[
https://issues.apache.org/jira/browse/HIVE-17213?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Chao Sun updated HIVE-17213:
----------------------------
Attachment: HIVE-17213.5.patch
Patch v4 used INSERT OVERWRITE DIRECTORY, which didn't work for mini
SparkOnYarn test. It seems currently all the qfiles in
{{spark.only.query.files}} are in default run using the
{{MiniSparkOnYarnCliDriver}}, which is causing problem for this case since we
want it to run using {{TestSparkCliDriver}}.
Patch v5 further divides the existing property {{spark.only.query.files}} into
two:
- {{spark.only.query.files}}: contains all qfiles that are Spark only AND
should only be tested using {{TestSparkCliDriver}},
- {{miniSparkOnYarn.only.query.files}}: contains all qfiles that are Spark only
AND should only be tested using {{MiniSparkOnYarnCliDriver}}.
> HoS: file merging doesn't work for union all
> --------------------------------------------
>
> Key: HIVE-17213
> URL: https://issues.apache.org/jira/browse/HIVE-17213
> Project: Hive
> Issue Type: Bug
> Components: Spark
> Reporter: Chao Sun
> Assignee: Chao Sun
> Attachments: HIVE-17213.0.patch, HIVE-17213.1.patch,
> HIVE-17213.2.patch, HIVE-17213.3.patch, HIVE-17213.4.patch, HIVE-17213.5.patch
>
>
> HoS file merging doesn't work properly since it doesn't set linked file sinks
> properly which is used to generate move tasks.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
