[
https://issues.apache.org/jira/browse/YARN-5696?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15544076#comment-15544076
]
kyungwan nam commented on YARN-5696:
------------------------------------
for example, there are input directories /data/input/0, /data/input/1 in HDFS.
there’s no way to assign an input directory per container.
/data/input/0 is assigned to container1
/data/input/1 is assigned to container2
It will be achieved by assigning “container_tag" per container.
It seems like mapper in MapReduce. but, we can use existing program without
changing.
> add container_tag to Distributedshell's env
> -------------------------------------------
>
> Key: YARN-5696
> URL: https://issues.apache.org/jira/browse/YARN-5696
> Project: Hadoop YARN
> Issue Type: Improvement
> Components: applications/distributed-shell
> Reporter: kyungwan nam
> Assignee: kyungwan nam
> Priority: Minor
> Attachments: YARN-5696-branch-2.7.3.001.patch
>
>
> a number of containers can be allocated by "num_containers" option.
> but, there is no way to assign different input to each container.
> each container has a unique id between 0 and (num_containers - 1).
> so, user can assign different input (or role) per container.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
