[ https://issues.apache.org/jira/browse/FLINK-13426?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16895684#comment-16895684 ]
Guowei Ma commented on FLINK-13426: ----------------------------------- TaskExecutor should use the taskExecutorRegistrationId after it successfully sendSlotReport the first time. Or it might use the old one in the heartbeat with ResourceManager. Since that TaskExecutor should monitor the ResourceManager at slotReportResponseFuture complete callback. > TaskExecutor uses the wrong Registrationid in the heartbeat with RM. > -------------------------------------------------------------------- > > Key: FLINK-13426 > URL: https://issues.apache.org/jira/browse/FLINK-13426 > Project: Flink > Issue Type: Bug > Components: Runtime / Coordination > Affects Versions: 1.8.1, 1.9.0 > Reporter: Guowei Ma > Priority: Minor > Attachments: image-2019-07-25-17-57-03-537.png > > > 1. First-time TaskExecutor register to rm successfully. If it fails to send > SlotReport to SlotMaanger, TaskExecutor will reconnect to RM again. However, > TaskExecutor still uses the old registration id in the > EstablishedResourceManagerConnection. > 2. Second-time TaskExecutor registers to rm successfully and gets a new > registration id. > 3. First-round and second-round has a race condition. Since that the task > executor maybe use the old registration id in heartbeat with rm. > > !image-2019-07-25-17-57-03-537.png! -- This message was sent by Atlassian JIRA (v7.6.14#76016)