[ 
https://issues.apache.org/jira/browse/YARN-3547?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14526210#comment-14526210
 ] 

Xianyin Xin commented on YARN-3547:
-----------------------------------

Thanks [~leftnoteasy]! 
Attached a new patch using Set. When i review the patch, I found patch-001 is 
ill-considered, because unlike CS, Fair uses list to organize the apps at the 
very beginning, thus the newly added sorted set may become un-sorted because of 
the "update resource usage" or "update fair share" operations. We have two 
choices, the first is we continue to use the solution in patch-001, but fix it 
with considerations that we reorder the set when we update resource usage or 
fair shares just like in CS; the second is like patch-003, before we do assign, 
we scan the runningApps, and pick out the apps with real request to construct a 
set. Patch-003 is more efficient than patch-002, and the change of code is less 
than patch-001 (if fixed). What do you think, [~leftnoteasy]?
Change the name "pendingApps" to "pendingForResourceApps" because "pending" is 
ambiguous. Maybe we need a better name.

> FairScheduler: Apps that have no resource demand should not participate 
> scheduling
> ----------------------------------------------------------------------------------
>
>                 Key: YARN-3547
>                 URL: https://issues.apache.org/jira/browse/YARN-3547
>             Project: Hadoop YARN
>          Issue Type: Improvement
>          Components: fairscheduler
>            Reporter: Xianyin Xin
>            Assignee: Xianyin Xin
>         Attachments: YARN-3547.001.patch, YARN-3547.002.patch
>
>
> At present, all of the 'running' apps participate the scheduling process, 
> however, most of them may have no resource demand on a production cluster, as 
> the app's status is running other than waiting for resource at the most of 
> the app's lifetime. It's not a wise way we sort all the 'running' apps and 
> try to fulfill them, especially on a large-scale cluster which has heavy 
> scheduling load. 



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

Reply via email to