[ https://issues.apache.org/jira/browse/YARN-9809?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17110356#comment-17110356 ]
Craig Condit commented on YARN-9809: ------------------------------------ Since health check scripts are by nature different for every deployment, it seems that neither the current behavior nor what is proposed makes sense in all cases. However, I believe there can be some middle ground. I propose we keep the existing behavior as default to avoid causing pain for existing users, but allow a configuration to opt-in to a single synchronous execution of a health check on startup before node check-in (controlled via a new *{{yarn.nodemanager.health-checker.preflight.enabled}}* boolean configuration). It may also be desirable to kill the NM upon repeated failures of this preflight check. We could add a new config *{{yarn.nodemanager.health-checker.preflight.retries}}* to control the number of retries before aborting the NM (or -1 for infinite). > NMs should supply a health status when registering with RM > ---------------------------------------------------------- > > Key: YARN-9809 > URL: https://issues.apache.org/jira/browse/YARN-9809 > Project: Hadoop YARN > Issue Type: Bug > Reporter: Eric Badger > Assignee: Eric Badger > Priority: Major > > Currently if the NM registers with the RM and it is unhealthy, it can be > scheduled many containers before the first heartbeat. After the first > heartbeat, the RM will mark the NM as unhealthy and kill all of the > containers. -- 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