Github user bbende commented on the issue:
https://github.com/apache/nifi/pull/2820
I haven't looked at any of the code so keep that in mind :)
If we are talking about NiFi templates, then they are just examples of how
to use a processor or set of processors, which you can then give to someone to
help them get started. We have a wiki page where people have posted some for
the community:
https://cwiki.apache.org/confluence/display/NIFI/Example+Dataflow+Templates
I think the recent work @ottobackwards did with the syslog record readers
is a good example of what we can do for this case. Meaning later on we can
implement a NetflowV5RecordReader, which then lets you use stuff like
ConvertRecord to go from netflow to any format like JSON, CSV, Avro.
For now, if this processor always produces JSON, then documenting the
expected output format in additionalDetails.html seems sufficient to me.
I don't really know enough about what people do with netflow data to know
if keeping the raw message with the parsed one makes sense. I would say
typically we wouldn't keep the raw message with the parsed one, but you could
always have an option to control that if you thought it was necessary. The
processor can also have an original relationship as Otto suggested, although at
that point the original and parsed data are completely separate.
---
