Github user StefanRRichter commented on the issue:
https://github.com/apache/flink/pull/4722
I have one more talking point for this: the behaviour of the iterators for
RocksDB and heap backends are different now: for RocksDB we iterate over a
snapshot of the data and modifications concurrently (by that, I mean not in a
different thread, but just interleaved with the iteration in the same thread)
to the iteration will work and not be reflected in the iterator. In the heap
backend, the concurrent modifications will make the iterator fail. Is this a
problem for the intended use case and what is the expected behaviour?
---
