sudo apt-get update
sudo apt-get install -y apt-transport-https ca-certificates curl

sudo curl -fsSLo /usr/share/keyrings/kubernetes-archive-keyring.gpg https://packages.cloud.google.com/apt/doc/apt-key.gpg

echo "deb [signed-by=/usr/share/keyrings/kubernetes-archive-keyring.gpg] https://apt.kubernetes.io/ kubernetes-xenial main" | sudo tee /etc/apt/sources.list.d/kubernetes.list

sudo apt-get update
sudo apt-get install -y containerd conntrack socat kubelet kubeadm kubectl

Kubelet ≥ 1.26 requires containerd ≥ 1.6.0.

Initialise cluster

We are going to use cilium in place of kube-proxy https://docs.cilium.io/en/v1.12/gettingstarted/kubeproxy-free/

sudo kubeadm init --skip-phases=addon/kube-proxy

Set up kubectl

https://kubernetes.io/docs/tasks/tools/

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

For remote kubectl copy the config file to local machine

scp gauss@192.168.1.12:/home/gauss/.kube/config ~/.kube/config

(Optional) Remove taint for single node use

Get taints on nodes

kubectl get nodes -o json | jq '.items[].spec.taints'

Remove taint on master node to allow scheduling of all deployments

kubectl taint nodes --all node-role.kubernetes.io/control-plane-

Install Cilium as CNI (Container Network Interface)

To bootstrap the cluster we can install Cilium using its namesake CLI.

For Linux this can be done by running

CILIUM_CLI_VERSION=$(curl -s https://raw.githubusercontent.com/cilium/cilium-cli/main/stable.txt)
CLI_ARCH=amd64
if [ "$(uname -m)" = "aarch64" ]; then CLI_ARCH=arm64; fi
curl -L --fail --remote-name-all https://github.com/cilium/cilium-cli/releases/download/${CILIUM_CLI_VERSION}/cilium-linux-${CLI_ARCH}.tar.gz{,.sha256sum}
sha256sum --check cilium-linux-${CLI_ARCH}.tar.gz.sha256sum
sudo tar xzvfC cilium-linux-${CLI_ARCH}.tar.gz /usr/local/bin
rm cilium-linux-${CLI_ARCH}.tar.gz{,.sha256sum}

See the Cilium official docs for more options.

Next we install Cilium in Kube proxy replacement mode and enable L2 announcements to reply to ARP requests. To not run into rate limiting while doing L2 announcements we also increase the k8s rate limits.

cilium install \
  --set kubeProxyReplacement=true \
  --set l2announcements.enabled=true \
  --set externalIPs.enabled=true \
  --set k8sClientRateLimit.qps=50 \
  --set k8sClientRateLimit.burst=100

See this blog post for more details.

Validate install

cilium status

Cilium LB IPAM

For Cilium to act as a load balancer and start assigning IPs to LoadBalancer Service resources we need to create a CiliumLoadBalancerIPPool with a valid pool.

Edit the cidr range to fit your network before applying it

kubectl apply infra/cilium/ip-pool.yaml

Next create a CiliumL2AnnouncementPolicy to announce the assigned IPs. Leaving the interfaces field empty announces on all interfaces.

kubectl apply infra/cilium/announce.yaml

Sealed Secrets

Used to create encrypted secrets

kubectl apply -k infra/sealed-secrets

Be sure to store the generated sealed secret key in a safa place!

kubectl -n kube-system get secrets

NB!: There will be errors if you use my sealed secrets as you (hopefully) don't have the decryption key

Traefik

Remove the deployment.dnsConfig from infra/traefik/values.yaml and change the io.cilium/lb-ipam-ips annotation to a valid IP address for your network.

Install Traefik

kubectl kustomize --enable-helm infra/traefik | kubectl apply -f -

Port forward Traefik

Port forward Traefik ports in router from 8000 to 80 for http and 4443 to 443 for https. IP can be found with kubectl get svc (it should be the same as the one you gave in the annotation).

Test-application (Optional)

Deploy a test-application by editing the manifests in apps/test/whoami and apply them

kubectl apply -k apps/test/whoami

An unsecured test-application whoami should be available at https://test.${DOMAIN}. If you configured apps/test/whoami/traefik-forward-auth correctly a secured version should be available at https://whoami.${DOMAIN}.

ArgoCD

ArgoCD is used to bootstrap the rest of the cluster. The cluster uses a combination of Helm and Kustomize to configure infrastructure and applications. For more details read this blog post

kubectl apply -k infra/argocd

Get ArgoCD initial secret by running

kubectl -n argocd get secrets argocd-initial-admin-secret -o json | jq -r .data.password | base64 -d

Kubernetes Dashboard

An OIDC (traefik-forward-auth) protected Kubernetes Dashboard can be deployed using

kubectl apply -k infra/dashboard

Create a token

kubectl -n kubernetes-dashboard create token admin-user

ApplicationSets

NB!: This will not work before you've changed all the domain names and IP addresses.

Once you've tested everything get the ball rolling with

kubectl apply -k sets

Cleanup

kubectl drain gauss --delete-emptydir-data --force --ignore-daemonsets
sudo kubeadm reset
sudo iptables -F && sudo iptables -t nat -F && sudo iptables -t mangle -F && sudo iptables -X
sudo ipvsadm -C

Troubleshooting

Kubernetes 1.26 requires containerd 1.6.0 or later due to the removal of support for CRI version v1alpha2 (link).

Make sure that runc is properly configured in containerd.

NB: Make sure the correct containerd daemon is running. (Check the loaded containerd service definition as reported by systemctl status containerd) Follow https://github.com/containerd/containerd/blob/main/docs/getting-started.md for further instructions.

sudo cat /etc/containerd/config.toml

[plugins."io.containerd.grpc.v1.cri".containerd.runtimes.runc]
runtime_path = "/usr/bin/runc"
runtime_type = "io.containerd.runc.v2"

Sealed Secrets

Restart pod after applying master-key.