[
https://issues.apache.org/jira/browse/YARN-7682?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16305626#comment-16305626
]
Arun Suresh commented on YARN-7682:
-----------------------------------
[~pgaref], thanks for the patch.
bq. should we promote affinity on Nodes where cMin is < 5 or just ensure cMax
is <= 10 ?
Yeah, I get what you mean. Lets keep things simple - and for the time being
allow users to only specify max-cardinality with affinity (I think our design
doc also only has max-cardinality + affinity expressions) in which case we
ignore the cMin - except for anti-affinity case where we ensure the value is 0.
Other comments on the patch:
* Instead of string literals "node" and "rack", use
PlacementConstraint.NODE_SCOPE / RACK_SCOPE. You have to make them public too I
guess.
* Lets have a test for affinity + max-cardinalty
* Nit: No need to have a javadoc for the private method - just put in a '//'
comment - if there are any specific assumptions you'd like to make.
Otherwise, the patch looks good to me.
> Expose canAssign method in the PlacementConstraintManager
> ---------------------------------------------------------
>
> Key: YARN-7682
> URL: https://issues.apache.org/jira/browse/YARN-7682
> Project: Hadoop YARN
> Issue Type: Sub-task
> Reporter: Arun Suresh
> Assignee: Panagiotis Garefalakis
> Attachments: YARN-7682-YARN-6592.001.patch, YARN-7682.wip.patch
>
>
> As per discussion in YARN-7613. Lets expose {{canAssign}} method in the
> PlacementConstraintManager that takes a sourceTags, applicationId,
> SchedulerNode and AllocationTagsManager and returns true if constraints are
> not violated by placing the container on the node.
> I prefer not passing in the SchedulingRequest, since it can have > 1
> numAllocations. We want this api to be called for single allocations.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
