Your failover plan is a DNS record you've never tested.
Most teams' cross-region failover is a weighted record, a health check someone configured once, and hope. ZeusK8s gives every service one name across every cluster — health-checked from inside and outside, on the mesh and on public DNS — and reroutes when something actually breaks.
Route 53 and Cloud DNS are great — inside their own cloud.
The managed DNS and health-check products from AWS and Google are genuinely good. They're also how you stay locked in: the records, the health checks, and the failover logic all live in one provider's console, wired to that provider's infrastructure. The day you need traffic to move between GCP and AWS — or to your own hardware — is the day they stop helping.
Running your own global routing layer instead has traditionally meant specialist work, or an expensive third-party traffic manager sitting in front of everything.
Zeus does this for you. One routing layer across every cloud and every cluster you own, with automatic health monitoring and failover routing — using Route 53 or Cloud DNS underneath as plain registries, never as the brain. Public multi-region publish is part of the product today.
Mesh names and public multi-region DNS — same brain.
Internal (mesh)
Stable names for services and databases inside your fabric. Local-first when healthy; fail over when a cluster dies. Fail-closed when state is ambiguous. Last-known-good answers survive control-plane blips.
Public (internet)
Customer-facing hostnames published to Route53 and Cloud DNS. Weights and failover driven by live health and multi-region probers — Zeus decides; the cloud DNS products are registries, not the brain.
Related: fabric path for private traffic · domains for TLS and white-label.
One name for a service, no matter which cluster answers.
One name, every cluster
Give a service or database a single stable name across all of your clusters. Route by latency, by weight, or union the healthy targets across clusters — without hand-writing provider DNS rules.
Continuously proven on live health
Routing decisions are pushed when health changes — not waiting on tribal runbooks. Traffic moves on live verdicts from inside the cluster and from multi-region probers.
Health checks, inside and out
Local probes watch every target from inside the cluster. Independent probers in multiple regions verify public endpoints from the outside — so “up” means reachable, not just running.
Fails safe, not silent
When state is ambiguous, Zeus answers fail-closed instead of pointing traffic at a dead target. Each cluster keeps its last-known-good routing on disk, so resolution keeps working even if the control plane is unreachable.
Works with the fabric
The same routing layer covers private cross-cluster traffic on the encrypted fabric and public traffic on your domains. One mental model for both.
Domains & TLS included
Bring your domains, get automatic certificates and per-brand white-label DNS on top — same control plane, no extra tooling.
Related: global fabric · databases · architecture · domains, certificates & white-label
DNS failover either works by accident, or it's the mess nobody wants to touch.
- · Hand-written weighted records in a provider console someone set up once
- · Health checks that test the load balancer, not the application behind it
- · Failover that has never been exercised outside of an actual outage
- · Every change made carefully, because nobody is sure what depends on what
- · Endpoints, targets, and health policies defined in one place, visible to the whole team
- · Internal mesh names and public hostnames on the same health model
- · Health verdicts from inside the cluster and from independent outside probers
- · Failover that runs the same way every time — because it runs continuously