A Multi-Cluster, Multi-Cloud Infrastructure with GitOps at CERN

Explore a conference talk detailing CERN's implementation of a multi-cluster, multi-cloud infrastructure using GitOps principles. Discover how CERN manages its distributed infrastructure to handle petabytes of experimental data annually, leveraging Kubernetes, cloud-native technologies, and public cloud resources. Learn about the evolution of their deployment strategies, incorporating GitOps and tools like Flux and ArgoCD for centralized and standardized management. Gain insights into the three-layer deployment approach, including the "underlay" infrastructure, base services, and end-user services. Understand how CERN utilizes tools such as Crossplane for efficient cluster management across on-premises and multiple public cloud regions. Explore topics like monitoring with Prometheus, OPA policies, cost calculation, and the integration of machine learning workflows. Witness a demonstration of rapid cluster addition and removal, and grasp the complexities of managing a large-scale, distributed scientific computing environment.

Introduction
Why Multiple Clusters
Clusters as Cattle
Release Channels
Umbrella Homechart
GitOps Training
Public Cloud
Deployment
Machine Learning
Public Cloud Stack
Components
RCD
Deployment at CERN
Infrastructure layer
Conclusion
Questions