Hi!
Is the number of partitions of your Kafka topic smaller than the number of
parallelisms of the job? If yes some parallelism will be idle and will not
emit watermarks unless you set idleness for them. See [1].
I guess the original behavior of Flink 1.12 is not the expected behavior
but I don't
Hello,
Recently, I've decided to migrate one of my toy projects to Flink 1.14.0
and I realized that JVM cluster behaviour changed. It no longer respects
event time. On Flink 1.12.5 I didn't experience any issues. I tried some
debugging and it seems that InternalTimerServiceImpl's watermark is not
