When compliance meets sovereignty.

Körber Pharma Software modernises its OpenStack platform on SCS together with B1 Systems GmbH.

Technical Overview:

The migration was carried out during ongoing operations based on the following architecture:

  • New platform: HARALD (SCS-compliant, OpenStack via OSISM)
  • 6,000+ instances, 100 compute nodes, 45,000 vCPUs, 104 TB RAM
  • Network: Spine-leaf topology without a spine, two independent fabrics, OSPF + BFD for multipathing
  • Storage: Nimble iSCSI (~15,000 volumes), Ceph HCI currently being set up
  • Automation: Python scripts, phased migration on a project-by-project basis
  • Test environment: GEMINI
  • GMP compliance: VM metadata verifiably unchanged following migration

This approach is particularly suitable for companies in regulated industries that need to modernise an existing OpenStack environment without disrupting ongoing operations.

Created using:

Certified integrator

Background

The Körber Pharma Software GmbH Körber Pharma Software GmbH is part of Körber AG, which has around 13,000 employees at more than 100 locations worldwide. With around 1,000 employees of its own, the company develops software for the documentation and pro- duction of pharmaceuticals – software that runs in the background but is mission-critical. The internal OpenStack cloud, with around 400 to 500 active users, is the backbone of day-to-day work. But the platform it ran on had reached the end of its life – and a migration without interrupting operations was inevitable.

The existing setup was based on TripleO, the lifecycle management tooling for OpenStack primarily driven by Red Hat, which also underpinned Red Hat OpenStack Platform. Version 16.x was in use, based on the open-source components. In practice, this model brought increasing constraints: deployments across 100 compute nodes took up to 17 hours, rollback was not supported, and the complexity of the system made troubleshooting and further development significantly harder.

The decisive turning point came with Red Hat’s strategic decision to discontinue TripleO in favour of OpenStack Services on OpenShift: further upgrades became unavailable, and access to modern OpenStack API features could no longer be guaranteed.

Procedure

Replacing the hardware was not an option. Körber Pharma Software wanted to continue using the existing compute nodes, network infrastructure, and Nimble storage with around 15,000 volumes in total – a complete data transfer was therefore off the table. Together with B1 Systems GmbH B1 Systems – a certified SCS IaaS Integrator (Silver) and OpenStack specialist with more than 150 experts and experience since 2011 – Körber Pharma Software developed a migration concept that met exactly these requirements.

The target was an SCS-compliant platform: open, certifiable standards and a modular open-source stack make it independent of individual vendors’ product decisions – the key lesson from the end of TripleO. OSISM is used for deployment and lifecycle management.

An additional critical factor came into play: as a pharmaceutical software provider, Körber Pharma Software is subject to the requirements of Good Manufacturing Practice (GMP). This meant that every VM had to carry exactly the same metadata after the migration as before – chipset type, enabled flags, database entries. What had been qualified under the digital four-eyes (dual-control) principle had to remain demonstrably identical. Full automation was therefore not a nice-to-have but a prerequisite for reliably meeting GMP requirements.

Result

The new platform HARALD is now in production. Looking back, three factors proved particularly decisive: the deliber- ate decision to prioritise quality over speed and not to set an external deadline; the consistent use of the GEMINI test environment, in which all migration and update scenarios were validated in advance; and the full automation of the migration process, without which a reliable transfer of around 15,000 volumes and more than 6,000 instances would not have been achievable at the required quality or within a reasonable timeframe.

Equally important were the in-depth expertise of the engineers at B1 Systems GmbH and the rapid provision of targeted bug fixes – by the B1 team upstream in OpenStack and by OSISM GmbH in OSISM.

The advantages of the new platform are clearly evident in day-to-day operations: a major OpenStack upgrade used to take 48 to 72 hours with TripleO – accompanied by hours-long outages that regularly had to be scheduled for weekends. With SCS and OSISM, the same upgrade can be carried out with API outages of just seconds within a single morning – in the middle of the working week.

Want more details?

Download the full use case with all the background and information!

Further use cases