Home | About | Sematext search-lucene.com search-hadoop.com
 Search Hadoop and all its subprojects:

Switch to Threaded View
Hadoop >> mail # dev >> Re: [DISCUSS] Getting to Hadoop 2.X beta

Copy link to this message
Re: [DISCUSS] Getting to Hadoop 2.X beta
To add to it, here's what have been experiences as a disconnect during the
release cycle of 2.0.4-alpha using BigTop 0.6.6 as the integration platform and
management stack of choice:

    1. Need for improved feedback from the downstream projects
    2. Need for improved feedback from the DevOps

Which boils down to getting sufficient buy-in from downstreams and devops
communities that they are now willing to consider branch-2 as being landing
spot for their Hadoop 2.X needs.

I belive the recent effort around 2.0.4-alpha and BigTop 0.6.0 might be
sufficient to convince both communities to:
     1. migrate their Hadoop 2.X profiles to at least 2.0.4-alpha
     2. start to run integration tests against Hadoop 2.X profile
     3. start to publish Maven artifacts
     4. in general agree to treat Hadoop 2.X issues with at least
        the same priority as the issues in default profiles are treated (which
        are mostly Hadoop 1.X profiles).

that would be a good start. BigTop can certainly help with a lot of
techicalities and moving parts of the process. That would require tighter
coordination between releases, of course.

The question I don't have an answer for is how to engage DevOps community.


On Wed, Apr 24, 2013 at 03:38PM, Roman Shaposhnik wrote:
> On Wed, Apr 24, 2013 at 9:42 AM, Steve Loughran <[EMAIL PROTECTED]> wrote:
> >> There's quite a few preconditions to be met for a piece of
> >> software to reach beta status. Quite a few of them are now
> >> being discussed in a neighboring thread 'Compatibility in
> >> Apache Hadoop' ( http://s.apache.org/VE1 ) but I'd like to
> >> kick off a broader discussion: what do you think a *complete*
> >> list should look like, before we can honestly signal this
> >> change to the rest of the world.
> >>
> >> It would be very nice if the things that are offered as part
> >> of the criteria are easily quantifiable, but lets brainstorm
> >> first anyway. We can always stack-rank and quantify later.
> >>
> > I need to get YARN-117 in before the service lifecycle is frozen forever
> We certainly need to get a list of blocker JIRAs for what would be
> our first beta release. This actually, raises a practical, question:
> what do we tag those?
> Or to put it another way: can we all agree that 2.0.5 should be our
> first beta release?
> Thanks,
> Roman.