I think a heads up that it is a needed change given we want to evolve and
Hadoop has stopped releasing < 2.7.

On Thu, Apr 26, 2018 at 7:31 AM, Sean Busbey <[email protected]> wrote:

> How do we want to phrase the note to user@hbase? as a heads-up that
> this is a needed change due to the practicalities of having Hadoop as
> a dependency? or as something we're looking for feedback on from the
> user community before we declare consensus reached?
>
> On Thu, Apr 26, 2018 at 8:44 AM, Josh Elser <[email protected]> wrote:
> > +1 no objections here.
> >
> >
> > On 4/24/18 11:37 PM, Andrew Purtell wrote:
> >>
> >> Let's consider branching for HBase 1.5.
> >>
> >> The new feature justifying a minor increment is storage class aware
> >> placement (HBASE-19858), and a required update in Hadoop minimum
> version.
> >> It would be marked experimental. However, some of our Chinese colleagues
> >> have been running equivalent changes in production for a couple of years
> >> and all tests I've done with it look positive.
> >>
> >> Hadoop hasn't released 2.6 or below in ages. In all recent discussions I
> >> have found on their public lists, there are no plans to do so. They are
> >> still releasing 2.7. Therefore, I think it fair to conclude the earliest
> >> supported version of Hadoop by the Hadoop community is 2.7, and we can
> >> adopt this position too. When putting together a 1.5.0 release I would
> >> update documentation to reflect that the minimum supported Hadoop
> version
> >> is now 2.7.0, and put a note to this effect in the release notes.
> >>
> >> Are there any concerns?
> >>
> >> Related, a nice to have for the new HBase 1.5 code line would be
> >> successful compilation against Hadoop 2.9, 3.0 and 3.1.
> >>
> >>
> >
>



-- 
Best regards,
Andrew

Words like orphans lost among the crosstalk, meaning torn from truth's
decrepit hands
   - A23, Crosstalk

Reply via email to