I also thought that the checkpointing state size was growing due to growing
key space, i registered processing time timer on 'onevent' and wrote a
custom trigger and still the checkpointing state size was growing.
Our code is linked with flink 1.0.0 jar but was running on flink 1.1.1
(yarn session
Hi!
Count windows are a little tricky. If you have a growing key space, the
state keeps growing.
Lets say you have a bunch of values for key "A". You will fire the count
windows for every two elements and keep one element. If "A" never comes
again after that, the one element will still be kept ar
Hi,
I am noticing that the checkpointing state has been constantly growing for
the below subtask. Only the current active window elements should be
checkpointed ? why is it constantly growing ?
finalStream.keyBy("<>").countWindow(2,1)
.apply((_, _, input: scala.Iterable[], out: Collector[]) =>
