[
https://issues.apache.org/jira/browse/HIVE-22707?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
László Bodor updated HIVE-22707:
--------------------------------
Description: Given a scenario, when there are 2 different buckets, and the
output is written to another bucket than the source. Under specific
circumstances, FileSinkOperator is only used in Reducer stages, and if a root
work is a merge join work, it's not scanned for output uris/paths, therefore
needed delegation tokens are not fetched for e.g. the output s3 bucket.
> MergeJoinWork should be considered while collecting DAG credentials
> -------------------------------------------------------------------
>
> Key: HIVE-22707
> URL: https://issues.apache.org/jira/browse/HIVE-22707
> Project: Hive
> Issue Type: Bug
> Reporter: László Bodor
> Priority: Major
>
> Given a scenario, when there are 2 different buckets, and the output is
> written to another bucket than the source. Under specific circumstances,
> FileSinkOperator is only used in Reducer stages, and if a root work is a
> merge join work, it's not scanned for output uris/paths, therefore needed
> delegation tokens are not fetched for e.g. the output s3 bucket.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
