[
https://issues.apache.org/jira/browse/FLINK-11318?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16742965#comment-16742965
]
Edward Rojas commented on FLINK-11318:
--------------------------------------
Yes [~kkl0u], I did verify that this could happen. This in the scenario I
mentioned in the description, when trying to migrate a job using BucketingSink
to now use StreamingFileSink, as I would like to use the same buckets that in
the previous version of the job, finished part files are overwritten (I'm using
HDFS).
> [Regression] StreamingFileSink can overwrite existing files
> -----------------------------------------------------------
>
> Key: FLINK-11318
> URL: https://issues.apache.org/jira/browse/FLINK-11318
> Project: Flink
> Issue Type: Bug
> Components: filesystem-connector
> Affects Versions: 1.6.3, 1.7.1
> Reporter: Edward Rojas
> Assignee: Kostas Kloudas
> Priority: Major
>
> StreamingFileSink does not validate if a file with the same name of the new
> part file already exists and this could result in overwriting a file.
> The BucketingSink perform this kind of validations in the "openNewPartFile"
> method here:
> https://github.com/apache/flink/blob/master/flink-connectors/flink-connector-filesystem/src/main/java/org/apache/flink/streaming/connectors/fs/bucketing/BucketingSink.java#L549-L561
> So this seems to be a regression an in the "old" BucketingSink this works.
>
> This can be problematic for example when migrating a job using Bucketing to
> to use the StreamingFileSink, file could be overwritten.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
