Given an SLA of some x milliseconds of end-to-end latency, is there a way to know the end-to-end latency at record level (time taken by a record to be processed by the entire pipeline and written to sink) for a flink job? This latency tracking documentation gives only the percentile latencies from source to a particular operator or subtask. From an SLA perspective, I was looking for a way to know absolute latency at record-level so we can visualise it to see the distribution.
Locked
