Spark on Kubernetes (which is still not fully production ready) will be the clearest indicator of k8s taking over the space occupied by Mesos. Mesos/Yarn is us

Spark on Kubernetes (which is still not fully production ready) will be the clearest indicator of k8s taking over the space occupied by Mesos. Mesos/Yarn is used by spark to schedule batch jobs in a distributed manner on a spark cluster. So they have to have very deep knowledge of the resource availability and the needs of the jobs being scheduled. We are not talking static resource allocation like a container, but scheduling a particular piece of processing in real time to one of the nodes managed by Mesos/Yarn. Kubernetes does not come with this level of awareness today. Mesos was always built with this in mind. So for Mesos, the scheduling of containers is a subset of its functionality. However kubernetes is getting there - https://github.com/kubernetes/kubernetes/issues/34377 https://github.com/kubernetes/kubernetes/issues/34377 https://github.com/apache-spark-on-k8s/spark/issues/4 https://github.com/apache-spark-on-k8s/spark/issues/4 https://issues.apache.org/jira/plugins/servlet/mobile#issue/SPARK-18278 https://issues.apache.org/jira/plugins/servlet/mobile#issue/... Today if you want to deploy a fleet of containers and schedule a bunch of processes across them in a resource aware manner...you have to use both kubernetes and mesos.