Hmm, not entirely sure why the snippet might not work for your setup. Is it possible that you are running on a PostgreSQL version that's not compatible with the COPY command in the snippet? I was running PostgreSQL 9.5.x on RDS.

As for the flatten, I think there is an implicit conversion (https://github.com/scala/scala/blob/v2.10.5/src/library/scala/collection/TraversableOnce.scala#L389 that allowed me to flatten the Iterator[Array[Byte]] to an Iterator[Byte]. An IDE may complain about that or you may have that disabled somehow but it was working in a notebook on Spark 1.6.1 for me.