OpenShift command line client has a nice little feature. With oc cluster up command you can easily bring in local OpenShift cluster using only Docker engine and nothing else. It will run OpenShift as a Docker container and provide you all important OpenShift fetarues. Unlike MiniShift it doesn't run in any virtual machine. It runs directly on your PC. That has some major advantages when you are developing software for Kubernetes. For example you share the network in which your OpenShift pods and services run. That means you can access them directly using the IP address from applications running locally. You don't have to forward ports from the Minishift machine or proxy the connection.
This project tries to create experience very similar to the one provided by oc cluster up but using pure Kubernetes instead of OpenShift. Currently it works, but it is not yet as nice and simple as OpenShift is.
kube-cluster-up currently supports only Linux OS. On MacOS X and on Windows Docker runs inside virtual machine. Therefore even when you run something similar to oc cluster up you will not be able to connect directly to the pods and services - and that would mean loosing the main advantage over Minishift or Minikube.
kube-cluster-up currently relies on kubeadm and kubelet being installed locally. Both kubeadm and kubelet are provided as RPM and DEB packages for the most popular Linux distributions. Follow the installation guide to install them. kube-cluster-up uses kubeadm to configure and deploy other components of Kubernetes cluster.
To install kube-cluster-up you can simply clone the repository or download the release. The shell script in this repository can be placed anywhere in your system.
kube-cluster-up can be started by running the kube-cluster-up.sh script. This currently requires root privileges to configure and start some of the services. Therefore sou should call it through sudo:
sudo kube-cluster-up.sh
The startup script will enable and start the kubelet service and run kubeadm to configure and deploy the remainning Kubernetes services (API server, Controller, etcd, ...). Once kubeadm is finished, the scirpt will deploy Calico network and install Heapster for resource monitoring and Kubernetes Dashboard. It will also remove labels and taints from the only node so that it can be used to schedule workloads.
kubeconfig file will be created in /home/<your-username>/kubeconfig. You can use this with kubectl to communicate with the cluster:
export KUBECONFIG=/home/<your-username>/kubeconfig
kubectl get nodes
You can use kubectl to deploy your own services.
To communicate with a pod or service directly, use kubectl to find the IP address. Use:
kubectl get service
or:
kubectl get pods -o wide
You can use the IP adress together with the directly
Use the method above to find the Dashboard IP address and enter it into browser:
kubectl get service kubernetes-dashboard --namespace kube-system
The cluster can be stopped using kube-cluster-down.sh. It will reset the kubeadm setup (stop all workloads and Kubernetes system containers) and stop the kubelet service.
- Run
kubeletas docker container and not as system service - Run without
sudo - Run without
kubeadm