[
https://issues.apache.org/jira/browse/FLINK-1722?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16848158#comment-16848158
]
Jingsong Lee commented on FLINK-1722:
-------------------------------------
Now, blink planner build batch sql on the StreamGraph, So we need to implement
finalizeGlobal to meet the requirements of Hadoop output formats(and the hive
output formats). We have fixed this problem in Blink's code.
[~aljoscha] [~rmetzger] [~maguowei] Can you take a fresh look at this issue?
CC: [~ykt836]
> Streaming not respecting FinalizeOnMaster for output formats
> ------------------------------------------------------------
>
> Key: FLINK-1722
> URL: https://issues.apache.org/jira/browse/FLINK-1722
> Project: Flink
> Issue Type: Bug
> Components: API / DataStream
> Reporter: Robert Metzger
> Priority: Major
>
> The Hadoop output formats execute a process in the end to move the produced
> files from a temp directory to the final location.
> The batch API is annotating output formats that execute something in the end
> with the {{FinalizeOnMaster}} interface.
> The streaming component is not respecting this interface. Hence,
> HadoopOutputFormats aren't writing their final data into the desired
> destination.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
