1

I have a Zeppelin (0.8.2) deployment on a Kubernetes cluster and I have packed Spark (2.4.0) binaries in the container. Then I set $SPARK_HOME to the Spark installation directory. The spark interpreter seems to work but when I try to change the serviceAccount, it still uses default.

Here are the configurations I tried based on Spark 2.4.0 docs:

spark.kubernetes.authenticate.driver.serviceAccountName <my_serviceAccount>
spark.kubernetes.authenticate.serviceAccountName        <my_serviceAccount>

I tried setting those via $SPARK_SUBMIT_OPTIONS, Zeppelin Interpreter Settings and via spark-defaults.conf but they all give the same results. Am I missing something?

Any help would be greatly appreciated. Thanks!

Joshua Villanueva
  • 179
  • 1
  • 5
  • 13

1 Answers1

0

I was able to use a non-default serviceAccount by adding .pod.spec.serviceAccountName. Provided some links on how you can do it.

[1] How to configure a non-default serviceAccount on a deployment
[2] https://kubernetes.io/docs/tasks/configure-pod-container/configure-service-account/

Joshua Villanueva
  • 179
  • 1
  • 5
  • 13