12.04.2024 Problem:
The client encountered an error in the apiservice v1beta1.custom.metrics.k8s.io, with the status showing “False” and a message indicating that the service/prometheus-adapter in a specific namespace is not listening on port 6443. The client provided details, including the helm chart files used for deployment, seeking a resolution as the issue persisted across AWS and Azure cloud environments.
Solution:
The expert identified a mismatch between the v1beta1.custom.metrics.k8s.io apiservice port and the prometheus-adapter service port. After analyzing the helm chart, it was discovered that an old version of the chart contained a bug, leading to the issue. The expert recommended uninstalling and reinstalling Prometheus-adapter on the OpenShift Container Platform (OCP) cluster. The client was advised to adopt the latest helm charts for Prometheus-adapter from the GitHub repository. During the client-expert meeting follow-up, various aspects were discussed, including error investigation, bug fixing, service deployment, and troubleshooting. The client is working on implementing the suggested changes, and further updates will be provided after thorough testing. Additionally, a new issue related to the Prometheus endpoint not getting created was addressed, and the expert recommended upgrading to the latest version of Prometheus-adapter and shared steps for custom metrics verification.
Conclusion:
The expert found a port mismatch in Kubernetes API service & Prometheus-adapter due to the bug in the old helm chart. Advised client to reinstall with latest helm charts from GitHub. Discussions included error investigation & troubleshooting. Client implementing changes, updates pending after testing. Also addressed Prometheus endpoint issue by recommending upgrade & providing metrics verification steps.