Hi Michiel, I really appreciate your offer, that sounds very useful for the HDP users. But to be honest, I'm afraid I'm a bit reluctant to include it into Bigtop, due to the following reasons.
1. Ambari has already been retired this January [1]. Keeping retired software as a Bigtop component brings us security risks and maintenance cost. Especially, Ambari depends on several obsolete softwares such as Python2 and Bower. So I personally think we should drop Ambari (and also Mpack) at some point in the future, unless it revives. 2. Cloudera has changed the license of their product [2], so I'm not sure if we could do that. Doesn't it violate their license to install and use HDP without subscription? If the new Mpack is derived from their product, could we include it in our distributions from the viewpoint of license compatibility [3]? But the above is just my personal opinion, so I'd like to hear from others. [1]: https://lists.apache.org/thread/m5jrpn4j28kn3wfn4zzxvy0g450vdlr1 [2]: https://www.cloudera.com/downloads/paywall-expansion.html [3]: https://www.apache.org/legal/resolved.html Kengo Seki <[email protected]<http://apache.org/>> Hi Kengo, Sorry if you get this twice; I wasn't subscribed when I made my initial reply. In regards to #2, licensing. Even if HDP did change the license when they expanded their paywall that could only apply to newly downloaded copies. The terms of the Apache, GPL, and most other open source licenses do not allow the author to rescind rights to users who previously obtained the software. So anyone who created a mirror of the legacy HDP repositories prior to the paywall deployment (3.1.4 and earlier, I believe) would benefit from Michiel's work. Users without an existing legacy mirror of the HDP repositories could only benefit if they paid to access Cloudera's repositories. I'm not sure how many existing HDP users would like to move to BigTop, but this would be a nice stepping stone. The lack of a working MPack in BigTop is currently limiting our use of BigTop. Are there any obvious replacements for Ambari going forward? Or would we just be left without a stack-management API and GUI? The dependence upon Python 2.7 is not good, but also not a showstopper for our deployments on CentOS 7/RedHat 7, which continues to backport fixes to Python 2.7 (because a whole bunch of that OS's software relies upon Python 2.7). John Gibson
