I'm trying to deploy zeppelin 0.10 on k8s, using following manual build: mvn clean package -DskipTests -Pspark-scala-2.12 -Pinclude-hadoop -Pspark-3.0 -Phadoop2 -Pbuild-distr -pl zeppelin-interpreter,zeppelin-zengine,spark/interpreter,spark/spark-dependencies,zeppelin-web,zeppelin-server,zeppelin-distribion,jdbc,zeppelin-plugins/notebookrepo/filesystem,zeppelin-plugins/launcher/k8s-standard -am
Spark itself is configured to use mesos as resource manager. It seems as if when trying to start the spark interpreter, K8sRemoteInterpreterProcess tries to find a sidecar pod for spark interpreter: Pod pod = client.pods().inNamespace(namespace).withName(podName).get(); Is there any option not to have spark interpreter as a separate pod, and instead just create the spark context within the zeppelin process? I'm trying to understand if I could make zeppelin use K8sStandardInterpreterLauncher instead (I assume it's an alternative to the remote interpreter?) Thanks, Lior
