Kubernetes Service has no endpoints
A Kubernetes Service currently has no ready backend addresses to receive traffic, even though its virtual IP and DNS name may still exist.
Endpoints: <none>Also searched as: no endpoints · EndpointSlice empty
What it means
The Service selector may match no Pods, matching Pods may be unready, or a selectorless/external endpoint definition may be absent. The Service virtual IP can exist while every request still has nowhere to go.
Typical trigger conditions
- Selector does not match Pod labels
- All matching Pods fail readiness
- Wrong namespace
- Manual EndpointSlice missing or malformed
First diagnostic checks
Inspect selector and endpoints
Compare the Service selector with EndpointSlices generated for it.
kubectl describe service <service> -n <namespace>; kubectl get endpointslices -n <namespace> -l kubernetes.io/service-name=<service>
List matching Pods
Apply the exact selector and inspect readiness rather than listing all Pods.
kubectl get pods -n <namespace> -l '<selector>' -o wide
Trace readiness failures
A matching but unready Pod is intentionally excluded from normal Service endpoints.
Version and platform notes
EndpointSlice is the current scaling mechanism; legacy Endpoints behaviour and deprecation status depend on Kubernetes version.
Related engineering templates
Follow Kubernetes failures and fixes
A useful digest about twice a week with new error references, tools, integration notes and production lessons. No daily noise.