I am trying to check the flink performance using kafka source.
My process logic contains the aggregating the values.
When I see the metrics, .aggregate after .keyBy and .window (TumblingWindow for 60sec) caused the backpressured and it eventually affects the kafka source throughput.
I think that aggregting is just count the value using incoming data but my experiment showed me that it cause the upstream performance.
Is this real in flink?