Hi everyone , please help me here
So I m using flink kafka connector to read data with custom watermark generation strategy (watermark updates on every event and 10sec out of orderness) now this kafka topic has 5 partitions and no key (hence messages are in round robin fashion and there can be out of orderness within the partitions as well) and i m trying to aggregate the data over tumbling window of 1 min but the issue is on every retry the window is giving different result (no new event being generated) this is because if in parallel stream the closing watermark causing event is processed first it will close the window and mark other events coming from different partition as late where as in another retry the closing watermark event is processed late and the event from another partition is no more late and added to aggregation
I want every time the result to be consistent how to achieve this in flink