This guide walks a deployer through launching a multi-node Kubernetes cluster on bare metal servers running CoreOS. After completing this guide, a deployer will be able to interact with the Kubernetes API from their workstation using the kubectl CLI tool.
For all nodes running Kubernetes components (masters & workers), you must use CoreOS version 773.1.0+ on the Alpha or Beta channel for the kubelet to be present in the image.
Use the official CoreOS bare metal guides for installation instructions:
Mixing multiple methods is possible. For example, doing an install to disk for the machines running the etcd cluster and Kubernetes master nodes, but PXE-booting the worker machines.
The following guides assume the use of flannel as a software-defined overlay network to manage routing of the pod network. However, bare metal is a common platform where a self-managed network is used, due to the flexbility provided by physical networking gear.
See the Kubernetes networking documentation for more information on self-managed networking options.
Did you install CoreOS on your machines? An SSH connection to each machine is all that's needed. We'll start the configuration next.
I'm ready to get started