Kubernetes Debugging

Kubernetes Debugging

Diagnose and troubleshoot Kubernetes workloads using safe kubectl inspection commands. Fast incident response for pod crashes, scheduling failures, and service issues.

What I Do

• Interpret pod statuses and identify root causes (OOMKilled, CrashLoopBackOff, ImagePullBackOff)
• Extract and analyze container logs including previous crash logs
• Debug scheduling failures, resource constraints, and node issues
• Inspect network connectivity between pods and services
• Use ephemeral debug containers for distroless/crashed pods

When to Use Me

• Pod stuck in CrashLoopBackOff, ImagePullBackOff, Pending, or Evicted
• Container exits with OOMKilled or non-zero exit code
• Deployment rollout hangs or fails
• Service endpoints missing or not routing traffic
• Debugging distroless or crashed containers

Quick Start

kubectl get pods -o wide                    # Pod status
kubectl describe pod <name>                 # Details + events
kubectl logs <name>                         # Container logs
kubectl logs <name> --previous              # Crash logs
kubectl get events --sort-by='.metadata.creationTimestamp'  # Recent events (reliable)

Context and Namespace

# Check current context
kubectl config current-context
kubectl config get-contexts

# Switch namespace
kubectl config set-context --current --namespace=my-ns

# Explicit namespace (safer)
kubectl get pods -n production
kubectl logs -n production pod-name

Diagnostic Commands

Logs

kubectl logs <pod> -c <container>           # Specific container
kubectl logs <pod> --previous               # Previous crash
kubectl logs -f <pod>                       # Follow real-time
kubectl logs --tail=100 --since=1h <pod>    # Recent lines

Interactive Shell

kubectl exec -it <pod> -- /bin/sh
kubectl exec -it <pod> -c <container> -- /bin/sh
kubectl exec <pod> -- cat /etc/config

Ephemeral Debug Containers

# Add debug container to running pod (in-place, requires shareProcessNamespace)
kubectl debug -it <pod> --image=busybox --target=<container>

# Create a copy of the pod with a debug container attached
kubectl debug <pod> --copy-to=debug-pod --container=<container> --target=<container> --image=<image>
# Example:
kubectl debug myapp-7d9f8 --copy-to=debug-pod \
  --container=app \
  --target=app \
  --image=nicolaka/netshoot -- bash

# Debug node (host filesystem at /host)
kubectl debug node/<node> -it --image=busybox

The --copy-to form is the recommended pattern for distroless or crashed containers — it spins up a brand-new pod sharing the original's volumes, namespace, and labels, so you can attach a shell-capable debug image without modifying the running pod.

Resources

kubectl top pods --containers
kubectl top nodes

Pod Status Interpretation

Common Failure Modes

CrashLoopBackOff

kubectl describe pod <name>           # Exit code in events
kubectl logs <name> --previous        # Crash logs

Causes: App error, missing config, dependency unavailable, OOM

ImagePullBackOff

kubectl describe pod <name>           # Error in events

Causes: Typo, image doesn't exist, missing imagePullSecret

Pending Pod

kubectl describe pod <name>           # Scheduler events
kubectl describe nodes                # Available resources

Causes: Insufficient resources, taints, PVC pending, quota exceeded

Service Not Working

kubectl get endpoints <service>       # Pod IPs registered?
kubectl describe svc <service>        # Verify selector

Networking Triage

# Check network policies
kubectl get netpol -A

# DNS debugging (from inside pod)
kubectl exec -it <pod> -- nslookup kubernetes.default
kubectl exec -it <pod> -- cat /etc/resolv.conf

# Service endpoints
kubectl get endpoints <service>
kubectl describe svc <service>

# Using netshoot for network debugging
kubectl debug <pod> -it --image=nicolaka/netshoot --target=<container>

Rollout Status

kubectl rollout status deployment/<name>
kubectl rollout history deployment/<name>
kubectl get replicasets -l app=<name>
kubectl describe rs <replicaset-name>

Quick Decision Matrix

Context7 Integration

Use Context7 MCP server for up-to-date kubectl documentation:

• Latest kubectl command syntax
• New debugging features
• Ephemeral container options

Debugging Checklist

Tier 1 (30s): get pods -> describe pod -> logs Tier 2 (2-5m): logs --previous -> get events -> top pods -> exec Tier 3: kubectl debug -> full YAML review -> network policies

Common Errors

Useful Debug Images

• busybox - Lightweight shell
• nicolaka/netshoot - Network tools (curl, dig, nmap)
• curlimages/curl - HTTP testing

See references/research.md for detailed examples and advanced patterns.

Related Skills

• gke-deployment - GKE-specific patterns
• helm-charts - Helm release debugging

Resources

• Debug Pods
• Debug Running Pods
• kubectl debug