Metrics server on EKS: No metrics available to the server

It appears that the issue was related to a version incompatibility with Helm, which is causing the sharedIndexInformer probe failures and consequently preventing access to metrics via Kubernetes. The error messages indicate attempts at probing for storage readiness but encounter “no metrics to serve.” This situation suggests problems either within your cluster’s metric collection or that certain components are not initialized properly, making the Metrics API unavailable.

Switching to a different Helm chart as per the resolved action seems like an effective solution. Here’s what this step might entail:

Identify Compatible Chart - You’ve identified that another version of metrics-server is more compatible with your cluster configuration and software stack, presumably managed by Helm.
Update the Resource Definition (Helm Recipe) – Your provided code snippet represents a resource definition in Terraform for installing or updating an application called ‘metrics_server’ using Helm within Kubernetes namespace kube-system. This is done via defining variables such as name, repository URL (https://kubernetes-sigs.github.io/metrics-server), and chart version.
Apply the New Chart – With Terraform configured to use Helm for deploying resources within your cluster (or update existing ones if already present), executing terraform apply would trigger this configuration, installing or updating ‘metrics_server’ based on those parameters in Kubernetes.
Verify Metric Availability – After applying the changes with Terraform and Helm, you should verify that your cluster is now ready to serve metrics by executing kubectl top pod -A or similar commands as shown previously where it was failing due to a lack of available monitoring data.

By following these steps based on common procedures for managing Kubernetes resources with Helm and Terraform, you should be able to resolve the issue at hand. This method also provides centralized control over infrastructure provisioning which is particularly useful when dealing with complex multi-cloud environments or hybrid deployments that often occur in DevOps workflows.

Remember this approach assumes familiarity with these tools and their interaction within your cluster environment, as well as understanding why a chart version conflict might be causing the issue. If further questions arise while troubleshooting similar issues: feel free to ask!