metrics-server pod metrics not available in OrbStack Kubernetes

[suyashbhawsar]

Describe the bug

Environment

• OrbStack Version: 2.0.4 (2000400)
• macOS Version: 26.0.1 (Build 25A362)
• Kubernetes Version: v1.32.6+orb1
• metrics-server Version: v0.8.0
• Container Runtime: docker://28.3.3

Problem Description

The metrics-server successfully collects node metrics but fails to collect pod metrics. Node-level metrics work correctly (kubectl top node), but pod-level metrics consistently fail with "podmetrics.metrics.k8s.io not found" errors.

### Current metrics-server Configuration
```yaml
args:
  - --cert-dir=/tmp
  - --secure-port=10250
  - --kubelet-preferred-address-types=InternalIP,ExternalIP,Hostname
  - --kubelet-use-node-status-port
  - --metric-resolution=15s
  - --kubelet-insecure-tls

Observations

• metrics-server pod is running and healthy
• metrics-server API service shows True status
• Node metrics work correctly
• No errors in metrics-server logs
• Pod metrics never become available (even after 30+ minutes)

Impact

• HorizontalPodAutoscaler (HPA) cannot function without pod metrics
• Vertical Pod Autoscaler (VPA) cannot make recommendations
• Pod-level resource monitoring is not possible

To Reproduce

1. Install metrics-server from official manifest
2. Patch with --kubelet-insecure-tls flag (required for local clusters)
3. Deploy any application with resource requests/limits defined
4. Wait 2-3 minutes for metrics collection

Expected behavior

Node metrics (working):

$ kubectl top node
NAME       CPU(cores)   CPU(%)   MEMORY(bytes)   MEMORY(%)
orbstack   97m          1%       778Mi           9%

Pod metrics (failing):

$ kubectl top pod
Error from server (NotFound): podmetrics.metrics.k8s.io "default/pod-name" not found

$ kubectl get --raw "/apis/metrics.k8s.io/v1beta1/namespaces/default/pods"
Error from server (NotFound): podmetrics.metrics.k8s.io "default/pod-name" not found

### Diagnostic report (REQUIRED)

OrbStack info:
  Version: 2.0.4
  Commit: bdc68f44f20561e398d2a70f45d00385fa55eadf (v2.0.4)

System info:
  macOS: 26.0.1 (25A362)
  CPU: arm64, 8 cores
  CPU model: Apple M1
  Model: MacBookAir10,1
  Memory: 16 GiB

Full report: https://orbstack.dev/_admin/diag/orbstack-diagreport_2025-11-02T08-20-47.453054Z.zip

### Screenshots and additional context (optional)

Node metrics work fine, suggesting the kubelet connection is functional, but pod-level metric collection fails silently without errors. Other local Kubernetes distributions (Minikube, Kind, Docker Desktop) have working pod metrics with identical metrics-server configuration.

This makes OrbStack Kubernetes unsuitable for testing HPA or any pod-level resource monitoring scenarios.

[slinorb]

Please check the workaround in #2143

[mturnaviotov]

JFYI. I set up successfully prometheus node exporter via their helm chart, and can see metrics properly from node

[suyashbhawsar]

JFYI. I set up successfully prometheus node exporter via their helm chart, and can see metrics properly from node

This one worked for me: #2143