[jira] [Commented] (FLINK-9371) High Availability JobManager Registration Failure

["Mike Urbach (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/FLINK-9371?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16483003#comment-16483003 ] 

Mike Urbach commented on FLINK-9371:
------------------------------------

I have encountered this too, specifically the error stating "because there is currently no valid leader id known". I have not encountered "because the expected leader session ID ... did not equal the received leader session ID". Does anyone have some insight on this issue? I have not dug into it at all.

We run our cluster on Kubernetes, and I have worked around this by deleting all of our resources, cleaning up the state in Zookeeper, and re-creating everything from scratch. If it is helpful, I can provide more info about our setup.

> High Availability JobManager Registration Failure
> -------------------------------------------------
>
>                 Key: FLINK-9371
>                 URL: https://issues.apache.org/jira/browse/FLINK-9371
>             Project: Flink
>          Issue Type: Bug
>          Components: Core
>    Affects Versions: 1.4.2
>            Reporter: Jason Kania
>            Priority: Major
>
> The following error is happening intermittently on an 3 node cluster with 2 Job Managers configured in HA mode. When this happens, the two JobManager instances are associated with one another.
> 2018-05-15 19:00:06,400 INFO  org.apache.flink.runtime.clusterframework.standalone.StandaloneResourceManager  - Trying to associate with JobManager leader akka.tcp://flink@aaa-1:50000/user/jobmanager
> 2018-05-15 19:00:06,404 WARN  org.apache.flink.runtime.jobmanager.JobManager                - Discard message LeaderSessionMessage(0bbe70c4-2642-4a08-912f-6cc09646281f,RegisterResourceManager akka://flink/user/resourcemanager-d6567c5d-85f4-4b18-8eac-cf9725d076a5) because there is currently no valid leader id known.
> 2018-05-15 19:00:16,418 ERROR org.apache.flink.runtime.clusterframework.standalone.StandaloneResourceManager  - Resource manager could not register at JobManager
> akka.pattern.AskTimeoutException: Ask timed out on [ActorSelection[Anchor(akka://flink/), Path(/user/jobmanager)]] after [10000 ms]. Sender[null] sent message of type "org.apache.flink.runtime.messages.JobManagerMessages$LeaderSessionMessage".
>         at akka.pattern.PromiseActorRef$$anonfun$1.apply$mcV$sp(AskSupport.scala:604)
>         at akka.actor.Scheduler$$anon$4.run(Scheduler.scala:126)
>         at scala.concurrent.Future$InternalCallbackExecutor$.unbatchedExecute(Future.scala:601)
>         at scala.concurrent.BatchingExecutor$class.execute(BatchingExecutor.scala:109)
>         at scala.concurrent.Future$InternalCallbackExecutor$.execute(Future.scala:599)
>         at akka.actor.LightArrayRevolverScheduler$TaskHolder.executeTask(LightArrayRevolverScheduler.scala:329)
>         at akka.actor.LightArrayRevolverScheduler$$anon$4.executeBucket$1(LightArrayRevolverScheduler.scala:280)
>         at akka.actor.LightArrayRevolverScheduler$$anon$4.nextTick(LightArrayRevolverScheduler.scala:284)
>         at akka.actor.LightArrayRevolverScheduler$$anon$4.run(LightArrayRevolverScheduler.scala:236)
>         at java.lang.Thread.run(Thread.java:748)
>  
> Sometimes the following type of log also comes out following the previous log:
> 2018-05-15 19:13:47,525 WARN  org.apache.flink.runtime.clusterframework.standalone.StandaloneResourceManager  - Discard message LeaderSessionMessage(5cab29b9-10d3-4b25-b934-f06b82be15b5,TriggerRegistrationAtJobManager akka.tcp://flink@aaa-1:50000/user/jobmanager) because the expected leader session ID 61075587-51da-4e58-ac4f-9ea118ccdde9 did not equal the received leader session ID 5cab29b9-10d3-4b25-b934-f06b82be15b5.



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