[ 
https://issues.apache.org/jira/browse/YARN-7612?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16287790#comment-16287790
 ] 

Arun Suresh commented on YARN-7612:
-----------------------------------

Many thanks for the review [~pgaref]

bq. Currently PlacementConstraintsManager interface considers constraints as 
part of appIDs but we also need to support *cluster wide constraints - so I 
would add another more generic setter and getter as well
Agreed - but I think we might need an extra API that needs to be invoked by the 
cluster operator etc. I was thinking, as a first cut, lets go with intra-app 
placements. In any case, inter app placement can still be accomplished by one 
app referring to the other apps allocation tags. Also, we can flesh up the 
PlacementConstraintManager further in YARN-6596.

w.r.t the PlacementAlgorithm
bq. ..One way to solve this would be having the algorithm implementation to 
keep an extra data structure with the placed tags, another would be to extend 
the TagManager to keep a temporary mapping. 
Yup - That was my intention actually (also I think I commented the same on the 
YARN-7522). When you tackle YARN-7613, feel free to add any missing api to the 
TagsManager and to use the tagsManager for intermediate storage. You should 
also consider storing internal state in the final Algorithm implementation.

w.r.t the PlacementAlgorithm abstract class not in the spi. : So I have a 
generic Algorithm interface in the spi package (which can be used for other 
algorithms pertaining to GlobalScheuling). The placementAlgorithm would be 
specific to our processor based implementation of placement constraints.

w.r.t the enum names and the other nits: Agreed - will post a subsequent patch 
with the fix.


> Add Placement Processor and planner framework
> ---------------------------------------------
>
>                 Key: YARN-7612
>                 URL: https://issues.apache.org/jira/browse/YARN-7612
>             Project: Hadoop YARN
>          Issue Type: Sub-task
>            Reporter: Arun Suresh
>            Assignee: Arun Suresh
>         Attachments: YARN-7612-YARN-6592.001.patch, 
> YARN-7612-YARN-6592.002.patch, YARN-7612-YARN-6592.003.patch, 
> YARN-7612-YARN-6592.004.patch, YARN-7612-YARN-6592.005.patch, 
> YARN-7612-v2.wip.patch, YARN-7612.wip.patch
>
>
> This introduces a Placement Processor and a Planning algorithm framework to 
> handle placement constraints and scheduling requests from an app and places 
> them on nodes.
> The actual planning algorithm(s) will be handled in a YARN-7613.



--
This message was sent by Atlassian JIRA
(v6.4.14#64029)

---------------------------------------------------------------------
To unsubscribe, e-mail: yarn-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: yarn-issues-h...@hadoop.apache.org

Reply via email to