[
https://issues.apache.org/jira/browse/STORM-1625?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15207149#comment-15207149
]
ASF GitHub Bot commented on STORM-1625:
---------------------------------------
Github user ptgoetz commented on the pull request:
https://github.com/apache/storm/pull/1239#issuecomment-199989200
@revans2 The `storm sql` command (via the `StormSql` class) builds a fat
topology jar based on the class path, and submits it.
I have some concerns around this approach, as it seems brittle to me.
Specifically the code that builds the jar does not seem to be
resource/MANIFEST-aware (unlike the shade plugin). That worries me with respect
to Hadoop integration (*-site.xml, etc.). There's also the issue with having to
manually make the kafka jar available.
> Move storm-sql dependencies out of lib folder
> ---------------------------------------------
>
> Key: STORM-1625
> URL: https://issues.apache.org/jira/browse/STORM-1625
> Project: Apache Storm
> Issue Type: Bug
> Components: storm-core, storm-sql
> Reporter: Abhishek Agarwal
> Assignee: Abhishek Agarwal
>
> We shade guava classes inside storm-core so that client can work with any
> version of guava without any clashes. However, storm-sql-core has a
> transitive dependency on guava and thus the guava jar is still getting
> shipped in lib folder.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
