[
https://issues.apache.org/jira/browse/FLINK-21129?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17271278#comment-17271278
]
Chesnay Schepler commented on FLINK-21129:
------------------------------------------
We do already limit operator names to 80 characters in the metric system, and
could easily do the same for task names.
At the same time, I feel like the discussion FLINK-20388 captures the root
issue in a better way; the operator name _should just not contain so much
information_.
> Long metric names cause Graphite to not store them
> --------------------------------------------------
>
> Key: FLINK-21129
> URL: https://issues.apache.org/jira/browse/FLINK-21129
> Project: Flink
> Issue Type: Bug
> Components: Runtime / Metrics
> Affects Versions: 1.11.1
> Reporter: Till Rohrmann
> Priority: Major
>
> While using the {{GraphiteReporter}} and a simple job with a window function,
> I noticed that Flink generates metric names which are too long for Graphite
> to handle.
> {code}
> githubCommits
> .global()
> .windowAll(TumblingEventTimeWindows.of(Time.days(1)))
> .apply(new MyWindowFunction())
> {code}
> The generated metric name was
> {{/opt/graphite/storage/whisper/172-20-3-229/taskmanager/flink-taskmanager-1-1/Commit-rate-per-day-program/TriggerWindow(TumblingEventTimeWindows(86400000),-ListStateDescriptor{name=window-contents,-defaultValue=null,-serializer=org-apache-flink-api-common-typeutils-base-ListSerializer@da6e90ca},-EventTimeTrigger(),-AllWindowedStream-runJob(CommitRateProgram-java:96))-->-Sink:-Unnamed/0/numRecordsInPerSecond/m1_rate.wsp}}
> The error only shows in the logs of Graphite.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
