[
https://issues.apache.org/jira/browse/FLINK-2736?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14904134#comment-14904134
]
Till Rohrmann commented on FLINK-2736:
--------------------------------------
Ok, [~rmetzger] just told me that the {{LeaderNotAvailableException}} comes
from constantly polling the Kafka state to see whether a leader has been
elected. The problem seems to originate from something else. I also encountered
failing KafkaITCases recently:
https://s3.amazonaws.com/archive.travis-ci.org/jobs/81574197/log.txt
https://s3.amazonaws.com/archive.travis-ci.org/jobs/81551553/log.txt
https://s3.amazonaws.com/archive.travis-ci.org/jobs/81551623/log.txt
The second log is especially interesting because there we couldn't cancel a
vertex because of a {{StackOverflowError}}.
> Failing KafkaConsumerTestBase.runCancelingOnEmptyInputTest
> ----------------------------------------------------------
>
> Key: FLINK-2736
> URL: https://issues.apache.org/jira/browse/FLINK-2736
> Project: Flink
> Issue Type: Bug
> Components: Kafka Connector
> Reporter: Ufuk Celebi
> Assignee: Robert Metzger
> Priority: Minor
> Labels: test-stability
>
> https://s3.amazonaws.com/archive.travis-ci.org/jobs/81591058/log.txt
> https://flink.a.o.uce.east.s3.amazonaws.com/travis-artifacts/uce/flink/673/673.1.tar.gz
> ---
> {{KafkaConsumer}} fails with {{LeaderNotAvailableException}} in the
> constructor. The job fails to submit and the test concurrently cancels it.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
