The following manual explains how to install a Kubernetes cluster with kubeadm.
Setup a server and make sure that you can SSH into the server.
In the following examples we assume that our server is named k1.
For this we use an Ubuntu image:
Before you proceed you should be able to SSH into the server:
ssh k1
How you setup your server is up to you. If you are on a Linux machine with KVM and Libvirt an easy way to quickly setup a VM is using vu.
The tool vu provides an easy way to setup a virtual machine based on a cloud-init image.
vu image add https://cloud-images.ubuntu.com/minimal/releases/focal/release-20220810/ubuntu-20.04-minimal-cloudimg-amd64.img
vu create --disk-size 100GB --memory 4G --cpu 2 ubuntu-20.04-minimal-cloudimg-amd64.img k1
If the server is up and running, we first run the setup script on the server to install kubelet, kubeadm, containerd, etc. The setup script basically performs the tasks described under https://kubernetes.io/docs/setup/production-environment/tools/kubeadm/install-kubeadm/.
Run the setup script on the server:
ssh k1 <setup.sh
Then you can bootstrap the cluster with kubeadm:
ssh k1 sudo kubeadm init --pod-network-cidr=10.244.0.0/16
To test the WASM extension a single node is enough.
If you still want more nodes in your cluster, you can repeat the actions above to add more nodes to a cluster (setup VM, run setup script).
Instead of kubeadm init you have to run the appropriate kubeadm join command which is shown after kubeadm init.
You also can generate a new join command like this:
token=$( ssh k1 sudo kubeadm token generate )
join_cmd=$( ssh k1 sudo kubeadm token create $token --print-join-command )
Then run the join command on the additional node k2:
ssh k2 sudo $join_cmd
To test the features of the WASM extension you don't need a network plugin. If you want to deploy workload in your cluster you have to install a network plugin like Flannel:
ssh k1 sudo kubectl --kubeconfig /etc/kubernetes/admin.conf apply -f https://raw.githubusercontent.com/coreos/flannel/master/Documentation/kube-flannel.yml
To connect to your test cluster from your local machine you have to configure kubectl accordingly.
The easiest way to do so is to copy the admin configuration from your server:
ssh k1 sudo cat /etc/kubernetes/admin.conf > ~/.kube/test_config
To not overwrite our existing configuration we put the configuration in a different location and then set the location of the config using the KUBECONFIG environment variable.
export KUBECONFIG=~/.kube/test_config
A kubectl get nodes should now list the available nodes.
Publish the kube-apiserver Docker image of your custom build to a Docker registry.
In the following example I use the image dvob/kube-apiserver:dev.
Login to the k1 server and update the manifest of the API server /etc/kubernetes/manifests/kube-apiserver.yaml under spec.containers[0].image and set it to dvob/kube-apiserver:dev.
Alternatively, you can run this command:
ssh k1 'sudo kubectl set image -f /etc/kubernetes/manifests/kube-apiserver.yaml kube-apiserver=dvob/kube-apiserver:dev --local -o yaml | sudo sponge /etc/kubernetes/manifests/kube-apiserver.yaml'
Usually it takes some time until the kubelet picks up the new manifest and restarts the API server.
To speed up the restart of the API server you can login to the server k1 and force a restart as follows:
ssh k1
sudo su -
crictl ps --name kube-apiserver | awk 'NR != 1{ print $1}' | xargs crictl stop
systemctl restart kubelet