[ https://issues.apache.org/jira/browse/YARN-4274?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15221871#comment-15221871 ]
Daniel Templeton commented on YARN-4274: ---------------------------------------- It looks to me like YARN-4288 should also have solved this JIRA. [~djp], can you confirm? > NodeStatusUpdaterImpl should register to RM again after a non-fatal exception > happen before > ------------------------------------------------------------------------------------------- > > Key: YARN-4274 > URL: https://issues.apache.org/jira/browse/YARN-4274 > Project: Hadoop YARN > Issue Type: Improvement > Reporter: Junping Du > Assignee: Junping Du > > From YARN-3896, an non-fatal exception like response ID mismatch between NM > and RM (due to a race condition) will cause NM stop working. I think we > should make it more robust to tolerant a few failure in registering to RM > with retry a few times. -- This message was sent by Atlassian JIRA (v6.3.4#6332)