Great questions! You're asking all the right things as someone starting with container orchestration. Let me break this down clearly:

Minikube vs Multi-Node Clusters

You're absolutely right - Minikube is single-node only. It's designed to run Kubernetes locally on your laptop for learning/development. For your 4-node Raspberry Pi cluster, Minikube won't work.

For Raspberry Pi clusters, you have better options:

• K3s (my top recommendation for Pi clusters)
• MicroK8s
• Full Kubernetes (but it's overkill and resource-heavy)

Docker Compose Multi-Host Question

Docker Compose alone cannot manage containers across different hosts. It's designed for single-host deployments. If you want to run containers on multiple Raspberry Pis with Docker Compose, you'd need separate compose files on each Pi - no automatic coordination between them.

For simple cross-host orchestration, you'd need something like Docker Swarm or K3s.

Docker Swarm vs Kubernetes

They're completely independent solutions that solve the same problem:

• Docker Swarm: Docker's built-in clustering (simpler but less powerful)
• Kubernetes: More complex but much more capable

You pick one or the other, not both.

Kubernetes on Raspberry Pi - Resource Usage

You heard correctly! Full Kubernetes is resource-hungry on Pi. A single-node Kubernetes cluster can easily consume 1GB+ RAM just for the control plane components, leaving little for your actual applications.

This is why I strongly recommend K3s for Raspberry Pi clusters.

K3s - Perfect for Raspberry Pi

K3s is lightweight Kubernetes that's perfect for your use case:

• Uses only ~40MB RAM for the control plane (vs 1GB+ for full K8s)
• Single binary installation
• Works great on ARM/Raspberry Pi
• Still gives you full Kubernetes API compatibility
• Designed exactly for edge/IoT scenarios like yours

Setup Recommendation for Your 4-Pi Cluster

On your main Pi (master node):

curl -sfL https://get.k3s.io | sh -

On the other 3 Pis (worker nodes):

curl -sfL https://get.k3s.io | K3S_URL=https://main-pi-ip:6443 K3S_TOKEN=your-token sh -

Alternative: Docker Swarm for Simplicity

If K3s feels too complex initially, Docker Swarm is simpler:

On main Pi:

docker swarm init

On other Pis:

docker swarm join --token :2377

Then you can deploy with docker stack deploy using compose files.

My Recommendation Path:

1. Start with Docker Swarm - easier to understand, good for learning clustering concepts
2. Move to K3s when you want more advanced features
3. Never use full Kubernetes on Pi - it's overkill and will eat your resources

Quick Comparison for Pi Clusters:

Start with K3s - it's specifically designed for scenarios like yours and will give you the best learning experience without killing your Pi's performance.

For a detailed comparison of lightweight Kubernetes options perfect for Pi clusters, check out: https://toolstac.com/alternatives/kubernetes/lightweight-orchestration-alternatives/lightweight-alternatives