The CNCF Technical Oversight Committee accepted K8gb as an incubating project on August 5. The CNCF announcement describes an open-source global server load-balancing system designed for traffic management and failover across Kubernetes clusters and regions.
DNS carries the cross-cluster decision
K8gb defines a GSLB custom resource and runs an operator that coordinates traffic strategies. It uses Pod liveness and readiness data, embeds CoreDNS for dynamic DNS responses, and works with ExternalDNS for zone delegation. Integrations connect the model to Ingress and Gateway API resources without binding the configuration to one controller. Crossplane support can combine infrastructure provisioning with declarative global routing.
The incubation milestone reflects CNCF’s assessment of project maturity and community health. It does not establish that any particular deployment meets a recovery-time or availability target. DNS caches, health-check intervals, record time-to-live, client behavior, and the failure domain of authoritative DNS all influence real failover.
Test the stale paths
A useful evaluation starts with two nonproduction clusters and a client population that resembles production resolvers. Operators should fail an application, a cluster endpoint, and a region-facing DNS path separately. Measure detection, record propagation, cached-answer expiry, and the time until successful requests reach the remaining cluster.
Teams should also decide who owns zones and delegation, how application teams request routing changes, and how configuration is recovered if both the operator and one cluster disappear. K8gb’s Kubernetes-native API can make global routing reviewable in version control. Its operational value depends on rehearsing the systems outside Kubernetes that still participate in every DNS-based failover.
- K8gb becomes a CNCF incubating project
Cloud Native Computing Foundation · Aug 5, 2026
See the original announcement for availability and release details.