[ https://issues.apache.org/jira/browse/YARN-4930?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15229060#comment-15229060 ]
Jason Lowe commented on YARN-4930: ---------------------------------- Currently admins can accomplish the task via a kill -9 of the nodemanager process, but it would be better to encapsulate the specific mechanism into a CLI command. > Convenience command to perform a work-preserving stop of the nodemanager > ------------------------------------------------------------------------ > > Key: YARN-4930 > URL: https://issues.apache.org/jira/browse/YARN-4930 > Project: Hadoop YARN > Issue Type: Improvement > Reporter: Jason Lowe > Priority: Minor > > When the nodemanager is not under supervision > (yarn.nodemanager.recovery.supervised=false) there may be cases where an > admin wants to perform a work-preserving maintenance of the node but the > nodemanager stop CLI command will cause the NM to cleanup containers. There > should be a CLI command that shuts down the NM while preserving local > filesystem and container state to allow a subsequent nodemanager start to > recover containers. -- This message was sent by Atlassian JIRA (v6.3.4#6332)