The Flink Kafka consumer's watermark extractor claims:.
"The watermark extractor will run per Kafka partition, watermarks will be merged across partitions in the same way as in the Flink runtime, when streams are merged."
Suppose we have simple situation where two streams merge into one. Is the ordering of the merged stream deterministic upon retries?
My pipeline relies on this property to guarantee determinism due to how watermarks are emitted.
Thanks for any input