[
https://issues.apache.org/jira/browse/FLINK-16018?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17036957#comment-17036957
]
Stephan Ewen commented on FLINK-16018:
--------------------------------------
I don't think this is a misalignment of timeouts. Indeed, when an Akka ask
timeout propagates to the user, it means we are not handling failures correctly
and simply report back whatever happened in the RPC system.
In this specific issue, we also have the problem of blocking/synchronous
ExecutionGraph creation. Because some amount of synchronous initialization will
most likely always be there, we would need some better way to handle responses
to the client. Maybe an additional JobStatus (like INITIALIZING).
> Improve error reporting when submitting batch job (instead of
> AskTimeoutException)
> ----------------------------------------------------------------------------------
>
> Key: FLINK-16018
> URL: https://issues.apache.org/jira/browse/FLINK-16018
> Project: Flink
> Issue Type: Improvement
> Components: Runtime / Coordination
> Affects Versions: 1.11.0
> Reporter: Robert Metzger
> Priority: Major
>
> While debugging the {{Shaded Hadoop S3A end-to-end test (minio)}} pre-commit
> test, I noticed that the JobSubmission is not producing very helpful error
> messages.
> Environment:
> - A simple batch wordcount job
> - a unavailable minio s3 filesystem service
> What happens from a user's perspective:
> - The job submission fails after 10 seconds with a AskTimeoutException:
> {code}
> 2020-02-07T11:38:27.1189393Z akka.pattern.AskTimeoutException: Ask timed out
> on [Actor[akka://flink/user/dispatcher#-939201095]] after [10000 ms]. Message
> of type [org.apache.flink.runtime.rpc.messages.LocalFencedMessage]. A typical
> reason for `AskTimeoutException` is that the recipient actor didn't send a
> reply.
> 2020-02-07T11:38:27.1189538Z at
> akka.pattern.PromiseActorRef$$anonfun$2.apply(AskSupport.scala:635)
> 2020-02-07T11:38:27.1189616Z at
> akka.pattern.PromiseActorRef$$anonfun$2.apply(AskSupport.scala:635)
> 2020-02-07T11:38:27.1189713Z at
> akka.pattern.PromiseActorRef$$anonfun$1.apply$mcV$sp(AskSupport.scala:648)
> 2020-02-07T11:38:27.1189789Z at
> akka.actor.Scheduler$$anon$4.run(Scheduler.scala:205)
> 2020-02-07T11:38:27.1189883Z at
> scala.concurrent.Future$InternalCallbackExecutor$.unbatchedExecute(Future.scala:601)
> 2020-02-07T11:38:27.1189973Z at
> scala.concurrent.BatchingExecutor$class.execute(BatchingExecutor.scala:109)
> 2020-02-07T11:38:27.1190067Z at
> scala.concurrent.Future$InternalCallbackExecutor$.execute(Future.scala:599)
> 2020-02-07T11:38:27.1190159Z at
> akka.actor.LightArrayRevolverScheduler$TaskHolder.executeTask(LightArrayRevolverScheduler.scala:328)
> 2020-02-07T11:38:27.1190267Z at
> akka.actor.LightArrayRevolverScheduler$$anon$4.executeBucket$1(LightArrayRevolverScheduler.scala:279)
> 2020-02-07T11:38:27.1190358Z at
> akka.actor.LightArrayRevolverScheduler$$anon$4.nextTick(LightArrayRevolverScheduler.scala:283)
> 2020-02-07T11:38:27.1190465Z at
> akka.actor.LightArrayRevolverScheduler$$anon$4.run(LightArrayRevolverScheduler.scala:235)
> 2020-02-07T11:38:27.1190540Z at java.lang.Thread.run(Thread.java:748)
> {code}
> What a user would expect:
> - An error message indicating why the job submission failed.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
