[ https://issues.apache.org/jira/browse/YARN-5392?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15386628#comment-15386628 ]
Wangda Tan commented on YARN-5392: ---------------------------------- [~asuresh], I can understand that YARN-4888 needs a class to combine Priority/AllocationRequestId like: {code} Map<NodeId, Map<SchedulingKey, Map<ContainerId, SchedContainerChangeRequest>>> {code} Instead of adding one more hierarchy such as: {code} Map<NodeId, Map<Priority, Map<AllocationRequestId, Map<ContainerId, SchedContainerChangeRequest>>>> {code} But I still not quite understand why we need to change all existing logic which uses Priority to use SchedulingKey, could you share your thoughts about this? > Replace use of Priority in the Scheduling infrastructure with an opaque > ShedulerKey > ----------------------------------------------------------------------------------- > > Key: YARN-5392 > URL: https://issues.apache.org/jira/browse/YARN-5392 > Project: Hadoop YARN > Issue Type: Sub-task > Reporter: Arun Suresh > Assignee: Arun Suresh > Attachments: YARN-5392.001.patch, YARN-5392.002.patch, > YARN-5392.003.patch, YARN-5392.004.patch > > > Based on discussions in YARN-4888, this jira proposes to replace the use of > {{Priority}} in the Scheduler infrastructure (Scheduler, Queues, SchedulerApp > / Node etc.) with a more opaque and extensible {{SchedulerKey}}. > Note: Even though {{SchedulerKey}} will be used by the internal scheduling > infrastructure, It will not be exposed to the Client or the AM. The > SchdulerKey is meant to be an internal construct that is derived from > attributes of the ResourceRequest / ApplicationSubmissionContext / Scheduler > Configuration etc. -- This message was sent by Atlassian JIRA (v6.3.4#6332) --------------------------------------------------------------------- To unsubscribe, e-mail: yarn-issues-unsubscr...@hadoop.apache.org For additional commands, e-mail: yarn-issues-h...@hadoop.apache.org