[
https://issues.apache.org/jira/browse/FLINK-14033?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16931732#comment-16931732
]
Zhenqiu Huang commented on FLINK-14033:
---------------------------------------
Agree. It is not for a jar. The issue happens on another path. Basically, if a
user wants to put a hdfs file as resource in the distributed cache, the file is
not accessible in runtime due to the blob item is not prepared for per job
mode. For session mode, the cache files will be created through rest-client
before job graph submission.
> Distributed caches are not registered in Yarn Per Job Cluster Mode
> ------------------------------------------------------------------
>
> Key: FLINK-14033
> URL: https://issues.apache.org/jira/browse/FLINK-14033
> Project: Flink
> Issue Type: Bug
> Components: Deployment / YARN
> Affects Versions: 1.9.0
> Reporter: Zhenqiu Huang
> Assignee: Zhenqiu Huang
> Priority: Minor
>
> CacheFiles in StreamExecutionEnvironment is not used in Job Submission in the
> Yarn per job cluster mode. Compare to the job submission in session cluster
> mode that will upload distributed cache files onto http server in application
> master, we should get the cache files in job graph and register them into
> blob store in YarnJobClusterEntrypoint.
--
This message was sent by Atlassian Jira
(v8.3.2#803003)
