hawq-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Roman Shaposhnik <ro...@shaposhnik.org>
Subject Re: Apache HAWQ binary release and Hadoop vendor neutrality
Date Tue, 04 Apr 2017 01:15:32 GMT
On Mon, Apr 3, 2017 at 6:06 PM, Keaton Adams <kadams@pivotal.io> wrote:
> Probably due to a lack of understanding on my part.
> Is HAWQ tested with the Open Source Apache Hadoop project?  Will it work with plan vanilla
> http://hadoop.apache.org/
> http://hadoop.apache.org/releases.html

That's a great question. Lets unpack what it means to be tested with
"plain vanilla Hadoop".
If you look at what kind of artifacts Apache Hadoop community is
releasing you will see
that they consist of source code and a tarball with Java binary
convenice artifacts. E.g.
here's how it looks like for Hadoop 2.7.3:
there's no RPMs of DEBs. That happens on Apache Bigtop side when Apache Bigtop
puts together a distribution around Apache Hadoop.

Putting this together means that if HAWQ wants to test its RPMs against
"plain vanilla Hadoop" the only game in town are "plain vanilla Hadoop" RPMs
produced by Bigtop.

At the same time, HAWQ may decide to utilize Hadoop Java tarball somewhere in
its pipeline -- that'd be fine too, but probably won't be super useful
to end users.


View raw message