Github user NicoK commented on the issue:
https://github.com/apache/flink/pull/4506
I just tested this on a real yarn cluster with `./bin/flink run -m
yarn-cluster -yn 1 -ys 2 -yjm 768 -ytm 1024 -yD
taskmanager.memory.off-heap=true -yD taskmanager.memory.size=260 -yD
taskmanager.memory.preallocate=true ./examples/batch/WordCount.jar` (the
equivalent of the test) and verified that it was failing on Flink 1.3.2 but
working with the proposed fix.
I also tested the counterpart without the three memory options, i.e.
`./bin/flink run -m yarn-cluster -yn 1 -ys 2 -yjm 768 -ytm 1024
./examples/batch/WordCount.jar` and verified it is working with the PR changes
---
