[
https://issues.apache.org/jira/browse/FLINK-20671?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17251570#comment-17251570
]
zhuxiaoshang edited comment on FLINK-20671 at 12/18/20, 8:00 AM:
-----------------------------------------------------------------
https://issues.apache.org/jira/browse/FLINK-20213 didn't solve this problem and
still exists in 1.12.
I think the reason is that check bucket is active in
`notifyCheckpointComplete`,it is a async function, inProgressPart is not null
so the Bucket#isActive will return true.
was (Author: zhushang):
https://issues.apache.org/jira/browse/FLINK-20213 didn't solve this problem and
still exists in 1.12.
I think the reason is that check bucket is active in
`notifyCheckpointComplete`,it is a async function, so the Bucket#isActive will
return true.
> Partition doesn't commit until the end of partition
> ---------------------------------------------------
>
> Key: FLINK-20671
> URL: https://issues.apache.org/jira/browse/FLINK-20671
> Project: Flink
> Issue Type: Bug
> Components: Connectors / FileSystem
> Affects Versions: 1.12.0, 1.11.1
> Reporter: zhuxiaoshang
> Priority: Major
>
> According to DOC
> [https://ci.apache.org/projects/flink/flink-docs-master/dev/table/connectors/filesystem.html#partition-commit]
> the partition can be commited multiple times.
> Actually, the partition is commited until the end of partition.For example
> hourly partition, the 10th partition will be commited at 11 clock, even the
> `sink.partition-commit.delay` is 0s.
> In theory,if delay is 0s the partition should be commited on every checkpoint.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
