Problem Statement: Sometimes application pods get stuck in CrashLoopBackOff or fail due to transient issues. As an SRE, we need a simple tool that can detect such pods and restart them automatically.
Detailed Scenario: In a large OpenShift cluster, developers may report that their pods are stuck or not recovering automatically. Instead of manually running oc delete pod, you can build a Python utility that scans all namespaces, identifies unhealthy pods (e.g., status CrashLoopBackOff, Error, ImagePullBackOff), and restarts them by deleting the pod (so the ReplicaSet/Deployment brings it back).
Use Case Approach: oc client to fetch pod statues, execute oc delete pod command
Tools and Modules: subprocess
══════════════ ⭑ ⭑ ⭑ ⭑ ⭑ ══════════════
Approach:
- Use the oc CLI or Kubernetes Python client to fetch pod statuses.
- Filter pods that are in bad states.
- Restart them by running oc delete pod -n .
- Log all restarts and generate a small HTML/CSV report for visibility.
══════════════ ⭑ ⭑ ⭑ ⭑ ⭑ ══════════════
- Uses
oc get pods -o jsonto list pods (all namespaces or a single namespace) - Detects unhealthy pods (CrashLoopBackOff, ImagePullBackOff, ErrImagePull, etc.)
- Restarts them by deleting the pod (Deployment/ReplicaSet/StatefulSet will recreate)
- Supports --dry-run to preview actions
Restart unhealthy pods in ALL namespaces
python oc_pod_restarter.py
Only a specific namespace
python oc_pod_restarter.py -n my-namespace
Filter by label (e.g., only app=nginx)
python oc_pod_restarter.py -n my-namespace --selector app=myapp
