[ https://issues.apache.org/jira/browse/YARN-1879?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13954943#comment-13954943 ]
Tsuyoshi OZAWA commented on YARN-1879: -------------------------------------- {quote} I suppose this is what MR-AM does today? we cannot assume each AM does the same. {quote} Jian, Thank you for the point. I supposed DistributedShell's application master in the sentence. I checked that MRAppMaster stops because it doesn't retry when Exception occurs at server-side. Therefore, as you mentioned, we cannot assume each AM doesn't the same. We should make them "AtMostOnce" with RetryCache-like mechanism. I'll create a patch based on the discussion. > Mark Idempotent/AtMostOnce annotations to ApplicationMasterProtocol > ------------------------------------------------------------------- > > Key: YARN-1879 > URL: https://issues.apache.org/jira/browse/YARN-1879 > Project: Hadoop YARN > Issue Type: Sub-task > Components: resourcemanager > Reporter: Jian He > Assignee: Tsuyoshi OZAWA > Priority: Critical > Attachments: YARN-1879.1.patch > > -- This message was sent by Atlassian JIRA (v6.2#6252)