Over time, the openstack databases will grow into infinity, because nothing is actually. Deleted resources will only be "soft deleted" in the database. All modules has commands for housekeeping.

nova

root@novaapi1:~# nova-manage db archive_deleted_rows --verbose --until-complete --purge

# if this fails, try to run first. Select a fairly recent date (you should get a hint from the error)
root@novaapi1:~# nova-manage db purge --before dd.mm.yyyy --all --verbose --all-cells

cinder

# <days> means, "purge all records older than <days>"
root@cinder1:~# cinder-manage db purge <days>

If you run into a bug, like this. Some manual housekeeping will be required:

# List all volumes that has been deleted, but the attachment is not marked as deleted:
MariaDB [cinder]> SELECT volumes.id,volumes.deleted_at,volume_attachment.id,volume_attachment.deleted_at FROM volumes INNER JOIN volume_attachment ON volumes.id=volume_attachment.volume_id WHERE volumes.deleted_at IS NOT NULL AND volume_attachment.deleted_at IS NULL;

+--------------------------------------+---------------------+--------------------------------------+------------+
| id                                   | deleted_at          | id                                   | deleted_at |
+--------------------------------------+---------------------+--------------------------------------+------------+
| 1588f08b-8014-4997-b874-9268fbf99315 | 2018-06-21 13:56:43 | ad9d1a73-4754-4db0-a9a6-d1f2a96d3c87 | NULL       |
+--------------------------------------+---------------------+--------------------------------------+------------+

# If this returns something, then update the deleted_at column for every volume_attachment listed. Just use the timstamp from volumes.deleted_at
UPDATE volume_attachment SET deleted=1, deleted_at='2018-06-21 13:56:43' WHERE id='ad9d1a73-4754-4db0-a9a6-d1f2a96d3c87';

heat

# <days> means, "purge all records older than <days>"
root@heatapi1:~# heat-manage purge_deleted <days>