You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@fluss.apache.org by "platinumhamburg (via GitHub)" <gi...@apache.org> on 2026/10/23 14:24:53 UTC
[I] [metrics] Define empty-window semantics for expiring histogram reporters [fluss]
platinumhamburg opened a new issue, #4828:
URL: https://github.com/apache/fluss/issues/4828
### Search before asking
- [x] I searched the issues and found the related sample-expiration issue #4139 and PR #4167. They do not define how reporters should represent an empty sample window.
### Motivation
Fluss histograms currently retain the last N observations indefinitely. Issue #4139 covers aging samples out, but expiration alone leaves an observable ambiguity: Prometheus exports quantile samples without recent observations, InfluxDB writes distribution fields on every report, and JMX exposes long-valued min/max that become zero for an empty sample. A cumulative count cannot indicate whether the current distribution contains any samples. This makes an idle request latency look like an active latency or a zero latency.
### Solution
Define one bounded sample-window contract for Fluss-owned histograms: last N observations within a fixed maximum age, while the lifetime count remains cumulative. On read, expire old observations even if no new request arrives. Reporters must represent an empty window explicitly: Prometheus omits quantile series while retaining cumulative _count; InfluxDB publishes the cumulative count and window sample count without distribution fields; JMX adds a window sample count so its existing numeric attributes can be interpreted correctly. Cover idle expiration and recovery in deterministic tests.
Keep RocksDB native statistics, client last-value gauges, and dashboard queries outside this issue. This issue is related to #4139 and extends its reporter contract.
### Anything else?
The in-flight PR #4167 overlaps the sample-expiration portion. A single end-to-end PR may supersede that portion while adding reporter handling.
### Willingness to contribute
- [x] I am willing to submit a PR!
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: issues-unsubscribe@fluss.apache.org.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [I] [metrics] Define empty-window semantics for expiring histogram reporters [fluss]
Posted by "polyzos (via GitHub)" <gi...@apache.org>.
polyzos closed issue #4828: [metrics] Define empty-window semantics for expiring histogram reporters
URL: https://github.com/apache/fluss/issues/4828
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: issues-unsubscribe@fluss.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org