this is a example configuration for Anthos Config Management. it has CRDs from the spark-on-k8s-operator which are in the cluster directory and it instantiates a spark operator resource using the spark-operator.yaml which lives in a dedicated namespace, spark-operator.
there is an example to show how the acme datascience team might set up jobs under a job directory and have RBAC configured automatically for respective subdirectories. the job there is based on the spark-pi-schedule.yaml example which runs a spark job every 5 minutes in the cluster.
this config set up works well with kind which uses docker not actual VMs for pods. if you're using a real enviroment like GKE, be sure to watch resources and in particular, CPUs. you might need more than default clusters provide because spark is pretty core hungry.