Details
-
Improvement
-
Status: Closed
-
Minor
-
Resolution: Fixed
-
1.7.1
Description
The documenation on Latency Tracking states:
All intermediate operators keep a list of the last n latencies from each source to compute a latency distribution. The sink operators keep a list from each source, and each parallel source instance to allow detecting latency issues caused by individual machines.
This is not correct anymore as this behavior is controlled by a configuration `metrics.latency.granularity` now.
Attachments
Issue Links
- links to