Aditya Kousik created KAFKA-21361:
-------------------------------------
Summary: Distribution snapshots for Kafka client metrics
Key: KAFKA-21361
URL: https://issues.apache.org/jira/browse/KAFKA-21361
Project: Kafka
Issue Type: Improvement
Components: clients, metrics
Reporter: Aditya Kousik
Assignee: Aditya Kousik
Kafka clients report latency and size only as an average and a maximum over the
last 30 seconds. Both don't show how values are spread out during that time.
One slow request can pull the average far from a typical one, and there's no
way to get a p50 or p99, or to combine numbers across clients or over an hour.
We should propose a distribution API that lets clients count how many values
fall into each range, so metrics backends such as Prometheus and OpenTelemetry
can show real histograms and compute any percentile over any period.
--
This message was sent by Atlassian Jira
(v10.3.26#10030026)
