[
https://issues.apache.org/jira/browse/FLINK-15438?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17005487#comment-17005487
]
Jörn Kottmann commented on FLINK-15438:
---------------------------------------
Yes, one of the counters I need is numRecordsIn for a specific operator. The
current implementation sends always the counter value for the total runtime of
the job, and not the counter value of for the reporting interval (the delta
since the last report).
I am happy to send a patch to resolve this issue.
> Counter metrics are incorrectly reported as total counts to DataDog
> -------------------------------------------------------------------
>
> Key: FLINK-15438
> URL: https://issues.apache.org/jira/browse/FLINK-15438
> Project: Flink
> Issue Type: Bug
> Components: Runtime / Metrics
> Affects Versions: 1.9.1
> Reporter: Jörn Kottmann
> Priority: Minor
>
> The Flink semantics of a counter are not matching with the counters in
> DataDog.
> In Flink a counter counts the total of increment and decrement calls.
> In DataDog a counter is a rate over the reporting interval.
> The Flink implementation of the DataDog reporter seems to send the Flink
> counter value each time the metrics are reported. Correct would be to send
> the delta of the counter since the last report.
> There are some features in DataDog which are easier to use if this could be
> fixed, e.g. alerts based on counters.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
