[DISCUSS] Dropping Hadoop 2.7 runtime support?

[Wei-Chiu Chuang <we...@apache.org>]

Hi community,

I want to bring up the Hadoop2 support for discussion -- is it possible to
drop 2.7?

I am asking because a number of features that I worked on requires Hadoop 3
APIs. Some of them can work with Hadoop 2.10 APIs. Examples: CompositeCRC
file checksum, StreamCapabilities APIs.

There are other useful Hadoop 3 APIs that will certainly improve Ozone
performance : ByteBufferPositionedReadable, Vectored IO... etc.

Dropping Hadoop 2.7 runtime, moving to a Hadoop 3 only runtime model will
enable a lot of innovations. I see that Ozone's doc on Spark support
requires hadoop 2.7 runtime (
https://ci-hadoop.apache.org/view/Hadoop%20Ozone/job/ozone-doc-master/lastSuccessfulBuild/artifact/hadoop-hdds/docs/public/recipe/sparkozonefsk8s.html),
but the latest Spark 3 can work with Hadoop3 runtime already. So is there
any use of Hadoop2 runtime compatibility?

Thanks