Spark can use MR3 as its execution backend. As MR3 supports Hadoop and Kubernetes, Spark on MR3 can run on Hadoop and Kubernetes. Thus Spark on MR3 provides an alternativey way to run Spark on Hadoop and Kubernetes. For the instruction on installing Spark on MR3, see Install Spark on MR3.
For an introduction to Spark on MR3, see the blog article Spark on MR3 - A New Way to Run Apache Spark.
For trying Spark on MR3 on Kubernetes, we recommend the quick start guide On Kubernetes.
For asking questions, please visit MR3 Google Group.