What Can I Do If Monitoring Fails to Be Enabled for a Cluster Due to Residual Add-on Resources?

Symptom

• When monitoring is enabled for a cluster, the API returns an error message containing the field release name already exists.
• The request for enabling cluster monitoring has been delivered, but the monitoring status is Installation failed or Unknown. On the page for enabling monitoring, check the kube-prometheus-stack add-on. The add-on installation failure cause contains the field resource that already exists.

Cause Analysis

The kube-prometheus-stack add-on has residual resources.

Troubleshooting

You can run the following commands to clear residual resources and enable monitoring again after the residual resources are cleared:

kubectl delete ns monitoring

kubectl delete ClusterRole cluster-problem-detector custom-metrics-resource-aggregated-reader event-exporter prometheus-operator prometheus-server ucsaddon-cie-collector-kube-state-metrics

kubectl delete ClusterRoleBinding ucsaddon-cie-collector-kube-state-metrics cluster-problem-detector event-exporter prometheus-operator prometheus-server

kubectl delete apiservice v1beta1.custom.metrics.k8s.io