[
https://issues.apache.org/jira/browse/YARN-6596?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16299462#comment-16299462
]
Konstantinos Karanasos commented on YARN-6596:
----------------------------------------------
bq. Would it not be better if we just expose addConstraint(sourceTags,
constraint, appId) and getConstraint(sourceTags, appId) and let the
PlacementConstaintManager decide from the tags if it is app specific or global
and perform the appropriate operation ?
I like what you say about the getConstraint. Indeed we should probably not have
a getGlobalConstraint. We can have a getConstraint that if the appId is empty,
it returns only global constraints or something along these lines. Also, what I
had in mind to do was that even if you request for the constraint with a
specific sourceTag and appId, it should be merged with any global constraint
too. So if you have for a specific appId that you should not have more than 5
HBase containers/rack, while a global says no more than 3/rack, you should
merge these too (the global wins here as it is more restrictive). I am planning
to add some transformation rules to the constraints to handle this soon.
For the addConstraint, I think it is fine to distinguish between app specific
and global, as the global should be used only by the admin API. But again we
can have a single addConstraint and when the appId is empty, it means it is a
global constraint.
> Introduce Placement Constraint Manager module
> ---------------------------------------------
>
> Key: YARN-6596
> URL: https://issues.apache.org/jira/browse/YARN-6596
> Project: Hadoop YARN
> Issue Type: Sub-task
> Reporter: Konstantinos Karanasos
> Assignee: Konstantinos Karanasos
> Attachments: YARN-6596-YARN-6592.001.patch
>
>
> This RM module will be responsible for storing placement constraints,
> allocation tags, and node attributes.
> It will be used when determining the placement of SchedulingRequests with
> constraints.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
