Kubernetes and Mesos, better together.
Google’s Kubernetes is a platform for building distributed applications using Docker containers that comes directly from Google Cloud Platform. In July we announced a collaboration between Mesosphere and Google to bring Kubernetes to Mesos, giving users the ability to run Kubernetes alongside other first-class datacenter services such as Hadoop, Spark, Marathon, and Chronos. This allows Kubernetes applications to run alongside all other types of workloads simultaneously, on the same set of servers. Mesos ensures resources are fairly allocated across services and that running applications are isolated from each other.
Shortly before Google officially announced collaboration with Mesosphere in July, we started working on the Kubernetes-Mesos project to integrate Kubernetes with the Mesosphere stack. Mesosphere CEO Florian Leibert gave an early demonstration together with Google’s John Wilkes during his MesosCon keynote presentation in Chicago this summer. Today we’re excited to announce that, like Kubernetes itself, the Kubernetes-Mesos project has matured rapidly and is ready for a larger audience.
Follow our Kubernetes on Mesos on Google Cloud Platform tutorial
Our partner Ishi Systems have been working with us on this project and play an important role supporting successful Apache Mesos deployments for their customers. In the screencast below, James DeFelice walks through the process of setting up Kubernetes on a Mesosphere for Google Cloud Platform cluster.
How does it work?
Kubernetes-Mesos integrates the Kubernetes scheduler API with the Mesos scheduler API, matching up requested pods to resource offers. Each scheduled pod is also a Mesos task. From a user’s perspective, the integration is transparent. You can set up replication controllers and services, and interact with the cluster using
kubecfg or the REST API.
We’re looking forward to the 2015 roadmap for both Mesos and Kubernetes. In particular, we’re excited about the forthcoming resource model implementation, which will provide a richer context for scheduling pods. So, stay tuned!
Talk to us on Twitter at @mesosphere or in the
#mesos IRC channel on Freenode.