EKS Networking Troubleshooting Triage

EKS Networking Troubleshooting Triage A workflow diagram generated by Archify. 01 / Symptom triage 02 / Diagnosis and remediation Network fault · Timeouts · Symptom triage Network fault Timeouts Pod networking · IP · comms · DNS · Symptom triage Pod networking IP · comms · DNS Service · LB · LB · Ingress · Symptom triage Service · LB LB · Ingress VPC · subnets · IP · routing · Symptom triage VPC · subnets IP · routing Deep diagnostics · netshoot · Diagnosis and remediation Deep diagnostics netshoot Confirm · fix · Fix · restart · Diagnosis and remediation Confirm · fix Fix · restart pods first not this not this go deeper act Legend Cluster layer Diagnostic tools AWS layer Symptom

Pod IP allocation fails

  • • Symptom: pods stay in ContainerCreating with an IP allocation failure message
  • • Cause: the node is out of free IPs, or it hit the ENI limit
  • • Fix: raise WARM_IP_TARGET, move to a larger instance type, or use custom networking
  • • Diagnose: kubectl describe node · kubectl logs -n kube-system -l k8s-app=aws-node

Pod-to-pod traffic and DNS

  • • Symptom: pod-to-pod connections fail, time out, or service names do not resolve
  • • Cause: a network policy blocks it, a security group rule, or CoreDNS trouble
  • • Diagnose: kubectl get networkpolicy, then run ping and nslookup through kubectl exec
  • • Check CoreDNS pod status and logs together

Load balancers and tooling

  • • For load balancer creation failures, check subnet tags, security groups, and IAM permissions first
  • • Required tags: kubernetes.io/role/elb=1 (public), kubernetes.io/role/internal-elb=1 (private)
  • • Run tcpdump and traceroute from a netshoot pod, then confirm the path with VPC Flow Logs