How to fix unable to fetch metrics from API in Kubernetes

Check if metrics-server exists and is healthy:

kubectl get deployment metrics-server -n kube-system
kubectl get pods -n kube-system -l k8s-app=metrics-server

If metrics-server is not found or in CrashLoopBackOff:

kubectl logs -n kube-system -l k8s-app=metrics-server | head -50

Common log errors include TLS failures or API server connectivity issues.

If metrics-server is missing, install it from the official release:

kubectl apply -f https://github.com/kubernetes-sigs/metrics-server/releases/latest/download/components.yaml

Wait for the deployment to be ready:

kubectl wait deployment metrics-server -n kube-system --for=condition=Available --timeout=120s

For clusters with self-signed certificates or insecure kubelets, add the --kubelet-insecure-tls flag:

kubectl patch deployment metrics-server -n kube-system --type='json' -p='[{"op":"add","path":"/spec/template/spec/containers/0/args/-","value":"--kubelet-insecure-tls"}]'

Verify the metrics API is registered and responding:

kubectl get --raw /apis/metrics.k8s.io/v1beta1/nodes
kubectl get --raw /apis/metrics.k8s.io/v1beta1/pods

If these return errors, check the APIService registration:

kubectl get apiservice v1beta1.metrics.k8s.io -o yaml

Look for conditions showing "Available: True". If False, the metrics-server service is not reachable.

Test that metrics collection is functioning:

kubectl top nodes
kubectl top pods -n <namespace>

If nodes work but pods don't, check kubelet configuration. If both fail, metrics-server is not properly connected.

Wait 60-90 seconds after installing metrics-server before testing, as initial metrics collection takes time.

Verify metrics-server has proper RBAC:

kubectl get clusterrolebinding | grep metrics
kubectl describe clusterrolebinding system:metrics-server

Check if network policies are blocking traffic:

kubectl get networkpolicies -A

If you have restrictive network policies, ensure metrics-server can reach kubelets on port 10250 and the HPA controller can reach metrics-server.

After fixing metrics-server, verify HPA is working:

kubectl describe hpa <hpa-name> -n <namespace>

Look for events without 'FailedGetResourceMetric' errors. The TARGETS column should show actual percentages instead of '<unknown>':

kubectl get hpa -n <namespace>

Allow 30-60 seconds for metrics to propagate before checking HPA status.