[
https://issues.apache.org/jira/browse/HIVE-12623?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15065022#comment-15065022
]
Gunther Hagleitner commented on HIVE-12623:
-------------------------------------------
Change and test LGTM +1. Submitted the patch to HiveQA. This add a new config:
hive.llap.task.scheduler.locality.delay. The value is used once at AM startup,
which means it's best to change it before starting HS2.
> Add an option to force allocation of fragments on requested nodes
> -----------------------------------------------------------------
>
> Key: HIVE-12623
> URL: https://issues.apache.org/jira/browse/HIVE-12623
> Project: Hive
> Issue Type: Improvement
> Components: llap
> Reporter: Siddharth Seth
> Assignee: Siddharth Seth
> Attachments: HIVE-12623.1.txt, HIVE-12623.1.wip.txt
>
>
> Currently, fragments are sent to random nodes if the requested node does not
> have capacity. In certain situations there's more to be gained by sending the
> fragments to the requested node only.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)