Thanks for pointing this out. It is impacting 0.7.0 as well. ________________________________ From: Michael Sells <mjse...@gmail.com> Sent: Wednesday, August 17, 2016 5:46:35 AM To: users@zeppelin.apache.org Subject: Re: Issue loading dependency with SPARK_SUBMIT_OPTIONS w/ 0.6.1
Thanks. I'll file a Jira. Looks like custom repos aren't loading either from either source. I'll try to replicate as well. Mike On Tue, Aug 16, 2016 at 5:55 PM Mina Lee <mina...@apache.org<mailto:mina...@apache.org>> wrote: I also could reproduce it with Spark 2.0.0, but not with Spark 1.6.1. If you want to use Zeppelin with Spark 2.0, one alternative you can try is using [1] "dependencies" in GUI interpreter menu. [1] http://zeppelin.apache.org/docs/0.6.1/manual/dependencymanagement.html On Wed, Aug 17, 2016 at 1:46 AM Jeff Zhang <zjf...@gmail.com<mailto:zjf...@gmail.com>> wrote: I can reproduce it in 0.6.1 & master branch, please file a ticket for that. On Wed, Aug 17, 2016 at 4:09 AM, Michael Sells <mjse...@gmail.com<mailto:mjse...@gmail.com>> wrote: Testing out 0.6.1 with Spark 2.0 and discovered the way we load dependencies doesn't seem to be working with the new update. We pass new dependencies in via a SPARK_SUBMIT_OPTIONS environment variable pass the following flags: --packages com.databricks:spark-avro_2.11:3.0.0 Now when I try to import it with: import com.databricks.spark.avro._ I get: <console>:25: error: object databricks is not a member of package com import com.databricks.spark.avro._ I checked the logs are there is no error retrieving the package. So it seems to be something with the classpath. This works in 0.6.0. Any idea if something changed or if we're doing something wrong? I tried this with a few internal packages as well and it doesn't work with those either. Thanks, Mike -- Best Regards Jeff Zhang
