kl0u commented on a change in pull request #13583:
URL: https://github.com/apache/flink/pull/13583#discussion_r503773953
##########
File path:
flink-clients/src/main/java/org/apache/flink/client/deployment/application/ApplicationDispatcherGatewayServiceFactory.java
##########
@@ -92,6 +91,7 @@ public ApplicationDispatcherGatewayServiceFactory(
throw new FlinkRuntimeException("Could not create the
Dispatcher rpc endpoint.", e);
}
+ bootstrap.setErrorHandler(exception ->
dispatcher.getShutDownFuture().completeExceptionally(exception));
Review comment:
Yes @aljoscha. The solution I gave was to set the error handler using a
setter, but it is not elegant as a construct. It is fragile, as there must be a
default handler that does nothing, and the user has to keep in mind to set it
when needed.
@tillrohrmann Errors that lead to job execution termination are fatal and
continuing to work does not really make sense. The semantics are that the
cluster is created for the application only, so if any job in the application
fails irrecoverably (even with status UNKNOWN), it should kill also the cluster
(potentially without deleting HA data).
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
