Suppose we have a time window of 10 milliseconds and we use EventTime.
First, we determine how Flink can get time and watermark from
incoming messages, after that, we set a key for the stream and set a time
window.
aggregatedTuple
.assignTimestampsAndWatermarks(new SampleTimestampExtractor())
.keyBy(1).timeWindow(Time.milliseconds(1000))/*.countWindow(3)*/
.reduce()
My understanding of the data flow in this scenario is the following:
Flink advanced time according to the timestamp of Incoming data into the
aggregatedTuple variable while for each message get the timestamp and
watermark.
As I use Periodic Watermarks, according to default watermark interval
(200ms), watermarks will be updated. After that according to the
current watermark, data with the timestamp between the last watermark and
current watermark will be released and go to the next steps (keyBy,
timeWindow, reduce). If Flink received a data but an appropriate watermark
didn't emit for that data yet, Flink didn't send that data to the next
steps and keep it until it's appropriate watermark will be emitted.
Is that correct?
