[
https://issues.apache.org/jira/browse/FLINK-22631?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17437787#comment-17437787
]
Li commented on FLINK-22631:
----------------------------
As discussed in FLIP-147 [1], checkpoints are supported after tasks finished,
but metrics not. As we known, metricReporters are reported periodically,
default 10 seconds.
If sources are bounded, the final metrics may not be reported to metric system
like pushgateway when task finished. This makes users unable to obtain the
correct metrics。
I have created a JIRA[2] before, but at that time the indicator system of flink
was considered to be used to monitor a running application. But now in Flink
1.14,
batch and stream processing are unified. I think it's also important to get the
metrics of bounded or mixed jobs even if those jobs are finished.
Therefore, we propose to also proactively push metrics after tasks finished.
What do you think about? cc[~gaoyunhaii]
[1]
https://cwiki.apache.org/confluence/display/FLINK/FLIP-147%3A+Support+Checkpoints+After+Tasks+Finished
[2] https://issues.apache.org/jira/browse/FLINK-22631
> Metrics are incorrect when task finished
> ----------------------------------------
>
> Key: FLINK-22631
> URL: https://issues.apache.org/jira/browse/FLINK-22631
> Project: Flink
> Issue Type: Improvement
> Components: Runtime / Metrics
> Reporter: Li
> Priority: Minor
> Labels: auto-deprioritized-major
> Attachments: image-2021-05-11-20-13-25-886.png,
> image-2021-05-11-20-14-29-290.png, image-2021-05-19-10-10-29-765.png,
> image-2021-05-19-10-11-02-764.png
>
>
> MetricReporters are reported periodically, default 10 seconds. The
> final metrics may not be reported to metric system like pushgateway when task
> finished. This makes users unable to obtain the correct metrics。
>
> Maybe metricReporters should be reported once manually before closed.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
