[ https://issues.apache.org/jira/browse/YARN-10571?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17295123#comment-17295123 ]
Andras Gyori commented on YARN-10571: ------------------------------------- This refactor intends to address two issues with the current approach of CapacityScheduler: * There are two auto queue creation logic (AutoCreatedLeafQueues/Dynamic queues), which is confusing. With this refactor, all the auto queue creation logic is decoupled from CapacityScheduler, and could be invoked from a single point (CSQueueManager#createQueue). This will not solve the entire issue in its entirety (it is much more complicated), but in my opinion it is a good step towards a more comprehensible system. * The mapping rule engine contains a lot of queue validation logic as well, which is often duplicated by the auto queue handling logic. It will be important to have a centralised queue management logic that is used by both system. > Refactor dynamic queue handling logic > ------------------------------------- > > Key: YARN-10571 > URL: https://issues.apache.org/jira/browse/YARN-10571 > Project: Hadoop YARN > Issue Type: Sub-task > Reporter: Andras Gyori > Assignee: Andras Gyori > Priority: Minor > > As per YARN-10506 we have introduced an other mode for auto queue creation > and a new class, which handles it. We should move the old, managed queue > related logic to CSAutoQueueHandler as well, and do additional cleanup > regarding queue management. -- This message was sent by Atlassian Jira (v8.3.4#803005) --------------------------------------------------------------------- To unsubscribe, e-mail: yarn-issues-unsubscr...@hadoop.apache.org For additional commands, e-mail: yarn-issues-h...@hadoop.apache.org