[
https://issues.apache.org/jira/browse/SPARK-36599?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17405917#comment-17405917
]
Steve Loughran commented on SPARK-36599:
----------------------------------------
I thought things had been fixed up so Hadoop's HTTP binding was not used
through classloaders. If that isn't the case, that is something to fix.
> ExecutorClassLoader no longer works with Http based Class Servers
> -----------------------------------------------------------------
>
> Key: SPARK-36599
> URL: https://issues.apache.org/jira/browse/SPARK-36599
> Project: Spark
> Issue Type: Bug
> Components: Spark Core
> Affects Versions: 3.1.2
> Reporter: VinothKumar Raman
> Priority: Major
>
> There are 2 primary issues,
> * If the classServer uri is `{{http://host:port}}` instead of
> `{{http://host:port/}}` the getPath on the URI object returns empty string
> and subsequently causes Path creation to fail with Path cannot be empty++
> * This is a regression issue and transitively a hadoop issue too, the Http
> Filesystem fails with URI not absolute error unlike other filesystems when
> the path doesn't have scheme or authority. This used to work when there was
> http specific implementation and no longer work with Filesystem based
> implementation
> https://issues.apache.org/jira/browse/HADOOP-17870
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
