You are viewing a plain text version of this content. The canonical link for it is here.

Posted to issues@flink.apache.org by "Till Rohrmann (JIRA)" <ji...@apache.org> on 2018/09/13 16:08:00 UTC

[jira] [Created] (FLINK-10344) Rethink SubmittedJobGraphListener

Till Rohrmann created FLINK-10344:
-------------------------------------

             Summary: Rethink SubmittedJobGraphListener
                 Key: FLINK-10344
                 URL: https://issues.apache.org/jira/browse/FLINK-10344
             Project: Flink
          Issue Type: Sub-task
          Components: Distributed Coordination
    Affects Versions: 1.6.0, 1.5.3, 1.7.0
            Reporter: Till Rohrmann
             Fix For: 1.7.0


The {{SubmittedJobGraphListener}} in {{ZooKeeperSubmittedJobGraphStore}} can return false positives. This is obviously problematic, because it causes the subsequent recovery operation to fail. Ideally we would not require the {{SubmittedJobGraphListener}}. One could, for example, periodically check from the main thread whether there are new jobs. That way we would know which jobs are currently running and which are being cleaned up. 

Alternatively it is necessary to tolerate false positives :-(



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)