Cloud sovereignty is not only about where workloads run or where data is stored. It is also about who controls reachability when infrastructure fails.

A platform may span multiple clouds, regions, or on-premises environments, yet still depend on a single provider’s load balancer, DNS service, or failover mechanism. That hidden dependency becomes critical precisely when resilience and operational autonomy matter most.

This talk explores how Kubernetes-native global traffic management can provide an open, provider-independent traffic and failover layer across clusters and infrastructure providers. Using k8gb, a CNCF Incubating project, and lessons from real-world adoption, we will walk through the architecture behind authoritative DNS-based failover, health checking, reconciliation, and traffic policies.

We will also examine the less glamorous operational realities: DNS delegation, TTLs and caching, split-brain scenarios, degraded workloads, observability, and testing failover before an actual incident.

Attendees will leave with a reusable architecture pattern for Kubernetes-native global failover, a practical checklist for operating it, and a clearer way to evaluate whether their multi-cluster platform is genuinely sovereign, or merely distributed across several providers.