Services, DNS and Ingress

Concept

Kubernetes provides several layers for exposing and discovering workloads:

• Service – stable virtual IP + DNS name in front of a set of Pods
• Cluster DNS (CoreDNS) – name resolution for Services and Pods
• Ingress – HTTP/HTTPS routing from outside the cluster to Services

Why it matters

Many “application is down” reports are actually:

• Service has no healthy endpoints
• DNS resolution failing inside the cluster
• Ingress controller or rules misconfigured
• Network policies blocking traffic

Mental Model

External Client
    ↓
Ingress Controller (optional)
    ↓
Service (ClusterIP / NodePort / LoadBalancer)
    ↓
Endpoints / EndpointSlices
    ↓
Pods

Inside the cluster, Pods talk to each other via Service DNS names:

my-service.my-namespace.svc.cluster.local

Key Commands

# Services and endpoints
kubectl get svc -A
kubectl get endpoints -A
kubectl describe svc <service> -n <ns>
 
# DNS debugging
kubectl get pods -n kube-system | grep coredns
kubectl run -it --rm debug --image=busybox --restart=Never -- nslookup my-service.my-namespace
 
# Ingress
kubectl get ingress -A
kubectl describe ingress <name> -n <ns>
 
# Test from inside the cluster
kubectl run -it --rm debug --image=curlimages/curl --restart=Never -- curl -v http://my-service.my-namespace

Common Failure Modes & Symptoms

Symptom	Likely cause	First checks
Service has no endpoints	Selector doesn’t match Pods / Pods not Ready	kubectl get endpoints, Pod labels & readiness
DNS name doesn’t resolve	CoreDNS issue or wrong name	CoreDNS pods, nslookup from a Pod
External traffic doesn’t reach app	Ingress / LoadBalancer / firewall	Ingress controller logs, Service type
Intermittent connectivity	Network policy, kube-proxy, CNI	Network policies, node health

Investigation Tips

• Always check Endpoints (or EndpointSlices) — a Service with zero endpoints will never work.
• Readiness probes directly affect whether a Pod is added to endpoints.
• Test DNS and connectivity from another Pod inside the cluster, not only from outside.
• Ingress problems are often in the controller logs or in the Ingress object’s events.

Related Notes

• Kubernetes Architecture
• Pod Troubleshooting
• TCP IP Troubleshooting Model
• DNS Resolution
• Troubleshooting Methodology

Personal Lessons Learned