I confirm,
Christopher was very kind helping me out here. The solution presented in the
linked doc worked perfectly. IMO it should be linked in the official Spark
documentation.
Thanks again,
Roberto
> On 20 Jun 2015, at 19:25, Bozeman, Christopher wrote:
>
> We worked it out. There was m
We worked it out. There was multiple items (like location of remote metastore
and db user auth) to make HiveContext happy in yarn-cluster mode.
For reference
https://github.com/awslabs/emr-bootstrap-actions/blob/master/spark/examples/using-hivecontext-yarn-cluster.md
-Christopher Bozeman
On
Hi Roberto,
I'm not an EMR person, but it looks like option -h is deploying the necessary
dataneucleus JARs for you.The req for HiveContext is the hive-site.xml and
dataneucleus JARs. As long as these 2 are there, and Spark is compiled with
-Phive, it should work.
spark-shell runs in yarn-client