[ https://issues.apache.org/jira/browse/YARN-1041?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13732207#comment-13732207 ]
Steve Loughran commented on YARN-1041: -------------------------------------- This is not something that MR could (currently) handle unless it was persisting its state somewhere, but for use cases like running HBase on YARN, we'd want to get a list of the containers running the AMs, so that the AM could bond to them again. To avoid having to persist any run-time state in an AM, the AM should be able to set/query some limited metadata to every container -such as adding a container role string/int that could differentiate the purpose of an existing container. > RM to bind and notify a restarted AM of existing containers > ----------------------------------------------------------- > > Key: YARN-1041 > URL: https://issues.apache.org/jira/browse/YARN-1041 > Project: Hadoop YARN > Issue Type: Sub-task > Components: resourcemanager > Affects Versions: 3.0.0 > Reporter: Steve Loughran > > For long lived containers we don't want the AM to be a SPOF. > When the RM restarts a (failed) AM, it should be given the list of containers > it had already been allocated. the AM should then be able to contact the NMs > to get details on them. NMs would also need to do any binding of the > containers needed to handle a moved/restarted AM. -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira