[jira] [Commented] (FALCON-349) Support "MapReduce" workflow in process entity

["Jean-Baptiste Onofré (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/FALCON-349?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14144698#comment-14144698 ] 

Jean-Baptiste Onofré commented on FALCON-349:
---------------------------------------------

What I have in mind as a first step: add some specific attributes/elements in the process entity XML (like mapper/reducer class, the location of the job jar). Falcon can create the oozie workflow.xml for the user using these metadata.

> Support "MapReduce" workflow in process entity
> ----------------------------------------------
>
>                 Key: FALCON-349
>                 URL: https://issues.apache.org/jira/browse/FALCON-349
>             Project: Falcon
>          Issue Type: Wish
>          Components: process
>            Reporter: Jean-Baptiste Onofré
>            Assignee: Jean-Baptiste Onofré
>
> Currently, a process entity supports the following workflow:
> - oozie
> - pig
> - hive
> If an user has a "pure" MapReduce job, the only way to use it in Falcon is via the oozie workflow. It means he has to create the workflow.xml describing the nodes (mapred.mapper.class, etc elements).
> So, it may look like a overhead for the user who just wants to "schedule" the job/process.
> I would propose to create a "mapreduce" workflow, taken directly a MapReduce jar from the filesystem, and behind the hood, create a simple workflow.xml and schedule it in oozie.
> Thoughts ?



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)