Uninstalling Kepler Operator

October 13, 2025 · View on GitHub

This guide provides step-by-step instructions for completely removing Kepler Operator from your cluster.

Pre-Uninstallation Checklist

Before uninstalling, consider:

  1. Backup PowerMonitor resources (if you might reinstall later):

    kubectl get powermonitor power-monitor -o yaml > powermonitor-backup.yaml

  2. Export Grafana dashboards (if customized):

    Export dashboards through Grafana UI or API

  3. Archive metrics (if needed for historical analysis):

    Use Prometheus federation or remote write to preserve metrics

  4. Notify stakeholders: Inform teams that energy monitoring will stop

Uninstalling via Helm (Kubernetes)

Step 1: Delete PowerMonitor Resources

Delete the PowerMonitor instance first:

# Check if PowerMonitor exists
kubectl get powermonitor power-monitor

# Delete the PowerMonitor
kubectl delete powermonitor power-monitor

This automatically removes Kepler DaemonSets and related resources.

Step 2: Uninstall the Operator

Using Makefile (from source repository):

make helm-uninstall

Using Helm directly:

helm uninstall kepler-operator --namespace kepler-operator

Step 3: Remove CRDs (Optional)

Warning: Deleting CRDs will permanently delete all PowerMonitor resources.

kubectl delete crd powermonitors.kepler.system.sustainable.computing.io
kubectl delete crd powermonitorinternals.kepler.system.sustainable.computing.io

Step 4: Clean Up Namespaces

Delete the operator namespace if no longer needed:

kubectl delete namespace kepler-operator

Step 5: Remove Webhooks (if not auto-deleted)

Check for remaining webhook configurations:

kubectl get validatingwebhookconfiguration | grep kepler
kubectl get mutatingwebhookconfiguration | grep kepler

Delete if present:

kubectl delete validatingwebhookconfiguration <webhook-name>
kubectl delete mutatingwebhookconfiguration <webhook-name>

Verify Complete Removal

Ensure all resources are removed:

# Check for remaining operator resources
kubectl get all -n kepler-operator

# Check for CRDs
kubectl get crd | grep kepler

# Check for webhooks
kubectl get validatingwebhookconfiguration | grep kepler
kubectl get mutatingwebhookconfiguration | grep kepler

# Check for ServiceMonitors (if using prometheus-operator)
kubectl get servicemonitor -A | grep power-monitor

Uninstalling via OLM (OpenShift)

The operator repository includes a comprehensive uninstall script:

# Clone repository if not already done
git clone https://github.com/sustainable-computing-io/kepler-operator.git
cd kepler-operator

# Dry run to see what will be deleted
./hack/uninstall-operator.sh

# Actually delete resources
./hack/uninstall-operator.sh --delete

The script automatically:

  • Lists all Kepler-related resources
  • Deletes PowerMonitor instances
  • Removes CSV (ClusterServiceVersion)
  • Deletes Subscription
  • Removes operator resources
  • Cleans up CRDs (with confirmation)

Method 2: Manual Uninstallation via Web Console

  1. Delete PowerMonitor Resources:

    • Navigate to Search → Select PowerMonitor kind
    • Delete the PowerMonitor instance

    Or via CLI:

    oc delete powermonitor power-monitor

  2. Uninstall Operator:

    • Navigate to Operators → Installed Operators
    • Find Kepler Operator
    • Click the three-dot menu → Uninstall Operator
    • Confirm deletion

  3. Verify Removal:

    • Check Installed Operators list
    • Verify Kepler Operator no longer appears

Method 3: Manual Uninstallation via CLI

Delete PowerMonitor Resources

oc delete powermonitor power-monitor

Wait for the Kepler DaemonSet to be removed.

Step 2: Delete the Subscription

oc delete subscription kepler-operator -n openshift-operators

Step 3: Delete the ClusterServiceVersion (CSV)

Find the CSV:

oc get csv -n openshift-operators | grep kepler

Delete it:

oc delete csv <kepler-operator-csv-name> -n openshift-operators

Step 4: Delete Custom Resource Definitions (Optional)

Warning: This permanently deletes all PowerMonitor resources.

oc delete crd powermonitors.kepler.system.sustainable.computing.io
oc delete crd powermonitorinternals.kepler.system.sustainable.computing.io

Step 5: Clean Up Additional Resources

Remove any remaining resources:

# Check for operator deployments
oc get deployment -n openshift-operators | grep kepler

# Delete if present
oc delete deployment <deployment-name> -n openshift-operators

# Check for service accounts
oc get sa -n openshift-operators | grep kepler

# Delete if present
oc delete sa <sa-name> -n openshift-operators

# Check for cluster roles and bindings
oc get clusterrole | grep kepler
oc get clusterrolebinding | grep kepler

# Delete if present
oc delete clusterrole <clusterrole-name>
oc delete clusterrolebinding <clusterrolebinding-name>

Verify Complete Removal on OpenShift

# Check CSV is gone
oc get csv -n openshift-operators | grep kepler

# Check subscription is gone
oc get subscription -n openshift-operators | grep kepler

# Check for remaining resources
oc get all -A -l "app.kubernetes.io/part-of=kepler-operator"

# Check CRDs
oc get crd | grep kepler

# Check ServiceMonitors
oc get servicemonitor -A | grep power-monitor

Removing Monitoring Components (Optional)

Remove ServiceMonitors

If using prometheus-operator:

kubectl delete servicemonitor -A -l "app.kubernetes.io/part-of=kepler-operator"

Remove Grafana Dashboards

OpenShift:

If you deployed Grafana using the deployment script:

oc delete deployment grafana -n kepler-operator
oc delete svc grafana -n kepler-operator
oc delete route grafana -n kepler-operator
oc delete configmap grafana-datasources -n kepler-operator
oc delete configmap grafana-dashboards -n kepler-operator

Kubernetes:

Delete dashboards via Grafana UI or remove dashboard ConfigMaps.

Preserve Historical Metrics (Optional)

If you want to keep historical energy data:

  1. Prometheus Federation: Set up a remote Prometheus to federate data
  2. Remote Write: Configure Prometheus to remote write to long-term storage
  3. Snapshot: Take Prometheus snapshots before uninstalling

Troubleshooting Uninstallation

Resources Not Deleting

If resources remain after deletion:

  1. Check for finalizers:

    kubectl get powermonitor power-monitor -o yaml | grep finalizers

    Remove finalizers if stuck:

    kubectl patch powermonitor power-monitor \
  --type json \
  --patch='[{"op": "remove", "path": "/metadata/finalizers"}]'

  2. Force delete:

    kubectl delete powermonitor power-monitor --grace-period=0 --force

Namespace Stuck in Terminating

If the kepler-operator namespace is stuck:

  1. Check for remaining resources:

    kubectl api-resources --verbs=list --namespaced -o name | xargs -n 1 kubectl get --show-kind --ignore-not-found -n kepler-operator

  2. Remove finalizers from namespace:

    kubectl get namespace kepler-operator -o json | \
  jq '.spec.finalizers = []' | \
  kubectl replace --raw "/api/v1/namespaces/kepler-operator/finalize" -f -

CRDs Cannot Be Deleted

If CRDs refuse to delete:

  1. Delete the instance first:

    kubectl delete powermonitor power-monitor

  2. Remove finalizers from CRD:

    kubectl patch crd powermonitors.kepler.system.sustainable.computing.io \
  --type json \
  --patch='[{"op": "remove", "path": "/metadata/finalizers"}]'

  3. Force delete:

    kubectl delete crd powermonitors.kepler.system.sustainable.computing.io --grace-period=0 --force

OLM Resources Remain on OpenShift

If OLM resources persist after CSV deletion:

  1. Check InstallPlan:

    oc get installplan -n openshift-operators | grep kepler

    Delete if present:

    oc delete installplan <installplan-name> -n openshift-operators

  2. Check OperatorGroup (only if no other operators in namespace):

    oc get operatorgroup -n openshift-operators

Post-Uninstallation

After complete removal:

  1. Verify no Kepler metrics in Prometheus:

    Query Prometheus for Kepler metrics - should return no results:

    {__name__=~"kepler_.*"}

  2. Check system resources are freed:

    kubectl top nodes

    CPU/memory usage should reflect Kepler removal.

  3. Optional: Remove monitoring stack (if no longer needed):

    Kubernetes:

    # If using kube-prometheus-stack
helm uninstall prometheus -n monitoring
helm uninstall grafana -n monitoring

    OpenShift:

    OpenShift monitoring is cluster-managed - do not remove.

  4. Optional: Remove cert-manager (Kubernetes only, if no other users):

    Warning: Only remove if no other operators/services use cert-manager.

    kubectl delete -f https://github.com/cert-manager/cert-manager/releases/download/v1.18.2/cert-manager.yaml

Reinstallation

If you want to reinstall Kepler Operator later:

  1. Follow the appropriate installation guide:

  2. Restore PowerMonitor resources from backup:

    kubectl apply -f powermonitor-backup.yaml

  3. Restore Grafana dashboards (if backed up)

Next Steps