Multi-Region Kubernetes at Scale: Lessons from Running Rancher and RKE2 in Production

Operating persistent workloads across globally distributed Kubernetes clusters is hard.

This talk dives into a realworld implementation of a "cluster of clusters" architecture built using opensource tools like Rancher, RKE2 and Helm

We will discuss -

1. Designing a “cluster of clusters” model with Rancher acting as a gcp(global control plane)
2. Simplifying operations with GitOps-driven templates and centralized policy management.
3. Configuring secure inter-cluster communication with VPC peering and VPN tunnels.
4. Automating cluster onboarding, upgrades and rollbacks with minimal downtime.
5. Addressing challenges like cross-region latency, HA configuration and team level access control.

EDIT -

From the author -

Hi review team,

Thank

you for the feedback on my CFP submission titled "Managing Multi-Region

Kubernetes Clusters with Rancher and RKE2" (link:

https://fossunited.org/c/indiafoss/2025/cfp/71vaqn1iin).

To address the feedback:

This talk is based on a real-world production use case where we had to support user-facing applications that serve users across multiple regions. These workloads are stateful, latency-sensitive, and require high availability, disaster recovery and data locality (due to regulations etc). A single region Kubernetes setup was not sufficient to meet these requirements.

The talk will begin by outlining these challenges in detail and then followed by how we addressed them using Rancher, RKE2 and other open-source tools.

I will focus on practical strategies for managing multi-region clusters, automating upgrades and handling failures in production.

Please let me know if you would like me to update the abstract on the portal and provide any additional documentation.

Thanks again for the helpful input