[
https://issues.apache.org/jira/browse/FLINK-20685?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17252648#comment-17252648
]
Yang Wang commented on FLINK-20685:
-----------------------------------
You just need to export {{HADOOP_CONF_DIR}} in the Flink client side before
submitting the job. But the root cause is not about this since the the Flink
job has been submitted successfully.
After checking the logs you attached, I find that the {{HADOOP_CONF_DIR}} is
not adding the {{CLASSPATH}} in {{launch_container.sh}}. I am not sure whether
you are familiar with Yarn. If it is, you need to update the yarn-site.xml of
NodeManager to set {{yarn.application.classpath}} containing the
{{HADOOP_CONF_DIR}}. I guess it does not be correctly set now. After then,
restarting the NodeManager and submit a new Flink job again. It should work now.
> Flink connect incorrect address of yarn's resouceManager scheduler
> ------------------------------------------------------------------
>
> Key: FLINK-20685
> URL: https://issues.apache.org/jira/browse/FLINK-20685
> Project: Flink
> Issue Type: Bug
> Components: flink-contrib
> Affects Versions: 1.12.0
> Environment: flink version: 1.12.0
> hadoop version: 3.1.3
> flink shaded hadoop version: flink-shaded-hadoop-3-uber-3.1.1.7.1.1.0
> Reporter: Spongebob
> Priority: Major
> Attachments: yarn.log
>
>
> flink version: 1.12.0
> hadoop version: 3.1.3
> flink shaded hadoop version:
> flink-shaded-hadoop-3-uber-3.1.1.7.1.1.0-565-9.0.jar
> Issue desc: When I start the flink session via script "yarn-session.sh", the
> console log shows " Deployment took more than 60 seconds. Please check if the
> requested resources are available in the YARN cluster". And yarn log shows
> this error "Retrying connect to server: 0.0.0.0/0.0.0.0:8030. Already tried 9
> time(s)", in fact the yarn scheduler address is "hadmaster:9901" in the
> yarn-site.xml for me.
> I am sure that my yarn cluster is running normally because I can run other
> application on yarn succefully and there are enough resources too. I had set
> `YARN_CONF_DIR`, `HADOOP_CONF_DIR`, `HADOOP_CLASSPATH` in system env also.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
