The brief

A major public-transportation operator was running production workloads on Amazon EKS. New national regulation around personal data, combined with stricter GDPR enforcement, meant the platform had to run inside Norwegian borders, on hardware they controlled. The brief was easy to describe and harder to deliver, rebuild the platform on-premises without breaking the services running on top of it.

What I did

I was the primary Linux resource on a four-person Orange consulting team, coordinating and executing the migration end-to-end. The new cluster runs on Kubernetes, with Traefik for ingress and a per-node monitoring stack of Telegraf for metrics and Fluent Bit for logs. Anything that touched configuration went through Terraform and a GitOps delivery pipeline, so every change has a paper trail.

A few things mattered more than the choice of tools.

• Image security. All pulled images went through Harbor with vulnerability scanning before they could run on the cluster.
• Day-two operations. The customer's engineers wrote and owned their own workload manifests. My role was to review their Kubernetes patterns and onboard them to the platform, so the handoff was durable rather than dependent on me.
• Quiet cutover. The blue-green migration covered the two clusters that made up the internal developer platform (test and prod). Both ran in parallel until the new side was verified, then traffic switched. The remaining three clusters (for a separate critical application) stayed under operational management throughout.

Why it mattered

For the customer, this was not only about compliance. It was about control of the data, the cost curve and the failure modes. A 30-node, 960 GB fleet moved from a managed cloud provider to a self-operated Kubernetes platform on dedicated hardware, with customers depending on it every morning, and zero incidents across both parallel tracks.