Docker Enterprise provides a production-ready CaaS platform, as we have been discussing throughout this chapter. In this section, we will review the minimum logical requirements for deploying Docker Enterprise in production.

We learned that Docker Swarm and Kubernetes require an odd number of master nodes to work properly. Docker Swarm does not require an external key-value store, while Kubernetes does. Docker Enterprise will deploy this key-value store with UCP, so a minimum of three manager nodes will be required to provide high availability. All managers will run the same services. In Docker Swarm and Kubernetes, we have a leader node that writes cluster changes in the database. Other managers will sync their data, but we can also run administration commands on any of them. We need to integrate an external load balancer to distribute API requests on all manager nodes.