[ https://issues.apache.org/jira/browse/YARN-1521?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13945905#comment-13945905 ]
Xuan Gong commented on YARN-1521: --------------------------------- Thanks for the review. bq. The bulk of test-specific hack in MiniYarnCluster can be moved to TestHAProtocol, as MiniYarnCluster is commonly used by others. Yes, you are right. Remove those changes from MiniYarnCluster to TestHAProtocol bq. Explicitly assert the app exists in RMContext after the 2nd RM comes Active./ Explicitly assert the App does not exist in RMStateStore or not. DONE bq. this doesn’t seem correct, user is possible to get multiple delegation tokens. Given this change, user can only get one token Remove all changes from ClientRMService#getDelegationToken(). Just let the method re-entry if failover happens. [~jianhe] Please correct me if I am wrong. The getDelegationToken() is used to get a new Token, but is not saved in zookeeper yet. So, even if failover happens during getDelegationToken call, we can not get previous generated Token back, so just let the method re-generate a new token, and it should be fine. > Mark appropriate protocol methods with the idempotent annotation or > AtMostOnce annotation > ----------------------------------------------------------------------------------------- > > Key: YARN-1521 > URL: https://issues.apache.org/jira/browse/YARN-1521 > Project: Hadoop YARN > Issue Type: Sub-task > Reporter: Xuan Gong > Assignee: Xuan Gong > Attachments: YARN-1521.0.patch, YARN-1521.1.patch > > > After YARN-1028, we add the automatically failover into RMProxy. This JIRA is > to identify whether we need to add idempotent annotation and which methods > can be marked as idempotent. -- This message was sent by Atlassian JIRA (v6.2#6252)