Thanks Sergey. Just wanted update our experience, may be it could help others.
Yes as you said Phoenix Bulk CSV upload using MR was the fastest, but this required additional step of generating CSV. We Went ahead with using Spark phoenix connector as anyway we need transform input data, so spark does hive(input) -> transform -> phoenix. -- Sent from: http://apache-phoenix-user-list.1124778.n5.nabble.com/
