How to append ORC file
ORC data files are subdivised in independent stripes; each stripe be created in a single atomic step. See the official documentation for details.
I don't believe you can directly append to an existing file on-the-fly. That would mean leaving a corrupt stripe (hence a corrupt file) in case of a job crash while writing.
But you can
- create a new ORC data file (which will contain 1..N stripes dependingon actual data volume vs.
orc.stripe.size
property) per reducer - then "concatenate" these data files -- and existing file(s) -- using Hive V0.14 and above