i built a Raspberry Pi SUPER COMPUTER!! // ft. Kubernetes (k3s cluster w/ Rancher)

NetworkChuck Technology Careers Career & Money

About this video

The creator builds a cluster from eight Raspberry Pis and presents it as a hands-on way to learn Kubernetes, which he describes as a sought-after IT skill. He explains clustering as orchestration rather than merging the machines into one computer: the Pis remain separate servers, but Kubernetes distributes containerized apps across them, can run multiple copies for load balancing and redundancy, and can scale or check health automatically. Because full Kubernetes is heavy for a Pi, he installs k3s, Rancher Labs’ lightweight version aimed at edge and IoT devices. The walkthrough covers a headless setup — imaging Raspberry Pi OS Lite, enabling cgroups, a static IP, 64-bit mode and SSH via the boot partition — then installing k3s on a first node as master/control plane and joining the others as worker nodes with a token from the master on port 6443, verified with kubectl get nodes. Optionally he installs the Rancher GUI on a separate Ubuntu VM to import and visualize the cluster, including an ARM64 agent image override for the Pis. From the master he applies YAML manifests: a deployment of six Nginx pods, exposed through a NodePort service on port 31111 because pod cluster IPs are not reachable from outside; a 50-replica hello-world deployment to visualize load balancing; an Ingress mapping a hostname to the service; and finally a Minecraft server installed via a Helm chart through Rancher, exposed on its game port 25565, which he then joins and plays. He stresses the ARM64 limitation — not all container images will run on Pis — and frames the project mainly as skills practice with bragging rights.

Key takeaways

A cluster orchestrates separate machines: Kubernetes schedules containers across nodes, distributes load, and maintains the replica count you request instead of merging hardware into one PC
k3s is presented as lightweight Kubernetes suited to Raspberry Pis and edge devices, with one master/control-plane node directing worker nodes that register using a token
Apps are declared in YAML manifests and applied with kubectl; pods wrap containers, and services are required to expose them
Cluster-internal pod IPs are unreachable externally by default — a NodePort service forwards a high port on every node to the app’s port and load-balances across pods
Raspberry Pi clusters use ARM64, so image compatibility must be checked; Helm charts and the Rancher GUI are offered as alternatives to hand-written manifests and CLI-only management

Episode details: published July 15, 2021 · 1:03:42 long. Summary written by Self Growth Videos from the episode itself. Business results and any income figures mentioned are NetworkChuck’s own claims, shared for education — they are not typical or guaranteed.

← Previous: DDNS on a Raspberry Pi using the Cloudflare API (Dynamic DNS) · Next: access EVERYTHING from your web browser!! (Linux and Windows Desktop, SSH) // Guacamole Install →

All NetworkChuck videos · NetworkChuck profile · NetworkChuck on YouTube