[
https://issues.apache.org/jira/browse/FLINK-20674?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17251825#comment-17251825
]
Piotr Nowojski commented on FLINK-20674:
----------------------------------------
Ideally we would have both bytes/records produced and bytes/records sent, but
if I remember correctly, the "sent" variant wouldn't be that trivial because of
the threading model (netty threads?). Either way, it would be nice to have
them, if not for the fact that it might be also confusing users with overloaded
stats?
But IMO just renaming this would already help a bit. If we had also tool tips
in the UI explaining the stats, that would solve the problem.
> Wrong send/received stats with UNION ALL
> ----------------------------------------
>
> Key: FLINK-20674
> URL: https://issues.apache.org/jira/browse/FLINK-20674
> Project: Flink
> Issue Type: Bug
> Components: Table SQL / Runtime
> Affects Versions: 1.12.0, 1.11.3
> Reporter: Nico Kruber
> Priority: Major
>
> When using {{UNION ALL}} to union the same table twice , the number of
> records and bytes sent is just half of what the next task receives:
> Reproducible with this:
> {code}
> CREATE TEMPORARY TABLE test (
> `number` SMALLINT
> )
> WITH (
> 'connector' = 'datagen',
> 'rows-per-second' = '1'
> );
> SELECT * FROM (
> (SELECT * FROM test)
> UNION ALL
> (SELECT * FROM test)
> )
> {code}
> Arguably, the use case is not too useful but other combinations may be
> affected, too.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
