// the find
k8gb-io/k8gb
A cloud native Kubernetes Global Balancer
k8gb is a Kubernetes operator that handles global server load balancing (GSLB) across multiple geographically dispersed clusters using DNS. It replaces proprietary hardware solutions like F5 or Infoblox GSLB with a single CRD and works with whatever ingress controller you already have. Aimed at platform teams running multi-region Kubernetes who need failover or weighted routing without buying appliances.
DNS-based approach is genuinely the right call for global load balancing — it's stateless, works across any network boundary, and TTL-based failover is predictable. No dedicated management cluster means no single point of failure baked into the architecture itself. Support matrix is wide: NGINX, Istio, Gateway API (HTTP, gRPC, TCP, TLS, UDP routes all have separate resolvers), and six EdgeDNS providers. The Terratest + Chainsaw E2E suite against real multi-cluster k3d setups is significantly better than most operator test strategies.
DNS TTL is a ceiling on failover speed — if your EdgeDNS provider caches aggressively or clients ignore TTLs, you'll see stale traffic long after a cluster goes down, and the operator can't do anything about it. Still in v1beta1 after years of development, so the API can change under you. The local quick-start spins up two k3d clusters plus CoreDNS config surgery, which means the gap between 'it works on my laptop' and 'I understand what's happening in production' is steep. The refresolver abstraction for Gateway API routes has ballooned into six nearly-identical packages (HTTP, gRPC, TCP, TLS, TLS v2, UDP) that will drift out of sync as the Gateway API spec evolves.