[
https://issues.apache.org/jira/browse/SPARK-18551?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15703081#comment-15703081
]
Steve Loughran commented on SPARK-18551:
----------------------------------------
hmm. If you are running HDFS unsecure, then all I need to do is set the
{{HDFS_USER}} to anything I want, and can delete files from the command line.
It's just harder to do than through a GUI.
Other filesystems though, they may have more security (example, I may not have
the delete perms to an s3 bucket). so going through the UI would add risk.
Now, if you do want security and restrict yourself to Kerberos only, then
obviously Kerberos is the answer. And the problem, as the web UI and REST APIs
will have to require SPNEGO auth, clients to set it up, kerberos, etc. Not
impossible, and potentially useful long term. If done thoroughly you'd even
have to think about list/view perms on jobs, how to set up the filesystem to
allow per-user logs (or use a different history store), etc. etc.
I would really not encourage going near Kerberos unless it somehow critical to
lock the history down, and there's no easy way to simply bring up a history
server as a standalone YARN AM. Believe me: kerberos is something to fear.
> Add functionality to delete event logs from the History Server UI
> -----------------------------------------------------------------
>
> Key: SPARK-18551
> URL: https://issues.apache.org/jira/browse/SPARK-18551
> Project: Spark
> Issue Type: New Feature
> Components: Spark Core, Web UI
> Reporter: Alex Bozarth
>
> Sometimes a Spark user will only have access to a History Server to interact
> with their (past) applications. But without access to the server they can
> only delete applications through use of the FS Cleaner feature, which itself
> can only clean logs older than a set date.
> I propose adding the ability to delete specific applications via the History
> Server UI with the default setting to off.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
