[openstack-dev] [nova] Can we deprecate the server backup API please?

Jim Rollenhagen jim at jimrollenhagen.com
Tue Nov 20 15:19:03 UTC 2018

On Tue, Nov 20, 2018 at 5:35 AM melanie witt <melwittt at gmail.com> wrote:

> On Mon, 19 Nov 2018 08:36:40 -0500, Jay Pipes wrote:
> > On 11/19/2018 08:15 AM, Matt Riedemann wrote:
> >> On 11/18/2018 6:51 AM, Alex Xu wrote:
> >>> Sounds make sense to me, and then we needn't fix this strange
> >>> behaviour also https://review.openstack.org/#/c/409644/
> >>
> >> The same discussion was had in the spec for that change:
> >>
> >> https://review.openstack.org/#/c/511825/
> >>
> >> Ultimately it amounted to a big "meh, let's just not fix the bug but
> >> also no one really cares about deprecating the API either".
> >
> > So we'll let the apathy of the past dictate the actions of the future.
> FWIW, my point of view on deprecating the API was/is, if people are
> using it to accomplish some task, why deprecate it if it's not hurting
> anything else? That is, I didn't want the aforementioned spec to amount
> to something like, someone proposes to fix a strange behavior they
> observed using the API and our answer is to deprecate the entire API.
> If it's clear that no one is using or benefiting from the API, then I am
> in support of deprecating it. But I haven't felt certain about whether
> that's the case so far.
> >> The only thing deprecating the API would do is signal that it probably
> >> shouldn't be used. We would still support it on older microversions. If
> >> all anyone cares about is signalling not to use the API then deprecation
> >> is probably fine, but I personally don't feel too strongly about it
> >> either way.
> >
> > Deprecating these kinds of APIs would, as I mentioned in my original
> > post, signal that the Nova team is actually serious about cleaning up
> > the cruft and getting rid of the debt from years past.
> >
> > And also that it is serious about Nova not being a dumping ground for
> > orchestration and out-of-scope APIs not related to a Compute API.
> That's fair enough, and if we can get a quick confirmation from
> operators we know (CERN, NeCTAR, Oath, Vexxhost, OVH, etc) that they
> don't use the API, I agree we should go ahead and deprecate it for the
> reasons you mention.

Oath doesn't have any internal documentation or regression testing around
the backup API - I can't guarantee it isn't used, but I think we'd be fine
with it going away.

// jim
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack-discuss/attachments/20181120/17acdbf2/attachment.html>

More information about the openstack-discuss mailing list