Kubernetes on your own hardware: built, documented, rehearsed

We build Kubernetes platforms with Ceph storage and GitOps, on-premise on your hardware (bare metal), on Proxmox or in your cloud, and either run them with you or hand them over. They are for software companies without a platform team. The result is a documented platform whose restore has been measured before it becomes yours.

Request an intro call

Fixed-scope entry offer

Disaster Recovery Drill

For whom: Companies that have backups but have never measured how long a restore actually takes. A common trigger: a customer's security questionnaire asks for restore tests, or a recent outage was a close call.

Scope

  • One agreed system, one exercise: either a restore from your existing backups or a planned failover.
  • Typically a database and the application that uses it, on Kubernetes, on virtual machines, or on AWS, GCP or Azure.
  • Before we start, we agree with you what "back in service" means, when the clock starts and stops, and where the drill runs: in a separate environment or, with your sign-off, on the live system.
  • We run the drill together with your team and log every step with a timestamp.
  • Delivered remotely.

DeliverableA written report: what was restored, the measured times, the gaps we ran into, and a prioritised list of fixes. If you have set a recovery time objective, the measured time is set against it.

Prerequisites

  • Backups or replicas of the system already exist. The drill tests what is there.
  • Someone on your team who knows the system is available during the drill.
  • Access to the systems involved, for the drill.
  • For a drill on the live system: a maintenance window you have approved.

Not included

  • Further systems or a second exercise (for example a failover on top of the restore). We quote those separately.
  • Fixing the gaps we find. That is a project of its own, with us or without us.
  • Setting up backups where none exist yet.
  • A commitment on how fast your system will be back in a real incident. The drill measures where you stand today.
  • An assessment of whether NIS2 or the Cyber Resilience Act applies to your company. Whether and how NIS2 applies to you is a question for your legal advisers; we implement the technical measures.

Request this offerfrybad.de/en/contact/?thema=kubernetes-plattform&angebot=ernstfall-probe

Where teams usually start

  • The cloud bill rises with every customer. Load is steady, data keeps growing, and you wonder whether your own hardware would cost less.
  • A customer asks about restore tests. The security questionnaire wants to know when you last rehearsed a restore and how long it took.
  • A cluster nobody wants to touch. Upgrades are overdue, the person who set it up has left, and every change feels like a risk.

GitOps & CI/CD

Every change to the platform is a commit. Flux keeps the cluster in line with the Git repository, so what is in Git is what runs. Secrets sit in the same repository, encrypted with SOPS and age; they never appear in Git in plain text.

For code and pipelines we use GitLab, self-hosted on your platform if you want it there. CI builds and checks every image, gives it a version and rolls it out to a test cluster first. Only what ran there goes to production. That is how we run our own platform too.

Backup & recovery

A backup only counts once the restore has been measured. We run databases with CloudNativePG on Kubernetes, with replicas and automatic failover. Fortlauf’s database backups are kept, encrypted, at a second location in Germany; since 5 Oct 2026 they have been encrypted before they leave our hardware. And we rehearse both: restoring from a backup and switching to a replica. Two measurements from our own operations:

Restore rehearsal on 30 Aug 2026: we restored both of Fortlauf’s production databases (product data and sign-in) to coordinated points in time, in a separate test environment. Both were back in service 3 min 22 s after the restore configuration was applied; a repeat run the same day took about 3 min 30 s. Not included: the time to decide and prepare, and files kept outside the databases.

Planned database failover on 4 Oct 2026: we switched Fortlauf’s product database to another replica while the service was running. The database cluster was fully back in service 19 s after the switch; writes paused for about 8 to 9 s, and no external request received a server error in that time. Not included: an unplanned outage.

The Disaster Recovery Drill above runs the same measurement on a system of your choice. More on location, encryption and backups: Where your data lives.

How we work

  1. Assessment.

    What runs where, who operates it, where it hurts. You get a short written finding.

  2. Target design.

    Kubernetes, storage (Ceph or what you already have), network, GitOps, backups and monitoring. You get an architecture document with the decisions and the reasons for them.

  3. Build.

    Everything as code in a Git repository, rolled out by GitOps, to a test environment first and then to production.

  4. Handover.

    Runbooks, a walkthrough for your team, and a measured restore rehearsal before sign-off. After that you run the platform yourselves, or we run it within the hours agreed in the contract.

What we run ourselves

We run our own hardware in a data centre in Munich.

Our product Fortlauf runs on this too. Moving off VMware? See VMware to Proxmox migration. For training and inference workloads on Kubernetes: GPU servers in Germany.

Stack

  • Virtualisation with Proxmox VE; our Kubernetes clusters run on it, with production and development kept apart.
  • Ceph storage, for Kubernetes volumes and as S3-compatible object storage.
  • GitOps with Flux, secrets with SOPS and age, code and CI on a self-hosted GitLab.
  • PostgreSQL with CloudNativePG.
  • Monitoring with VictoriaMetrics, Grafana and Loki.

Case studies

Our own product

Case study: running a SaaS on our own infrastructure

Fortlauf is our product. We built it, and we run it on our own hardware. In production since July 2026: tenant isolation in the database, immutable document storage, delivery by GitOps, measured recovery.

Company with its own Ceph cluster

Case study: a client Ceph cluster encrypted, zero downtime

A customer wanted its Ceph cluster encrypted after the migration. We encrypted every OSD while the cluster stayed in service, with zero downtime, and handed over secrets, documentation and runbooks.

Who you will talk to

One fixed contact from the intro call to the handover.

Send a message

What we do not do

  • No on-call cover. Operations run in agreed hours.
  • No resale of cloud capacity or licences.
  • No platform without a handover: what we build is documented, and your team can run it themselves.

Frequently asked questions

Is our own hardware worth it compared with EKS or GKE?

It depends on your load. Steady base load and large data volumes tend to favour your own hardware; sharp peaks and short projects tend to favour the cloud. We work through it with your numbers before you commit. If you stay on AWS, GCP or Azure, we build the platform there.

Who runs the platform after the handover?

Your team, with runbooks and a walkthrough. Or we do, within the hours agreed in the contract.

Do we need Ceph, or is our existing storage enough?

Not every platform needs Ceph. It pays off when several hosts need shared storage that survives a host failure and you do not want to buy a central storage system. If your existing storage is enough, we connect it. We decide this in the target design.

How is recovery proven?

By a rehearsal, not by a tick in the backup tool. We restore an agreed system or switch it to a replica, time every step and write up the result. As a separate offer, this is the Disaster Recovery Drill.

Can you build on our Proxmox environment?

Yes. Our own Kubernetes clusters run on Proxmox VE. If you are still on VMware, see VMware to Proxmox migration.

Insights

Want to talk about your platform?

Tell us what runs where today and what needs to change.