Hi Yingda,
Flink on Tez was dropped in preparation of the 1.0 release as discussed on
the mailing list [1].
In addition to the reasons in the thread, Tez did not support pipelined
shuffles which are a prerequisite for low-latency stream processing, the
most common use case for Flink.
Cheers,
Fabi
All,
Back in version 0.9, there has been discussion about building Flink on top
of Tez:
https://ci.apache.org/projects/flink/flink-docs-release-0.9/setup/flink_on_tez.html
But looks like things have gone the other way and there has been no follow
up about leveraging Tez for yarn environment. Won