I'm trying to understand this picture from https://ci.apache.org/projects/flink/flink-docs-stable/dev/event_timestamps_watermarks.html#watermark-strategies-and-the-kafka-connector
N|39
in it mean?W(33)
mean?W(17)
mean?(1)
or (2)
under Source, map, or window.Here's how I interpret this diagram:
N|39
indicates an event for key N with timestamp 39W(33)
is a Watermark on the wire with a timestamp of 33W(17)
show that W(17)
is part of the streaming data flowAlso, the four orange cylinders are Kafka partitions. Source(1)
and Source(2)
are each connected to two Kafka partitions. The FlinkKafkaConsumer
is tracking the maximum timestamp observed so far in each partition, and is emitting watermarks relative to the minimum of those per-partition maximums (after subtracting some bounded delay, in the case of a BoundedOutOfOrderness watermarking strategy).