Julian Wiley

Mapping the 2026 RPi Kubernetes Service Expansion

April 25, 2026· 1 min readRPi Kubernetes

How the root kustomization in rpi_kubernetes evolved into a full multi-service platform spanning data, MLOps, and observability.

RPi KubernetesKustomizeHomelabMLOpsInfrastructure

The Big Shift

The most important file in this phase of rpi_kubernetes is still kubernetes/kustomization.yaml.

It now acts as a platform map, not just a deployment list. The resource set spans:

  • base data services (minio, postgresql, mlflow, chromadb, ragflow, datahub)
  • observability (otel-collector, jaeger, victoriametrics, prometheus)
  • MLOps (argo-workflows, dagster, pipelines, bentoml)

That is a meaningful jump from a basic homelab to a real local platform.

Why This Matters On Raspberry Pi

On constrained hardware, integration mistakes are expensive. A single source-of-truth kustomization reduces drift and makes service composition explicit.

It also clarifies what is intentionally out-of-band. The file comments call out Helm-managed components like Loki and kube-prometheus-stack, which prevents false assumptions when kubectl apply -k kubernetes/ appears successful but dashboards are incomplete.

Practical Operational Benefit

This explicit map improves:

  • onboarding for new contributors
  • debugging of missing dependencies
  • upgrade planning across service layers

I treat this file as a deployment contract: if a new platform capability is "real," it should be represented here or intentionally documented as external.

What I Learned

A growing cluster fails less from missing YAML and more from unclear boundaries. The combination of kustomize resources plus documented Helm exceptions is a good pattern for mixed deployment models.

Practical Takeaway

If your homelab is becoming a platform, design your root kustomization as an architecture index with explicit inclusion and exclusion rules. It scales better than organic service sprawl.

Related Posts