Hi, Thank you Jay for your valuable feedback and suggestions! I got your point. Current BOM of k8s stuffs, it is just our previous practice on development of Bigtop stack... No special intention! at this point, I believe it's slightly meaningless at all. As you pointed out, we should focus on analytics stack, interoperability and practical blueprints of analytics architecture. We won't need any strict conditions or versions of things related k8s unless we have to keep our own customized k8s distribution or tools. Definitely hacking k8s itself, that is *out of scope*
So, feel free to correct or update the BOM and docs. If there are things we don't need, just remove it to avoid confusion. Thanks, Youngwoo On Wed, Oct 16, 2019 at 9:57 PM jay vyas <[email protected]> wrote: > hi youngwoo and others !!! , saw the bom JIRAs, !!! Thanks for bringing > this to life. > > the k8s components (like helm and stuff) are great to put in the bom, but > typically most people will *just *want the yaml snippets examples/docs on > how to integrate spark and minio / zookeeper / etc . most helm k8s distros > under the hood will work fine together, and even if they dont, that is > really secondary. I think theres *alot *of integration hacking to do to > make sure spark/kafka/hive/hbase versioning schemes all work well for end > to end analytics. > > just a quick thought , but obviously the more you put in the bom the > better, and im sure some people will find some value in our packages for > helm kubespray and so on. > > I kinda think the BOM should, though, mostly focus on things *above* k8s, > like, the stuff bigtop already packages, and specifically, YAMLs, helm > charts, and docker images for those projects. > > Again these are just my thoughts - your driving it so - any k8s solution is > gonna be light years ahead of what we have now, so im very supportive > either way ! > > -- > jay vyas >
