I also lean to the second option since v1alpha1 is a preview release.
But we need to be more careful to introduce other incompatible changes
after v1beta1.

Maybe we also need a simple manual for how to upgrade the operator
especially when the CRD version changed.

Best,
Yang

Gyula Fóra <gyula.f...@gmail.com> 于2022年4月5日周二 14:45写道:

> I agree that we can go with *v1beta1 .*
>
> Gyula
>
>
>
> On Mon, Apr 4, 2022 at 9:53 AM Őrhidi Mátyás <matyas.orh...@gmail.com>
> wrote:
>
> > I prefer the 2nd option Gyula, shouldn't we target v1beta1 already?
> >
> > On Mon, Apr 4, 2022 at 8:47 AM Gyula Fóra <gyf...@apache.org> wrote:
> >
> > > Hi all!
> > >
> > > With the preview release almost (hopefully today) officially out we
> need
> > to
> > > bump the CRD version of the deployment objects.
> > >
> > > As there will likely be a second preview release "soon" I suggest the
> > next
> > > version to be v1alpha2.
> > >
> > > The other question is what to do with the current v1alpha1 version as
> > there
> > > will probably be breaking changes.
> > >
> > > 1. We can either keep it and add all the versioning, migration logic
> > > necessary (
> > >
> > >
> >
> https://kubernetes.io/docs/tasks/extend-kubernetes/custom-resources/custom-resource-definition-versioning/
> > > )
> > > 2. Simply drop it and notify users that a one time manual migration
> step
> > is
> > > necessary when they upgrade on the next release.
> > >
> > > Given that we have clearly communicated this as a preview/experimental
> > > release, all interfaces are annotated accordingly and even the current
> > crd
> > > version implies this, I suggest that we go with option number 2 and
> > simply
> > > drop the previous version for the next release.
> > >
> > > This will avoid a lot of extra boilerplate and unnecessary complexity
> at
> > > this point when we want to do the opposite, stabilize and simplify the
> > API.
> > >
> > > Let me know what you think so we can proceed with
> > > https://issues.apache.org/jira/browse/FLINK-27005
> > >
> > > Cheers,
> > > Gyula
> > >
> >
>

Reply via email to