[
https://issues.apache.org/jira/browse/METRON-799?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15961291#comment-15961291
]
ASF GitHub Bot commented on METRON-799:
---------------------------------------
Github user justinleet commented on the issue:
https://github.com/apache/incubator-metron/pull/518
So coming back to the AutoTGT discussion and making it easier to see.
There's basically two approaches we can takes that (should) work.
1. The current use of AutoTGT. This requires setting up .storm/storm.yaml
for the Storm nodes, and users will have to do the same.
2. The AutoHDFS / AutoHBase solution mentioned by @dlyle65535. This
requires symlinking some jars and configs to make them available to Storm.
Should only be necessary for Nimbuses, but it does mean that HDFS/HBase
upgrades can make the symlinks stale.
Either solution requires Ambari acting on a different node (potentially)
than the one running the scripts. I don't know if Ambari has any resources for
handling that sort of things. It could potentially be a command over ssh to
another node, but presumably that requires passwordless ssh setup or someone to
manually create the symlinks (which may be acceptable for this pass).
I'm fairly strongly inclined towards the second one, primarily because it
requires less effort on the users part. Ambari work is fairly similar either
way.
> The MPack should function in a kerberized cluster
> -------------------------------------------------
>
> Key: METRON-799
> URL: https://issues.apache.org/jira/browse/METRON-799
> Project: Metron
> Issue Type: Improvement
> Reporter: Casey Stella
> Assignee: Justin Leet
> Labels: kerberos
>
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
