You are viewing a plain text version of this content. The canonical link for it is here.
Posted to mapreduce-issues@hadoop.apache.org by "Gera Shegalov (JIRA)" <ji...@apache.org> on 2016/03/18 21:08:33 UTC

[jira] [Commented] (MAPREDUCE-6653) JobSubmitter with unmanaged MRApplicationMaster

    [ https://issues.apache.org/jira/browse/MAPREDUCE-6653?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15202061#comment-15202061 ] 

Gera Shegalov commented on MAPREDUCE-6653:
------------------------------------------

In terms of user API we can introduce a new value for mapreduce.framework.name: yarn.local or yarn.unmanaged.

> JobSubmitter with unmanaged MRApplicationMaster  
> -------------------------------------------------
>
>                 Key: MAPREDUCE-6653
>                 URL: https://issues.apache.org/jira/browse/MAPREDUCE-6653
>             Project: Hadoop Map/Reduce
>          Issue Type: New Feature
>          Components: applicationmaster, client
>            Reporter: Gera Shegalov
>
> Long living DAG applications (pig, hive, etc) may launch many mapreduce jobs. Since we already have to allocate resources for these clients, we could reduce the latency of monitoring AM's, the overall number of network RPC's, the number of containers running on YARN cluster by embedding the MRAppMaster with the submitter. 



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