Kubernetes · Orchestration under pressure

How does a Kubernetes Service route traffic to pods?

EasyservicesnetworkingendpointsUpdated Oct 2026

Short answer

A Service gives a stable virtual IP and DNS name to a set of pods selected by labels. EndpointSlices track ready pod IPs and the dataplane forwards traffic to them.

Pods come and go, so clients cannot rely on pod IPs. A Service selects pods with a label selector and gets a stable ClusterIP. The control plane maintains EndpointSlices containing the IPs of matching, ready pods, and kube-proxy or an eBPF dataplane programs the forwarding rules.

Services select pods; they do not create them
Only ready pods are normally included in endpoints
The default protocol is TCP; UDP and SCTP are also supported
Session affinity is optional and can skew load distribution
A headless Service returns pod IPs directly for client-side discovery

yaml

apiVersion: v1
kind: Service
metadata:
  name: api
spec:
  selector:
    app: api
  ports:
    - name: http
      port: 80
      targetPort: 8080

Clients call api:80; traffic is forwarded to port 8080 on ready pods.

The most common Service bug is a selector mismatch. If the Service has no endpoints, compare its selector with pod labels and confirm the pods are ready. A Service with zero endpoints will typically reject or time out connections rather than return a helpful error.

A Service load balances connections, not requests. Long-lived HTTP/2 or gRPC connections can concentrate load on one pod, so use client-side balancing or a service mesh when that matters.

Follow-ups you should expect

Interviewers use these to check whether the first answer was understanding or memorisation.

1What happens if a Service selector matches no pods?
2Why do only ready pods appear in EndpointSlices?
3What is a headless Service used for?
4How would you debug a Service that times out?

Say it in the interview

A Service is a stable virtual IP and DNS name in front of a changing set of pods. It uses labels to select pods, and EndpointSlices track which pod IPs are ready. The dataplane then forwards traffic to those endpoints. When a Service fails, I check selector matches, endpoint readiness and target ports before looking at the application.

Use this as a model, not a script. Saying it in your own words is what makes it survive follow-up pressure.

Now say it out loud.

Understanding an answer and delivering it under pressure are different skills. Rehearse this topic in a live Kubernetes mock interview and get scored feedback on depth, clarity and confidence.

Start a free Kubernetes interview Browse all questions

Follow-ups you should expect

Related questions

Now say it out loud.