Private Cloud Infrastructure
Self-hosted Kubernetes cluster powering storage, compute, CI/CD, observability, and distributed services across ARM64 homelab nodes.
Requests flow through ingress controllers into Kubernetes workloads, backed by distributed storage, databases, and full observability stacks.
Building Private Cloud Infrastructure
Designing and operating a self-hosted Kubernetes cluster with distributed storage, networking, observability, CI/CD automation, and AI workloads on ARM64 nodes.
Core Principles
- Everything runs as a containerized workload — no bare-metal service installs
- Infrastructure is declarative (GitOps-first) — config is code, reviewed like code
- Failure is expected and designed for — rescheduling, retries, and circuit breakers by default
- Systems must be observable from day one — metrics, logs, and health checks before launch
- Edge and local compute reduce cloud dependency, cost, and latency for internal workloads
Infrastructure Failure Modes
- Node failure without pod rescheduling policies — workloads go dark silently
- Storage saturation across persistent volumes causing write failures mid-operation
- Network bottlenecks between services under burst load without pod-level limits
- Silent service degradation with no alerting — issues surface hours late
- Cluster resource fragmentation preventing scheduling of new pods despite available nodes
Infrastructure Layers
Compute Cluster
k3s on 8 Raspberry Pi 5 ARM64 nodes. Workload scheduling, resource quotas, and pod autoscaling.
Networking
Ingress NGINX for cluster-edge routing. Internal DNS, service mesh patterns, and cert management.
Storage
MinIO S3-compatible object storage for assets and model weights. Persistent volumes for stateful workloads.
CI/CD
Self-hosted Gitea with Actions runners. Automated build, test, and deploy pipelines triggered on push.
Observability Stack
Metrics
Prometheus scrapes node, container, and application metrics. Grafana dashboards surface CPU, memory, queue depth, and inference latency.
Logs
Centralized structured logging across all containers. Correlation IDs thread through API, worker, and AI service logs.
Tracing
Request tracing across distributed services. Health check endpoints polled by the load balancer per dependency.
System Architecture Diagram
Loading diagram…
Infrastructure Stack
Frequently Asked Questions
Power Consumption Tradeoffs
Dec 18, 2025
Balancing power consumption and performance in homelab clusters.
Gitea Runners Pipeline
Nov 15, 2025
CI/CD pipeline flow using Gitea runners in my homelab.
Cluster Architecture (Homelab)
Oct 26, 2025
Architecture of my homelab Kubernetes cluster.
ARM64 Considerations
Oct 24, 2025
Considerations for running infrastructure on ARM64 hardware.
Monitoring and Troubleshooting
Oct 18, 2025
Monitoring and troubleshooting a Raspberry Pi Kubernetes cluster.
Ingress Setup
Oct 13, 2025
Setting up ingress for a K3s Raspberry Pi cluster.
Storage Setup
Oct 10, 2025
Configuring storage for a Raspberry Pi Kubernetes cluster.
K3s on Raspberry Pis
Sep 7, 2025
Guide to setting up K3s on Raspberry Pi clusters.