Amazon EKS · Troubleshooting Playbooks
EKS Troubleshooting
For when something is broken right now. Each playbook starts from the exact error or status you see, walks a decision tree to the cause, and gives you the commands to confirm it, the fix, and how to keep it from coming back.
Want the why first? Each cause links to the Visual Walkthrough that explains it, and the EKS Concepts page lists them all in learning order.
| Problem | What you see | Covers | Playbook |
|---|---|---|---|
| Pod stuck in Pending | 0/3 nodes are available: 3 Insufficient cpu. |
Resources, taints, affinity and spread, volumes, max Pods, cordoned or missing nodes, and Pods still starting. | Open |
| CrashLoopBackOff and OOMKilled | Back-off restarting failed container |
App errors, bad config, failing liveness probes, and containers killed for using too much memory. | Open |
| ImagePullBackOff | Failed to pull image |
Wrong image names or tags, ECR permissions, private registries, and nodes that can't reach the registry. | Open |
| Node NotReady or won't join | NotReady |
Nodes that never register, lose the kubelet, or go NotReady: IAM, networking and resource pressure. | Open |
| Service not reachable | connection refused |
Selectors that match nothing, Pods that aren't ready, wrong ports, Network Policies and Pod security groups. | Open |
| DNS failures | no such host |
CoreDNS health and scaling, search paths and ndots, and VPC resolver limits. | Open |
| ALB 5xx and unhealthy targets | 502 Bad Gateway |
Health checks, target types, security groups, and Pods that drop connections during rollouts. | Open |
| Out of Pod IPs | failed to assign an IP address to container |
Subnet exhaustion, ENI and IP limits per instance type, warm pools and prefix delegation. | Open |