HIVE not creating _SUCCESS ?

[Yang <te...@gmail.com>]

I'm very surprised to find that HIVE actually does not create a _SUCCESS in
its output file dir..

I had thought that since every MR job by default creates a _SUCCESS, and
HIVE is compiled into MR anyway, HIVE kind of naturally behaves the same as
any MR program. I verified that the wc example and streaming job both
create _SUCCESS.

I checked hadoop and hive code, seems that they use a
param mapreduce.fileoutputcommitter.marksuccessfuljobs  to control the file
creation, by default it's true. we never mess with this param anywhere, and
actually see it set to true on job tracker.


so, why is HIVE not creating _SUCCESS file?


thanks