[ https://issues.apache.org/jira/browse/YARN-4198?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Alexey Tumanov updated YARN-4198: --------------------------------- Attachment: YARN-4198-v1.patch Please find attached the patch that relaxes some of the synchronized methods in YARN RM. Most of the synchronization relaxation is in handling child queues. This code was tested on a large cluster in September 2015 along with the rest of the code written for node label aware reservations (YARN-4193). In this patch, we isolate just the synchronization changes we've made. We're in the process or re-testing this isolated delta and will update this JIRA when the testing is complete. Please treat this post as a request for code review and comments. Thanks, Alexey > CapacityScheduler locking / synchronization improvements > -------------------------------------------------------- > > Key: YARN-4198 > URL: https://issues.apache.org/jira/browse/YARN-4198 > Project: Hadoop YARN > Issue Type: Sub-task > Reporter: Carlo Curino > Assignee: Alexey Tumanov > Attachments: YARN-4198-v1.patch > > > In the context of YARN-4193 (which stresses the RM/CS performance) we found > several performance problems with in the locking/synchronization of the > CapacityScheduler, as well as inconsistencies that do not normally surface > (incorrect locking-order of queues protected by CS locks etc). This JIRA > proposes several refactoring that improve this. -- This message was sent by Atlassian JIRA (v6.3.4#6332)