By default Spark restart policies for both executor and driver pods are set to "Never" and that will override all Kubernetes configurations: link.

In the case, you need to restart, I guess you have to do with bash, kubectl scripts.