Migrating a legacy estate and bringing Kubernetes on-premise

Context

The group runs mobility, home, health and legal assistance services in France and abroad. A large part of its application estate ran on operating system versions reaching end of support, and its container strategy stopped at the public cloud: workloads that had to stay on-premise had no modern platform.

Approach

Our team led DevOps on both fronts.

• Obsolescence programme. Analysed the technical base and dependencies of each application, defined a migration plan per application, automated the migration tasks with Ansible and moved applications on Red Hat, SLES and Windows Server across every environment.
• On-premise Kubernetes. Built a proof of concept, then a Rancher and RKE2 sandbox; defined the architecture per environment type with the architects; validated it with the system, virtualisation and network teams; extended the CI/CD chain to on-premise clusters and adapted the AKS services so they run on both.
• Build-to-run. Hardened the factory with Ansible-managed application profiles, level 2 and 3 support on AKS and Azure, Dynatrace dashboards for FinOps, and Kubernetes workshops for the teams.
• HR systems. Set up the delivery chain for a critical HR application, investigated overnight batch incidents and applied a rightsizing plan across its environments.

Results

• About a hundred applications and forty virtual machines moved off obsolete systems, with a repeatable, automated method.
• One delivery chain deploying to Azure Kubernetes and on-premise Kubernetes alike.
• Operations and FinOps teams equipped with shared dashboards, and teams trained to run Kubernetes themselves.