[ 
https://issues.apache.org/jira/browse/YARN-2497?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14585253#comment-14585253
 ] 

Karthik Kambatla edited comment on YARN-2497 at 6/14/15 8:28 PM:
-----------------------------------------------------------------

Thanks for working on this, [~Naganarasimha]. Before implementing it, it would 
be nice to align on the approach. Could you put up a doc with your thoughts on 
the same? 

[~asuresh] and I have been discussing changes to FairScheduler and how we would 
implement node-labels if we were starting from scratch. Today, the schedulers 
react to node heartbeats. Even continuous/asychronous scheduling in FS/CS 
iterate through the nodes. On the other hand, if we consider a pre-ordered 
hierarchy of queues/apps, we could look at nodes that satisfy a series of 
constraints - locality, available capacity and node labels.




was (Author: kasha):
Thanks for working on this, [~Naganarasimha]. Before implementing it, it would 
be nice to align on the approach. Could you put up a doc with your thoughts on 
the same? 

[~asuresh] and I have been discussing changes to FairScheduler and how we would 
implement node-labels if we were starting from scratch. Today, the schedulers 
react to node heartbeats. Even continuous/asychronous scheduling in FS/CS 
iterate through the nodes. On the other hand, if we consider an pre-ordered 
hierarchy of queues/apps, we could look at nodes that satisfy a series of 
constraints - locality, available capacity and node labels.



> Changes for fair scheduler to support allocate resource respect labels
> ----------------------------------------------------------------------
>
>                 Key: YARN-2497
>                 URL: https://issues.apache.org/jira/browse/YARN-2497
>             Project: Hadoop YARN
>          Issue Type: Sub-task
>          Components: resourcemanager
>            Reporter: Wangda Tan
>            Assignee: Naganarasimha G R
>




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

Reply via email to