[ https://issues.apache.org/jira/browse/FLINK-24113?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17459422#comment-17459422 ]
Robert Metzger commented on FLINK-24113: ---------------------------------------- Maybe this issue was introduced by https://issues.apache.org/jira/browse/FLINK-24156 ? It seems that the BlobServer.close() method is setting shutdownRequested = true, for the BlobServer.run() method to finish. However in my case, I'm stuck in {code:java} BlobServerConnection conn = new BlobServerConnection(NetUtils.acceptWithoutTimeout(serverSocket), this); {code} ... probably its not a good idea to "accept without timeout" in this case (the interrupt from the close() method is probably caught in the NetUtils.acceptWithoutTimeout() method). WDYT? > Introduce option in Application Mode to disable shutdown > -------------------------------------------------------- > > Key: FLINK-24113 > URL: https://issues.apache.org/jira/browse/FLINK-24113 > Project: Flink > Issue Type: Improvement > Components: Runtime / Coordination > Affects Versions: 1.15.0 > Reporter: Robert Metzger > Assignee: Nicolaus Weidner > Priority: Major > Labels: pull-request-available > Fix For: 1.15.0 > > Attachments: shutdown_issue.log > > > Currently a Flink JobManager started in Application Mode will shut down once > the job has completed. > When doing a "stop with savepoint" operation, we want to keep the JobManager > alive after the job has stopped to retrieve and persist the final savepoint > location. > Currently, Flink waits up to 5 minutes and then shuts down. > I'm proposing to introduce a new configuration flag "application mode > shutdown behavior": "keepalive" (naming things is hard ;) ) which will keep > the JobManager in ApplicationMode running until a REST call confirms that it > can shutdown. -- This message was sent by Atlassian Jira (v8.20.1#820001)