skills.rest
LiboMa
LiboMaCOMMUNITY
Explore all skills by this author →
GitHub

kubernetes-admin

Diagnose and remediate Kubernetes pod, node, networking, and storage failures using kubectl decision trees.

3|1|Updated Feb 14, 2026
Search Tags:#troubleshooting#remediation#kubernetes#eks#kubectl#crashloopbackoff#oomkilled
One-click install
npx skills add https://github.com/LiboMa/agenticops-chat --skill kubernetes-admin-liboma
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: kubernetes-admin
Source: https://github.com/LiboMa/agenticops-chat/tree/main/skills/kubernetes-admin
Command: npx skills add https://github.com/LiboMa/agenticops-chat --skill kubernetes-admin-liboma

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve? Kubernetes failures like CrashLoopBackOff, OOMKilled, ImagePullBackOff, Pending pods, and NotReady nodes require deep, scattered operational knowledge to diagnose and fix quickly. This Skill provides structured decision trees and ready-to-run kubectl commands that guide an agent from symptom to root cause to a risk-tiered remediation plan. ## Core Features & Use Cases - Failure Decision Trees: Step-by-step diagnostic flows for pod states (Pending, CrashLoopBackOff, ImagePullBackOff), node conditions (MemoryPressure, DiskPressure), service/networking, PVC/storage, and HPA/VPA autoscaling issues. - Fix Paths with Risk Levels: Remediation procedures for OOMKilled, ImagePullBackOff, ReplicasMismatch, Node NotReady, CoreDNS, HPA, and PVC Pending, each with investigation commands, sizing guidance, risk tier (L1/L2), and rollback steps. - EKS Deep-Dive References: Detailed references covering VPC CNI behavior, prefix delegation, security groups for pods, kubelet logs, taints, QoS classes, exit codes, and pod eviction thresholds. - Use Case: A deployment shows 2/5 replicas ready. The Skill walks through checking pod events, node capacity, and resource limits, identifies memory limits that are too tight, applies kubectl set resources with a 2x-peak sizing rule, and provides a kubectl rollout undo rollback. ## Quick Start Ask the agent to diagnose why a specific pod in a namespace is stuck in CrashLoopBackOff and propose a fix with rollback steps.

Frequently Asked Questions about kubernetes-admin

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I troubleshoot a pod stuck in CrashLoopBackOff?▼

Check the previous container logs with kubectl logs POD -c CONTAINER --previous, then inspect the exit code via kubectl describe pod. Exit code 137 indicates OOMKilled, 1 indicates an application error, and 127 means the entrypoint binary was not found.

How do I fix OOMKilled pods in Kubernetes?▼

Confirm OOMKilled via the container's lastState.terminated.reason, check peak memory with kubectl top pod, then set the memory limit to 2x the observed peak using kubectl set resources. For JVM apps, set -Xmx to 75% of the container memory limit.

Why is my pod stuck in ImagePullBackOff on EKS?▼

The two common causes are a bad image tag or expired ECR credentials, since ECR tokens expire every 12 hours. Check the image reference and pod events, then either roll back to a known good image or refresh the ECR pull secret.

Does this cover EKS-specific networking issues?▼

Yes, the references cover the VPC CNI plugin, ENI and IP limits per instance type, prefix delegation, custom networking, security groups for pods, and the AWS Load Balancer Controller. It includes pod density calculations and troubleshooting commands for aws-node.

What risk levels are assigned to Kubernetes remediation actions?▼

Fixes are tiered by risk: L1 covers low-risk actions like resource limit changes and CoreDNS restarts with rolling-update rollback, while L2 covers disruptive actions like node drains and PVC recreation. Every fix path includes an explicit rollback command.

How do I diagnose a node in NotReady status?▼

Check node conditions with kubectl describe node for MemoryPressure, DiskPressure, PIDPressure, or NetworkUnavailable, then inspect kubelet logs via journalctl or SSM. Remediation ranges from pruning images for DiskPressure to draining and restarting the kubelet for MemoryPressure.

Related Skills

Explore More Software Engineering Skills→
lloydchanglloydchang
2

health-check-monitor

Monitor Kubernetes cluster health across nodes, pods, and quotas.

Community
Advanced
cropsggcropsgg
15

incident-postmortem

Automate structured incident postmortems with metadata, timeline, impact, and root-cause analysis.

Community
Advanced
tjl8787tjl8787

ec2

Manages AWS EC2 instances including provisioning, configuration, security groups, and snapshots.

Community
Advanced
eric-sabeeric-sabe
1

iac-terraform

Enforce consistent Terraform infrastructure practices with state locking and pre-apply plan review.

Community
Advanced