Github user whai2kay commented on the issue:
https://github.com/apache/zeppelin/pull/3240
> Right. I just pushed a commit fa36c18, and it allows user override master
spark interpreter property. If master is not starting with k8s://, it does not
configure Spark on Kubernetes. So I think this opens a possibility to connect
other types of Spark cluster.
It would seem that when setting the master to: spark://spark-cluster:7077 ,
it would just use local[*] instead. Is connecting to a cluster using client
mode not yet supported?
---
