Skip to main content

Initial Release

· 2 min read
Jovica Zarić
Jovica Zarić
Founder at Infaloom

Today we are open-sourcing Nursery, the opinionated blueprint we use to run Infaloom's production workloads on a lean budget.

The project documents everything we've learned while standing up a six-node K3s cluster on Hetzner Cloud with Pulumi for provisioning, Ansible for day-two automation, and a security-first stance powered by OpenBao + External Secrets. The full stack comes in at EUR 187.34/month (November 2025) and is designed for small teams that need production reliability.

What you get in v1​

  • Core platform – Automated Hetzner bootstrap, HA control plane access via HAProxy, and Rook-Ceph backed storage across three 200 GB nodes for ~200 GB of triple replicated capacity.
  • Data services – CloudNativePG with scheduled S3-compatible off-site backups and Redis Sentinel (Bitnami legacy images) for stateful workloads.
  • App enablement – Harbor registry, ArgoCD GitOps, and ready-to-use Helm values for common building blocks.
  • Security + secrets – OpenBao as the source of truth with ExternalSecrets syncing into the cluster, plus documented ingress, TLS, and IP allowlist patterns.
  • Ops readiness – Step-by-step guides for monitoring, logging, disaster recovery, and everyday utilities, all pinned to tested tool versions for reproducibility.

Why it matters​

Managed Kubernetes is fantastic, but it quickly eats into the budget of bootstrapped startups. Nursery shows that you can own the stack end-to-end, keep costs predictable, and still follow best practices for backups, observability, and multi-environment automation.

Dive in​

  • Start with the introduction for motivation, audience, and cluster specs.
  • Follow the development environment and getting started guides to mirror our toolchain.
  • Explore the service deep dives (CNPG, Redis, Harbor, ArgoCD, storage, secrets, DR, etc.) in documentation and keep an eye on the roadmap for what is coming next.

Questions, ideas, or battle stories from your own clusters are very welcome — open an issue or start a discussion so we can keep improving the stack together.