Hello! I have some comments below, but you might also consider posting this question to openstack-...@lists.openstack.org as I'd be curious as to the developer opinion on this question. This list is mostly for users/operators of OpenStack clouds.

Tiantian Gao wrote:
Hi stackers,

Now when reclaim_instance_interval is great than 0, the delete action
will become soft_delete.

How about let user to specify the reclaim_instance_interval? does that
make sense? or anybody have the same requirement?

Because currently implemented the reclaim_instance_interval is system
level config. User can't control the expires time of instance. For
example, I want keep a important instance for 1 weeks, and some not so
important about 1 hour. That requirement make sense to me. I want to
know you options.

As a user and an operator I agree with your assessment that this feature should move from system level to user level (but retain some sort of max soft_delete timeout which can be configured at a system level.

Perhaps something like:

DELETE /v2/1234/servers/e14b69ce-eb17-4998-a8cb-114640db328b?delay_until=<utc_date>

If "delay_until" is set too far in the future, this DELETE call could return 400 Bad Request with details about the operator's maximum allowed reclaim interval.


For security issues, we can specify the system level
reclaim_instance_interval_max, instance are not allow to keep longer
than the max interval.

reclaim_instance_interval_min: If > 0 then all deletes will be "soft" deletes unless "hard" delete is explicitly requested by user.

reclaim_instance_interval_max: If = 0 then "soft" deletes are effectively disabled. Users can never request a "soft" delete farther in the future than this value allows.


--
Best regards,
Gtt

_______________________________________________
Mailing list: %(list_owner)s
Post to     : %(list_post)s
Unsubscribe : %(list_unsubscribe)s
More help   : %(list_help)s

_______________________________________________
Mailing list: %(list_owner)s
Post to     : %(list_post)s
Unsubscribe : %(list_unsubscribe)s
More help   : %(list_help)s

Reply via email to