[
https://issues.apache.org/jira/browse/FLINK-5118?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15823968#comment-15823968
]
ASF GitHub Bot commented on FLINK-5118:
---------------------------------------
Github user StephanEwen commented on the issue:
https://github.com/apache/flink/pull/3106
+1 to this approach.
Some suggestions for small performance improvements, not specific to this
case, but also applicable to other cases:
- It may make sense to type the counters at that level to the
"SimpleCounter", if we anyways provide them. The TaskIOMetricGroup could type
its counters to "SimpleCounter" as well. Since that class it not final, and
future mocks/checks can be implemented on that class as well
- We can make sure that the field `counter` is always initialized by
initially assigning a standalone SimpleCounter. That way we could drop the null
checks in the code.
> Inconsistent records sent/received metrics
> ------------------------------------------
>
> Key: FLINK-5118
> URL: https://issues.apache.org/jira/browse/FLINK-5118
> Project: Flink
> Issue Type: Bug
> Components: Metrics, Webfrontend
> Affects Versions: 1.2.0, 1.3.0
> Reporter: Ufuk Celebi
> Assignee: Chesnay Schepler
> Fix For: 1.2.0, 1.3.0
>
>
> In 1.2-SNAPSHOT running a large scale job you see that the counts for
> send/received records are inconsistent, e.g. in a simple word count job we
> see more received records/bytes than we see sent. This is a regression from
> 1.1 where everything works as expected.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
