[
https://issues.apache.org/jira/browse/OOZIE-3069?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16178665#comment-16178665
]
Peter Cseh commented on OOZIE-3069:
-----------------------------------
Oozie uses MR as a launcher because of MR1 support. This was dropped after
Oozie 4.3.0, the next release will use it's own Launcher without the MR
framework.
However it should be possible to run HoS from Oozie via Spark in Oozie 4.3.0 as
well. I've done it before in CDH5.12.1. Please post logs/config files/workflow
files onto the users@ list for guidance.
> Oozie runs Hive on Spark task inside a MR job
> ---------------------------------------------
>
> Key: OOZIE-3069
> URL: https://issues.apache.org/jira/browse/OOZIE-3069
> Project: Oozie
> Issue Type: Improvement
> Reporter: Mobin Ranjbar
>
> Hi there,
> If I am wrong, correct me. Oozie does run all workflows inside a MapReduce
> job. Even if you want to run Spark job, it will run by MapReduce job. In
> HiveOnSpark situation(with hive.exection.engine=spark and spark.master=yarn
> in job.properties), that job will be failed with no reason. Based on my
> logic, there is no reason to run spark job inside a MapReduce job.
> What do you think? It may be an issue I think.
> Best,
> Mobin
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
