0

I'm going to work on effect of parallelism on different basic operators in Flink. I tried to change the number of parallelism of some specific operators. It has been done but I don't know how can I obtain the processing time of a specific operator at all (with different number of parallelism) and the time of each instance of a specific operator. I saw that there are some metrics and APIs for different run time information but I completely new in using Flink and don't know what is the best way to obtain the processing time of an operator.

Folani
  • 31
  • 1
  • 6
  • Do you mean the time it takes to process a record or the clock time of an operator? – Richard Deurwaarder Oct 27 '18 at 14:39
  • Yes, I mean the processing time of each record with one specific operator. – Folani Nov 01 '18 at 06:15
  • this might answer your question (https://stackoverflow.com/a/44698429/2096986) although you are going to add latency when you try to measure it. I think that the best way is using the latency track default from Flink (which also adds latency by the way) or look at backpressure signals. Backpressure indicates that there is operators with high latency. – Felipe Dec 15 '20 at 16:34

0 Answers0