Hi Richard, thank you for rising this question! I’m in favor of dropping Hadoop 1 support in Sqoop 2 from multiple reasons:
1) Hadoop 2.0 is GA for very long time and offers significant advantage over Hadoop 1.0. 2) Keeping support for both versions is significant headache due to the huge differences (MR API, Security, …) 3) Keeping support for Hadoop 1.0 when we go “2.0” ourselves would mean to keep Hadoop 1.0 support for lifetime of the major version (as we don’t want to do backward incompatible changes) Hence I’m +1 on the idea and I’m wondering what are other people’s thoughts? Jarcec > On Jan 12, 2015, at 10:14 PM, Zhou, Richard <[email protected]> wrote: > > Hi all: > > AFAIK, the integration test of profile Sqoop2-hadoop100 has > failed for a long time, I cannot even remember when it started to fail. > Recently, Hari helped to disabled the polling of git for this build to avoid > FAILURE message on every commit. As We are all focusing on the implementation > of the feature on Hadoop 2, and no one, as I know (maybe not correct), are > working on Hadoop 1 support. I am wondering whether it is necessary to > support Hadoop 1 in Sqoop 2. > From my perspective, it makes sense that Sqoop 1 support > Hadoop 1, whilst Sqoop 2 support Hadoop 2. Actually it works like this way. > The Hadoop 1 support on Sqoop 2 does not work now. So I am encouraged to > raise this topic, is it ok to remove Hadoop 1 support on Sqoop 2? > Sorry for my rude to ask such a question, since the original > purpose for Sqoop 2 is to support all versions of Hadoop, like Hive. However, > it will cost huge effort on compatible work, which we could not offer now. As > a result, shall we remove the Hadoop 1 support for the time being at least? > And we could add it back in the future when we have enough resource on it. > Any comment will be appreciated. Thanks. > > Regards > Richard >
