Install a Kubernetes Cluster with kubeadm¶

Spin up a production-grade Kubernetes cluster on any Linux server — VM, EC2, bare-metal, microVM, or cloud VPS — with a single curl | bash command.

No local cloning. No manual step-by-step guides. No forgetting to disable swap.

The Problem This Solves¶

Setting up Kubernetes with kubeadm is not one command. It is 30+ manual steps spread across the official documentation: disable swap, load kernel modules, set sysctl parameters, install containerd, configure its cgroup driver, install runc, install crictl, add the Kubernetes APT repo, pin versions, run kubeadm init, configure kubectl, install a CNI — and if you miss one step or run them in the wrong order, the cluster silently breaks or never comes up at all.

SilverStack fills the gap: a fully automated, idempotent, version-aware bash automation layer that bootstraps a real upstream Kubernetes cluster via kubeadm on any server you own or rent.

Who Is This For¶

• A DevOps engineer who spun up an EC2 instance, a Hetzner VPS, or a DigitalOcean Droplet and needs a real Kubernetes cluster in minutes for testing Ingress, Gateway API, Helm charts, or service mesh configurations
• Anyone running a disposable lab on iximiuz, Killercoda, or a local microVM who does not want to re-type 30 steps every time
• Someone who wants to understand what a proper kubeadm bootstrap actually does — this code is readable, every phase is labeled, every decision is explained

Not for production. This is a lab and learning tool. The scripts are designed for fresh, disposable servers. They assume you are the owner and root on the machine.

Requirements¶

Quick Start — Two Commands for a Full Cluster¶

Step 1 — Initialize the Control Plane¶

Run this on the node that will be your control plane:

curl -fsSL https://raw.githubusercontent.com/ibtisam-iq/silver-stack/main/scripts/kubernetes/entrypoints/init-controlplane.sh | sudo bash

The script will:

1. Ask you 5 questions (your node IP, Kubernetes version, hostname, Pod CIDR, containerd install method) — all have sensible defaults, just press Enter
2. Disable swap, load kernel modules, apply sysctl settings
3. Install containerd with the correct cgroup driver configuration
4. Resolve the exact patch version for your chosen Kubernetes minor version
5. Install kubelet, kubeadm, kubectl, helm, and k9s
6. Detect and safely clean any existing Kubernetes state
7. Run kubeadm init with all required flags
8. Print the kubeadm join token for your worker nodes

Total time: 5–10 minutes on a fresh server with good internet.

After it finishes, run the kubeconfig helper (as your normal user, not root):

curl -fsSL https://raw.githubusercontent.com/ibtisam-iq/silver-stack/main/scripts/kubernetes/cluster/kubeconfig-helper.sh | bash

Then install a CNI plugin so your nodes become Ready:

curl -fsSL https://raw.githubusercontent.com/ibtisam-iq/silver-stack/main/scripts/kubernetes/cni/install-cni.sh | sudo bash
# Choose: 1) Calico  or  2) Flannel

Step 2 — Join Worker Nodes¶

Run this on each worker node, then run the kubeadm join ... command that init-controlplane.sh printed at the end:

curl -fsSL https://raw.githubusercontent.com/ibtisam-iq/silver-stack/main/scripts/kubernetes/entrypoints/init-worker-node.sh | sudo bash
# Then paste the kubeadm join command from the control plane output

Verify¶

kubectl get nodes          # All nodes should be Ready
kubectl get pods -A        # All system pods should be Running
kubectl cluster-info

Local Development — KinD Cluster¶

If you are on your laptop and Docker is already running, use the KinD path instead — no server required:

# curl -fsSL https://raw.githubusercontent.com/ibtisam-iq/silver-stack/main/scripts/kubernetes/entrypoints/create-kind-cluster.sh | bash

Two cluster configurations are available:

Prerequisites: Docker Desktop or Docker Engine running locally, kind CLI installed, kubectl installed.

Dry Run Mode¶

Every script accepts --dry-run. In this mode the script runs all preflight checks, resolves versions, and prints what it would do — but makes no system changes.

# curl -fsSL https://raw.githubusercontent.com/ibtisam-iq/silver-stack/main/scripts/kubernetes/entrypoints/init-controlplane.sh | sudo bash -s -- --dry-run

Use this to validate that your server meets all requirements before committing to the full install.

What Gets Installed¶

On Every Node (control plane and workers)¶

Control Plane Only (additional)¶

How Version Pinning Works¶

You provide a MAJOR.MINOR string (e.g. 1.36). The scripts:

1. Query https://dl.k8s.io/release/stable-1.36.txt to get the exact patch version (v1.36.0)
2. Convert it to the Debian package revision (1.36.0-1.1)
3. Install kubelet=1.36.0-1.1 and kubeadm=1.36.0-1.1 — no floating version installs
4. Run apt-mark hold on both — prevents accidental upgrade

This means your cluster will never silently move to a new patch version.

CNI Options¶

Two CNI plugins are supported. The script prompts you to choose after kubeadm init.

Calico (recommended for most cases)¶

Installed via the Tigera Operator. Supports full Kubernetes NetworkPolicy plus Calico-native policies.

• The script auto-detects your Pod CIDR from the cluster's ConfigMap — no manual CIDR editing
• Supports VXLAN (default) or IPIP encapsulation
• Resolves the latest stable Calico version automatically

kubectl -n calico-system get pods   # Verify
kubectl get nodes                   # Should show Ready

Flannel (simpler, no NetworkPolicy)¶

Installed via a patched manifest. Good for single-tenant labs where NetworkPolicy enforcement is not needed.

kubectl -n kube-flannel get pods    # Verify
kubectl get nodes                   # Should show Ready

Resetting a Cluster¶

To wipe the cluster and start fresh (useful in disposable labs):

# curl -fsSL https://raw.githubusercontent.com/ibtisam-iq/silver-stack/main/scripts/kubernetes/maintenance/reset-cluster.sh | sudo bash

The script: 1. Requires typed YES confirmation — will not run silently 2. Stops kubelet 3. Runs kubeadm reset -f 4. Removes /etc/kubernetes, /var/lib/kubelet, /var/lib/etcd, ~/.kube 5. Detects and kills any orphaned kube-apiserver process 6. Leaves containerd installed but stopped — ready for the next init

After reset, run init-controlplane.sh again. CNI binaries, OS prerequisites (swap, sysctl), and installed binaries (kubectl, helm, k9s) are not removed by reset and do not need to be re-installed.

Safety Features¶

These are things the scripts do that manual guides do not:

Known Limitations¶

Under the Hood¶

The automation is organized into clearly separated layers:

entrypoints/     ← Start here. Two scripts: control plane and worker.
cluster/         ← kubeadm init logic, cluster parameters wizard, safety guards
node/            ← OS prerequisites: swap, kernel modules, sysctl
runtime/         ← containerd, runc, crictl, CNI binaries
packages/        ← kubelet, kubeadm, kubectl, helm, k9s — with version pinning
cni/             ← Calico and Flannel installers
maintenance/     ← Reset scripts
manifests/       ← KinD cluster YAML configurations
lib/             ← Shared library: logging, remote execution, kubeconfig helpers

For a deep dive into what each script does, see the internal runbooks:

• Node Preparation
• Container Runtime
• Kubernetes Packages
• Cluster Bootstrap
• Kubeconfig & CNI
• Maintenance & Reset
• KinD Local Cluster

Source¶

All scripts are open source and available at:

github.com/ibtisam-iq/silver-stack

scripts/kubernetes/ is the root of everything described on this page.