Set Up Your Own Kubernetes Cluster via VirtualBox

The objective of this post will set up three nodes Kubernetes(K8S) cluster on VirtualBox and launch an application/nginx.

What you will learn

• Using vagrant spin up multiple virtual machines
• Install Kubernetes and set up Controller(master) and Workers
• Launch an application scale up/down and expose it.
• containerd as the container runtime. we will not use Docker

Kubernetes

Tools

• Virtualbox
• containerd will be installed via script
• Kubernetes will be installed via script
• Vagrant

PS: This tutorial was tested on Linux.

please make sure on your host machine Vagrant and VirtualBox are installed.

For sure you can play around with Kubernetes via minikube as well, but to be honest it does not feel like a real use case for K8S. I mean minikube is easy, we love complicated things:). All resources are accessible on my GitHub repo here( PS: Don’t hesitate give a star

). I have prepared some script which makes easy repeated tasks.

• control node at least ~ 2GB RAM
• Workers ca ~ 1GB RAM

$ git clone https://github.com/mbaykara/k8s-cluster.git
$ ls
k8s-setup-master.sh  README.md  Vagrantfile
$ vagrant up 

I have defined one control node and two workers, have look Vagrantfile for further details. vagrant up` command will create master and worker-1, worker-2

vagrant create VMS, it will take a few minutes

The master node is ready we can ssh into the machine via vagrant ssh master

From now on can initialize kubeadm on the master node to manage the workers as root. We will use kubeadm for setting up the cluster. That’s why this is not the really hard way of setting up a k8s cluster. Otherwise, all Kubernetes core components such as etcd, apiserver, controller manager, scheduler etc. have to be run as service on the master node. This will be an overhead for a beginner. Maybe in another post, we can set up really hardway. Anyway, let us move forward.

Run the following command as root, it will take a few mins

root@master:~# kubeadm init --apiserver-advertise-address 192.168.33.13 --pod-network-cidr=10.244.0.0/16

init process

When it is done, should looks as:

init done

As we see that it’s relatively straightforward and explain what should be done in the next step.

Exit from the root and run the following command as an arbitrary user

 $ mkdir -p $HOME/.kube
 $ sudo cp -i /etc/kubernetes/admin.conf $HOME/.kube/config
 $ sudo chown $(id -u):$(id -g) $HOME/.kube/config

If you now list nodes till now it is the only one which is master by kubectl get nodes` you will see, the STATUS` is not ready. Simply because we did not apply any network plugin.

I will use flannel net plugin for the cluster network.

$ kubectl apply -f https://raw.githubusercontent.com/coreos/flannel/master/Documentation/kube-flannel.yml

ca 2 min later, the node is ready!

At this point the Schedulerand Controller-panel might be unhealthy.

$ kubectl get cs

kubectl get componentstatus

Modify the following files on all master nodes:

$ sudo vim /etc/kubernetes/manifests/kube-scheduler.yaml
Clear the line (spec->containers->command) containing this phrase:  -- port=0
$ sudo vim /etc/kubernetes/manifests/kube-controller-manager.yaml
Clear the line (spec->containers->command) containing this phrase: --- port=0
$ sudo systemctl restart kubelet.service

Now master/controller node is ready, we can simply add workers.

step 1. vagrant ssh worker-1
step 2. run the command below as root
 kubeadm join 192.168.33.13:6443 --token bx86lo.agyszwr53ow5y53u \
    --discovery-token-ca-cert-hash sha256:536b10417f411de9ff9f11cb83d286f9217f5031845df93355b3a6a5ed96c066

Repeat both steps for each worker. Then go to the master node

cluster status

No, finally we are ready to deploy nginx! :D

I will deploy one Nginx service and 5 instances of the webserver

$ kubectl create deployment nginx --image=nginx --port 80
$ kubectl create deployment webserver --image=nginx --port 80 --replicas=5

To access from the internet we have to expose it as follow

vagrant@master:~$ kubectl expose deployment webserver --port 80 --type=NodePort

or via curl

Note that webserver launched on worker-2 that's why you have to access it via worker-2 IP Address.

Troubleshooting

1. By default, your cluster will not schedule Pods on the control-plane/master node for security reasons. If you want to be able to schedule Pods on the control-plane node, run:

kubectl taint nodes --all node-role.kubernetes.io/master-

2. For further issues using kubectl describe nodes [nodeName] will let you know where eventually the problem is.

Resources:
* https://v1-17.docs.kubernetes.io/docs/setup/production-environment/tools/kubeadm/create-cluster-kubeadm/
*https://github.com/coreos/flannel/blob/master/Documentation/troubleshooting.md
* https://kubernetes.io/docs/concepts/scheduling-eviction/taint-and-toleration/
* https://github.com/mmumshad/kubernetes-the-hard-way

PS: In case you face any issue through this tutorial, leave a comment please, I will respond as soon as possible.