[openstack-dev] [Cinder] Regarding manage_existing and unmanage

Duncan Thomas duncan.thomas at gmail.com
Thu Apr 10 17:20:15 UTC 2014


On 10 April 2014 09:02, Deepak Shetty <dpkshetty at gmail.com> wrote:

> Ok, agreed. But then when admin unmanages it, we shud rename it back to the
> name
> that it originally had before it was managed by cinder. At least thats what
> admin can hope
> to expect, since he is un-doing the managed_existing stuff, he expects his
> file name to be
> present as it was before he managed it w/ cinder.

I'd question this assertion. Once you've given a volume to cinder, it
is not an external volume any more, it is cinder's. Unmanage of any
volume should be consistent, regardless of whether it got into cinder
via a volume create or a 'cinder manage' command. It is far worse to
have unmanage  inconsistent at some point in the distant future than
it is for the storage admin to do some extra work in the short term if
he is experimenting with managing / unmanaging volumes.

As was discussed at the summit, manage / unmanage is *not* designed to
be a routine operation. If you're unmanaging volumes regularly then
you're not using the interface as intended, and we need to discuss
your use-case, not bake weird and inconsistent behaviour into the
current interface.

So, under what circumstances do you expect that the current behaviour
causes a significant problem?



More information about the OpenStack-dev mailing list