e-40b44d715537 |
+-----------------------+--------------------------------------------+

Found there is a backup with an error status, its parent backup is above backup.

+-----------------------+--------------------------------------+
| Property | Value |
+-----------------------+--------------------------------------+
| availability_zone | nova1 |
| container | None |
| created_at | 2020-12-10T07:52:43.000000 |
| data_timestamp | 2020-12-10T07:52:43.000000 |
| description | None |
| fail_reason | global name 'dd' is not defined |
| has_dependent_backups | False |
| id | 4de46669-a29c-4b92-9ca3-9fe527103dc8 |
| is_incremental | True |
| name | test2 |
| object_count | 0 |
| size | 1 |
| snapshot_id | None |
| status | error |
| updated_at | 2020-12-10T07:52:47.000000 |
| volume_id | 50ae12d3-1608-44cd-9a3e-40b44d715537 |

after delete the child backup, parent backup can be deleted

[Brian Rosmaita (brian-rosmaita)]

This needs to be fixed in Cinder, the python-cinderclient is just acting as an interface here.

Please read through the discussion at this week's cinder meeting:
http://eavesdrop.openstack.org/meetings/cinder/2020/cinder.2020-12-16-14.00.log.html#l-154

A few key points:

1) has_dependent_backups == False on the existing backup is correct, we don't want to count a failed backup as a dependent

2) we could implement a 'cascade' option for deletion that would delete a backup AND its dependent backups. This is probably a good idea independently of this particular bug. Its impact on this bug is that it would be useful in this situation, though a user wouldn't know to use it immediately because has_dependent_backups == False.

3) we could implement it so that if a backup is requested to be deleted, and it has a dependent backup in ERROR status, then it acts like 'cascade' even without the 'cascade' option.

Think it over, and we can discuss it in #openstack-cinder or at the weekly meeting or on the mailing list.

[FengJiankui (fengjiankui)]

Not only backups that failed to create, but also backups that failed to delete, etc. Why can these failed backups be considered dependencies?