Kubernetes Operations

Comprehensive kubectl assistance for debugging, resource management, and cluster operations with token-efficient scripts.

BEFORE YOU START

This skill prevents 5 common errors and saves ~70% tokens.

Known Issues This Skill Prevents

1. Running kubectl commands in wrong namespace/context
2. Verbose output flooding context with unnecessary data
3. Missing critical debugging steps (events, previous logs)
4. Exposing secrets in plain text output
5. Destructive operations without dry-run verification

Quick Start

Step 1: Verify Context

kubectl config current-context
kubectl config get-contexts

Why this matters: Running commands in the wrong cluster can cause production incidents.

Step 2: Debug a Pod

uv run scripts/debug_pod.py <pod-name> [-n namespace]

Why this matters: The script combines describe, logs, and events into a condensed summary, saving ~800 tokens.

Step 3: Check Cluster Health

uv run scripts/cluster_health.py

Why this matters: Quick overview of node status and unhealthy pods without verbose output.

Critical Rules

Always Do

• Always verify kubectl config current-context before operations
• Always use -n namespace to be explicit about target
• Always use --dry-run=client -o yaml before applying changes
• Always check events when debugging: kubectl get events --sort-by='.lastTimestamp'
• Always use --previous flag when pod is in CrashLoopBackOff

Never Do

• Never run kubectl delete without --dry-run first in production
• Never output secrets without filtering: avoid kubectl get secret -o yaml
• Never assume default namespace - always specify -n
• Never ignore resource limits when debugging OOMKilled pods
• Never skip describe when logs show no errors

Common Mistakes

Wrong:

kubectl logs my-pod

Correct:

kubectl logs my-pod -n my-namespace --tail=100 --timestamps

Why: Default namespace may not be correct, unlimited logs flood context, timestamps help correlate with events.

Known Issues Prevention

Debugging Workflows

Pod Not Starting

# 1. Get pod status and events
kubectl describe pod <name> -n <namespace>

# 2. Check logs (current or previous)
kubectl logs <name> -n <namespace> --tail=100
kubectl logs <name> -n <namespace> --previous  # If restarting

# 3. Check events for scheduling issues
kubectl get events -n <namespace> --sort-by='.lastTimestamp' | grep <name>

# 4. Interactive debugging
kubectl exec -it <name> -n <namespace> -- /bin/sh

Service Connectivity

# 1. Verify service exists and has endpoints
kubectl get svc <name> -n <namespace>
kubectl get endpoints <name> -n <namespace>

# 2. Check pod labels match service selector
kubectl get pods -n <namespace> --show-labels

# 3. Test from within cluster
kubectl run debug --rm -it --image=busybox -- wget -qO- http://<service>:<port>

# 4. Port-forward for local testing
kubectl port-forward svc/<name> 8080:80 -n <namespace>

Resource Management

Deployments

# List deployments
kubectl get deployments -n <namespace>

# Scale
kubectl scale deployment <name> --replicas=3 -n <namespace>

# Rollout status
kubectl rollout status deployment/<name> -n <namespace>

# Rollback
kubectl rollout undo deployment/<name> -n <namespace>

# History
kubectl rollout history deployment/<name> -n <namespace>

ConfigMaps and Secrets

# List
kubectl get configmaps -n <namespace>
kubectl get secrets -n <namespace>

# View ConfigMap data
kubectl get configmap <name> -n <namespace> -o jsonpath='{.data}'

# View Secret keys (NOT values)
kubectl get secret <name> -n <namespace> -o jsonpath='{.data}' | jq 'keys'

# Create from file
kubectl create configmap <name> --from-file=<path> -n <namespace> --dry-run=client -o yaml

Cluster Operations

Node Management

# List nodes with status
kubectl get nodes -o wide

# Node details
kubectl describe node <name>

# Cordon (prevent scheduling)
kubectl cordon <node>

# Drain (evict pods)
kubectl drain <node> --ignore-daemonsets --delete-emptydir-data

# Uncordon
kubectl uncordon <node>

Resource Usage

# Node resources
kubectl top nodes

# Pod resources
kubectl top pods -n <namespace>

# Sort by memory
kubectl top pods -n <namespace> --sort-by=memory

Bundled Resources

Scripts

Located in scripts/:

• debug_pod.py - Comprehensive pod debugging with condensed output
• get_resources.py - Resource summary using jsonpath for minimal tokens
• cluster_health.py - Quick cluster status overview

References

Located in references/:

• kubectl-cheatsheet.md - Condensed command reference
• jsonpath-patterns.md - Common JSONPath expressions
• debugging-flowchart.md - Decision tree for pod issues

Note: For deep dives on specific topics, see the reference files above.

Dependencies

Required

Optional

Official Documentation

• kubectl Quick Reference
• JSONPath Support
• kubectl Cheat Sheet
• Debug Running Pods

Troubleshooting

kubectl command not found

Symptoms: command not found: kubectl

Solution:

# macOS
brew install kubectl

# Verify
kubectl version --client

Context not set

Symptoms: error: no context is currently set

Solution:

# List available contexts
kubectl config get-contexts

# Set context
kubectl config use-context <context-name>

Permission denied

Symptoms: Error from server (Forbidden)

Solution:

# Check current user
kubectl auth whoami

# Check permissions
kubectl auth can-i get pods -n <namespace>
kubectl auth can-i --list -n <namespace>

Timeout connecting to cluster

Symptoms: Unable to connect to the server: dial tcp: i/o timeout

Solution:

# Check cluster endpoint
kubectl cluster-info

# Verify network connectivity
curl -k https://<cluster-api-endpoint>/healthz

# Check kubeconfig
cat ~/.kube/config

Setup Checklist

Before using this skill, verify:

• kubectl installed (kubectl version --client)
• Kubeconfig configured (~/.kube/config exists)
• Context set to correct cluster (kubectl config current-context)
• Permissions verified (kubectl auth can-i get pods)
• jq installed for JSON parsing (jq --version)