max_backup_count should keep valid backups #607
Description
If for any reasons your backups fails, the medusa purge task will delete what you have configured in max_backup_count, without taking the status in account.
So you can end up with only incompletes backups:
max_backup_count = 5:
$ medusa list-backups
[2023-07-26 17:41:44,995] INFO: Found credentials in shared credentials file: /etc/medusa/credentials-s3
scylladb01-production-20230719-1100 (started: 2023-07-19 11:00:40, finished: Incomplete [0 of 6 nodes finished])
scylladb01-production-20230721-1100 (started: 2023-07-21 11:01:24, finished: Incomplete [0 of 6 nodes finished])
scylladb01-production-20230723-1100 (started: 2023-07-23 11:00:55, finished: Incomplete [0 of 6 nodes finished])
scylladb01-production-20230724-1100 (started: 2023-07-24 11:01:20, finished: Incomplete [0 of 6 nodes finished])
scylladb01-production-20230725-1100 (started: 2023-07-25 11:00:43, finished: Incomplete [0 of 6 nodes finished])
In this case, we don't have any valid backup due to the purge, only incomplete ones.
We may be able to keep max_backup_count valid backups after a purge. Not sure how to implement that btw.
Yes I know, the real solution is to avoid failed backups :D
┆Issue is synchronized with this Jira Story by Unito
┆Issue Number: MED-26