Docker & Kubernetes on Raspberry Pi: Edge Orchestration

Cloud-native tools aren't just for data centers. Running Kubernetes at the edge gives you self-healing, rolling updates, and service discovery — right on your factory floor.

Why Containers at the Edge?

• Reproducibility — Same container runs in dev and production
• Isolation — Services don't interfere with each other
• Updates — Rolling deployments with zero downtime
• Scalability — Add nodes to handle more sensors

Hardware Setup

Our recommended edge cluster:

| Node | Hardware | Role | |------|----------|------| | Node 1 | RPi 4 (8GB) | K3s Server (Control Plane) | | Node 2 | RPi 4 (4GB) | K3s Agent (Worker) | | Node 3 | RPi 4 (4GB) | K3s Agent (Worker) |

Installing K3s

K3s is a lightweight Kubernetes distribution perfect for ARM devices.

Server Node

curl -sfL https://get.k3s.io | sh -s - \
    --write-kubeconfig-mode 644 \
    --disable traefik

Agent Nodes

curl -sfL https://get.k3s.io | K3S_URL=https://192.168.1.100:6443 \
    K3S_TOKEN=$(cat /var/lib/rancher/k3s/server/node-token) \
    sh -

Deploying an IoT Data Pipeline

Deployment: MQTT Broker

apiVersion: apps/v1
kind: Deployment
metadata:
  name: mosquitto
spec:
  replicas: 1
  selector:
    matchLabels:
      app: mosquitto
  template:
    metadata:
      labels:
        app: mosquitto
    spec:
      containers:
      - name: mosquitto
        image: eclipse-mosquitto:2
        ports:
        - containerPort: 1883
        volumeMounts:
        - name: config
          mountPath: /mosquitto/config

Deployment: Data Processor

apiVersion: apps/v1
kind: Deployment
metadata:
  name: data-processor
spec:
  replicas: 2
  selector:
    matchLabels:
      app: processor
  template:
    spec:
      containers:
      - name: processor
        image: ghcr.io/itsyourturn/iot-processor:latest
        env:
        - name: MQTT_BROKER
          value: "mosquitto.default.svc.cluster.local"
        resources:
          limits:
            memory: "256Mi"
            cpu: "500m"

Monitoring

Deploy Prometheus + Grafana for cluster observability:

• CPU/Memory per pod and node
• MQTT message rates via custom exporters
• Disk usage alerts before SD cards fill up
• Pod restart notifications via webhook

Production Tips

1. Use NVMe over USB — SD cards fail under write-heavy workloads
2. Set resource limits — Prevent a single pod from starving others
3. Enable pod disruption budgets — Ensure availability during updates
4. Back up etcd — K3s uses SQLite by default; consider external etcd for HA
5. Use private registries — Pull images from a local registry to save bandwidth

Build your own edge cluster in our Industrial IoT Architecture track.