Maximilian Michels created FLINK-3863:
-----------------------------------------

             Summary: Yarn Cluster shutdown may fail if leader changed recently
                 Key: FLINK-3863
                 URL: https://issues.apache.org/jira/browse/FLINK-3863
             Project: Flink
          Issue Type: Bug
          Components: YARN Client
    Affects Versions: 1.0.2, 1.1.0
            Reporter: Maximilian Michels
            Assignee: Maximilian Michels
            Priority: Minor
             Fix For: 1.1.0


The {{ApplicationClient}} sets {{yarnJobManager}} to {{None}} until it has 
connected to a newly elected JobManager. A shutdown message to the application 
master is discarded while the ApplicationClient tries to reconnect. The 
ApplicationClient should retry to shutdown the cluster when it is connected to 
the new leader. It may also time out (which currently is always the case).



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Reply via email to