[
https://issues.apache.org/jira/browse/FLINK-26126?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17492357#comment-17492357
]
Leonard Xu edited comment on FLINK-26126 at 2/15/22, 4:26 AM:
--------------------------------------------------------------
[~ruanhang1993] The blocker priority means we have to fix it in the coming
release.
CC: [~arvid] Could you help take a look this one?
was (Author: leonard xu):
[~ruanhang1993] The blocker priority means we have to fix to the coming
release.
CC: [~arvid] Could you help take a look this one?
> Sink V2 will cause error numRecordsOut metric
> ---------------------------------------------
>
> Key: FLINK-26126
> URL: https://issues.apache.org/jira/browse/FLINK-26126
> Project: Flink
> Issue Type: Bug
> Components: Connectors / Kafka
> Affects Versions: 1.15.0
> Reporter: Hang Ruan
> Priority: Blocker
>
> We found that the new sink v2 interface will have a wrong numRecordsOut
> metric for the sink writers. We send a fixed number of records to the source,
> but the numRecordsOut of the sink continues to increase by the time.
> The problem lies in the method `emitCommittables` in the class
> `SinkWriterOperator`. The field `output` in its parent class
> `AbstractStreamOperator` uses the same counter object as the `KafkaWriter`.
> It will cause the numRecordsOut increasing when doing the checkpoint.
> I found this problem when we implement the metric test in the testframe, now
> I disable this metric test in the
> [PR|https://github.com/apache/flink/pull/18496] We could reopen this test
> case after the fix.
--
This message was sent by Atlassian Jira
(v8.20.1#820001)
