Case Studies

Two recent client projects about reliable operations, reproducible infrastructure and recovery. Both clients are anonymised.

From a single server to an automated Hetzner Cloud platform Link to heading

Company running an online photo platform · 2025–2026

Context and problem Link to heading

A company running an online photo platform needed to move it from Azure to Hetzner Cloud. The existing setup depended on a single server. The migration was also the chance to make the platform more resilient and to build its infrastructure reproducibly.

Constraints and risks Link to heading

GDPR obligations and data-location requirements shaped the infrastructure decisions. The photo service had to stay available with as little interruption as possible during the migration, and the new setup had to cover application availability, database access and recovery.

Approach and engineering decisions Link to heading

I built the platform around load-balanced application nodes, with private networking between the application and the database. I wrote a MySQL availability design and a recovery plan.

OpenTofu and Ansible make the infrastructure and its configuration reproducible. CI/CD and TLS are part of how the platform is delivered and operated.

What was implemented Link to heading

  • Migration from Azure to Hetzner Cloud.
  • Load-balanced application nodes and private networking between application and database.
  • Infrastructure provisioning with OpenTofu and configuration automation with Ansible.
  • CI/CD, TLS and monitoring.

Outcome and lessons Link to heading

The platform moved to Hetzner Cloud with minimal interruption. Load-balanced application nodes reduced the reliance on a single application server, and automation made the environment reproducible and easier to maintain.

Application resilience, database availability and recovery have to be planned together. More application nodes solve only part of the availability problem; the database design and the recovery plan remain central to running the service.

Relevant technologies: Azure, Hetzner Cloud, OpenTofu, Ansible, MySQL, CI/CD, TLS, private networking and monitoring.

Building an encrypted and recoverable private photo cloud Link to heading

Private client

Context and problem Link to heading

A private client wanted a self-hosted photo platform that kept their data under their own control. The project covered the Immich deployment and everything needed to operate it: encryption, access controls, backups and recovery documentation.

Constraints and risks Link to heading

Media and backups had to be encrypted. Access to the service had to stay restricted, with a controlled route for remote administration. Recovery had to cover both the photo files and the application database.

Approach and engineering decisions Link to heading

I deployed Immich with Docker on Ubuntu. Media and backups are stored on encrypted drives; the service is protected by TLS and reachable only on the local network. Tailscale provides remote administration, and Samba provides local file access.

I treated backups, database dumps, storage monitoring and recovery documentation as part of the deployment, not as later additions.

What was implemented Link to heading

  • An Immich deployment with Docker on Ubuntu.
  • Encrypted media and backup drives.
  • Remote administration through Tailscale and local file access through Samba.
  • Automated backups and database dumps.
  • Storage monitoring and alerts.
  • TLS and access restricted to the local network.
  • A documented recovery procedure.

Outcome and lessons Link to heading

The client now has a private photo platform with encrypted storage, controlled access, automated backups and documented recovery steps.

A recoverable photo service needs more than a copy of its media files. Protecting the database and writing clear recovery instructions are part of maintaining the application and the data it manages.

Relevant technologies: Ubuntu, Docker, Immich, disk encryption, Tailscale, Samba, TLS, automated backups and monitoring.

Public engineering lab Link to heading

This earlier public lab shows how I approach development environments and application delivery.

Local Kubernetes with TLS Link to heading

Public development lab · 2025

An automated lab with Kind (Kubernetes in Docker), HTTPS endpoints and certificate validation. The published walkthrough shows HTTPS responses and successful certificate verification locally; it does not reproduce every production constraint.

Read the TLS lab write-up

Working on a similar problem? Link to heading

Discuss your project or explore my services.