[
https://issues.apache.org/jira/browse/YARN-7419?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16236712#comment-16236712
]
Wangda Tan commented on YARN-7419:
----------------------------------
Thanks [~suma.shivaprasad] for working on this.
Some comments:
1) Why changes of getClock are needed, seems nobody is using them?
2) Following methods can reuse leaf queue's methods (just pass
<parent-queue-path>.AUTO_CREATED_LEAF_QUEUE_TEMPLATE_PREFIX as queue path).
{code}
getAutoCreatedLeafQueueTemplateMaxCapacity(String)
getAutoCreatedQueuesMaxChildQueuesLimit(String)
setAutoCreatedLeafQueueTemplateCapacity(String, float)
setAutoCreatedLeafQueueTemplateMaxApplicationsPerQueue(String, int)
setAutoCreatedLeafQueueTemplateMaxCapacity(String, float)
{code}
And this config: {{AUTO_CREATE_QUEUE_MAX_QUEUES}} should be part of parent
queue instead of leaf queue templete.
3) Could you create a AutoCreatedLeafQueueTemplate class which encapsulate all
leaf queue's properties and replace all getter/setters of leaf queue properties.
4) sumOfChildCapacities/sumOfChildAbsCapacities can be moved to
AbstractManagedParentQueue, correct?
5) ApplicationPlacementContext, why keep PlacementRule as one of the fields?
6) There're two {{rmApp.getApplicationQueuePlacementContext()}} calls inside
CapacityScheduler, one of them can be removed. (Which we can avoid access
RMApps map twice).
> Implement Auto Queue Creation with modifications to queue mapping flow
> ----------------------------------------------------------------------
>
> Key: YARN-7419
> URL: https://issues.apache.org/jira/browse/YARN-7419
> Project: Hadoop YARN
> Issue Type: Sub-task
> Components: capacity scheduler
> Reporter: Suma Shivaprasad
> Assignee: Suma Shivaprasad
> Priority: Major
> Attachments: YARN-7419.1.patch, YARN-7419.patch
>
>
> This involves changes to queue mapping flow to pass along context information
> for auto queue creation. Auto creation of queues will be part of Capacity
> Scheduler flow while attempting to resolve queues during application
> submission. The leaf queues which do not exist are auto created under parent
> queues which have been explicitly enabled for auto queue creation . In order
> to determine which parent queue to create the leaf queues under - parent
> queues need to be specified in queue mapping configuration
--
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
