[
https://issues.apache.org/jira/browse/FLINK-19378?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17203182#comment-17203182
]
Robert Metzger commented on FLINK-19378:
----------------------------------------
Thank you for your responses. I personally think that from a user's
perspective, it would be very nice to print an error message ("can not load
hadoop dependencies to connect to running YARN Session"), instead of failing
with a "Connection refused". But I understand how this is difficult to
implement.
Given that this is just a minor usability improvement (that would probably
require quite some changes to make it nice), I'm closing this ticket for now.
Please comment on it if you've experienced this as well as a poor usability
experience, so that we can re-open it in the future, if necessary.
> Running "./bin/flink run" without HADOOP_CLASSPATH to submit job to running
> YARN session fails
> ----------------------------------------------------------------------------------------------
>
> Key: FLINK-19378
> URL: https://issues.apache.org/jira/browse/FLINK-19378
> Project: Flink
> Issue Type: Bug
> Components: Command Line Client, Deployment / YARN
> Affects Versions: 1.11.2
> Reporter: Robert Metzger
> Priority: Major
>
> Steps to reproduce:
> Shell A:
> 1. export HADOOP_CLASSPATH
> 2. start YARN session
> Shell B
> (unset HADOOP_CLASSPATH if it is set automatically)
> 3. start a Flink job using {{./bin/flink run}}.
> It will fail with {{Connection refused: localhost/127.0.0.1:8081}}.
> Expected behavior: connect to YARN cluster using YARN properties file.
> Workaround: set HADOOP_CLASSPATH, then the client will find the properties
> file.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
