[ 
https://issues.apache.org/jira/browse/YARN-451?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13795266#comment-13795266
 ] 

Jason Lowe commented on YARN-451:
---------------------------------

bq. current allocation can be seen from the scheduler page.

I took a look at the scheduler page, and all I see for current allocation is 
per-user-per-queue and not per app.  Where are you seeing the current 
assignment for each app on the scheduler page?

As for the instance you recently encountered, showing the current ask would 
have quickly isolated the issue as all 30K maps would have been asked for once 
the app launched.

My main concern with a current-plus-estimated-future approach is that it's 
optional for AMs to implement and requires an API change.  I see showing the 
current and/or ask as more robust across different app frameworks (doesn't 
require AMs to implement anything), easier to implement, and should solve most 
of the problems with identifying where the bottlenecks currently are in 
scheduling containers.  Doing so doesn't preclude adding a total estimate 
metric at some point.

Quick question on the estimate -- is it a calculation of the total app weight 
at the start of the app or do the values decrease as containers are granted?  
The former is useful as a gauge of how big an app is/was overall, while the 
latter is more useful for identifying upcoming demands if the application has 
been running for some time.

> Add more metrics to RM page
> ---------------------------
>
>                 Key: YARN-451
>                 URL: https://issues.apache.org/jira/browse/YARN-451
>             Project: Hadoop YARN
>          Issue Type: Improvement
>          Components: resourcemanager
>    Affects Versions: 2.0.3-alpha
>            Reporter: Lohit Vijayarenu
>            Assignee: Sangjin Lee
>            Priority: Blocker
>         Attachments: in_progress_2x.png, yarn-451-trunk-20130916.1.patch
>
>
> ResourceManager webUI shows list of RUNNING applications, but it does not 
> tell which applications are requesting more resource compared to others. With 
> cluster running hundreds of applications at once it would be useful to have 
> some kind of metric to show high-resource usage applications vs low-resource 
> usage ones. At the minimum showing number of containers is good option.



--
This message was sent by Atlassian JIRA
(v6.1#6144)

Reply via email to