Hi, We want to collect metrics for stream processing, typically counts aggregated over 1-minute buckets. However, we want these 1-minute boundaries determined by timestamps within the data records. Flink metrics do not handle this so we want to roll our own. How to proceed ? Some of our team members believe we can add methods in operator class code that can be called from the main Flink program, whist I am not sure this is supposed to be possible. Others consider using a side output stream with a record per input record and use Flink operators to do the aggregation. That may double the amount of records processed.
Can we extend the Flink metrics to provide such aggregation ? Regards, Joris
