Kubernetes HPA显示<unknown>问题求助：前端服务无法获取有效指标值

Let’s break down what’s going on here and fix your HPA issue step by step:

The core problem is clear: your Metrics Server deployment isn’t ready (shows 0/1 in your kubectl get deployment output). HPA depends entirely on Metrics Server to pull CPU/memory usage data from pods—if Metrics Server can’t run properly, HPA has no metrics to display, hence the <unknown> value.

Here’s how to resolve this:

1. Diagnose Why Metrics Server Is Failing #

First, check the logs of the Metrics Server pod to uncover the root cause:

kubectl logs deployment/metrics-server -n kube-system

Common issues here include:

• Certificate validation errors (most local/private clusters use self-signed certificates for kubelet)
• Network connectivity problems with the Kubernetes API Server
• Missing required startup flags for Metrics Server

2. Fix the Metrics Server Deployment #

The most frequent fix for a stuck Metrics Server is adding a flag to skip kubelet certificate verification (safe for non-production environments; adjust certificate settings properly for production):

1. Edit the Metrics Server deployment:

   kubectl edit deployment metrics-server -n kube-system
   

2. Find the args section under the metrics-server container, and add the --kubelet-insecure-tls flag:

   args:
     - --cert-dir=/tmp
     - --secure-port=4443
     - --kubelet-preferred-address-types=InternalIP,ExternalIP,Hostname
     - --kubelet-use-node-status-port
     - --kubelet-insecure-tls  # Add this line
   

3. Save and exit the editor—Kubernetes will automatically restart the Metrics Server pod.

3. Verify Metrics Server Is Working #

Wait 30 seconds, then check if the deployment is ready:

kubectl get deployment metrics-server -n kube-system

Once it shows 1/1 Ready, confirm it can fetch metrics:

# Check node-level metrics
kubectl top nodes
# Check metrics for your udagram-frontend pods
kubectl top pods

If these commands return valid CPU/memory values, Metrics Server is functioning correctly.

4. Recheck Your HPA Status #

Give HPA a minute to pull the new metrics, then run:

kubectl describe hpa udagram-frontend

You should now see actual CPU usage percentages instead of <unknown>.

A quick note: Your udagram-frontend deployment is correctly configured with resources.requests.cpu—this is a required setting for HPA to calculate usage percentages, so you don’t need to adjust anything there.

内容的提问来源于stack exchange，提问作者Oussama Bouchikhi