DevOps Coding Challenge #6: Kubernetes Pod Restarter

Level: Intermediate
Languages: Python or Go
Focus: Kubernetes API, Automation, Production Troubleshooting

---

Problem Statement

You’re on-call. Pods are stuck in a CrashLoopBackOff or Error state.
Instead of manually restarting them, automate it.

Your task:
Write a script (in Python or Go) that connects to a Kubernetes cluster and restarts pods in a given namespace, but only if:

• The pod has been in a bad state (CrashLoopBackOff or Error)

• The condition has persisted for more than 10 minutes

• The pod name matches a given pattern (e.g., "api")

---

Core Requirements

• Accept these inputs via CLI:

  • --namespace

  • --pattern

  • --threshold (in minutes)

• Restart the pod using the Kubernetes API (delete it; let the ReplicaSet recreate it)

• Print/log:

  • Which pods were deleted

  • Why they were deleted

---

Bonus Features

• Add a --dry-run flag to simulate deletion without taking action

• Add a summary output like:

Checked 18 pods  
Restarted 5 pods  
Skipped 13 pods

---

Tech Suggestions

If using Python:

• Use the kubernetes client (pip install kubernetes)

• Use argparse for CLI

• Use datetime to calculate age

• Docs: https://github.com/kubernetes-client/python

If using Go:

• Use client-go or controller-runtime

• Use Cobra or the flag package for CLI

• Watch for containerStatuses under .status.containerStatuses[*].state

---

Evaluation Criteria

• Correct filtering and logic

• Clean code structure

• Proper use of Kubernetes API

• Bonus: dry-run logic, helpful logging, CLI flags

---