Minikube中Kubernetes Pod DNS nameserver IP错误致服务解析失败

Got it, let's work through this DNS resolution issue you're hitting in Minikube. I've seen this exact scenario a few times, so here's how to diagnose and fix it properly—no temporary manual edits required!

问题根源 #

The core issue here is that your Pod's DNS configuration is pointing to the wrong kube-dns Cluster IP:

• Inside the Pod, /etc/resolv.conf lists 10.96.0.10 as the nameserver, but your actual kube-dns service has a Cluster IP of 10.0.0.10
• That's why using the service name web-server:8081/endpoint fails (it can't reach the correct DNS server to resolve the name), but accessing the direct IP works (it skips DNS entirely)

Quick Temporary Fix (You Already Did This!) #

Manually updating the nameserver in /etc/resolv.conf to 10.0.0.10 will get DNS working immediately, but this is a band-aid—restarting the Pod will reset the config. Let's fix the root cause.

Root Cause Troubleshooting Steps #

1. Confirm kube-dns's actual Cluster IP
   Run this command to verify the kube-dns service details:

   kubectl get service kube-dns -n kube-system
   

   Double-check that the CLUSTER-IP matches 10.0.0.10 (which you already confirmed it does).

2. Check the Pod's DNS Policy
   Look at your problematic Pod's configuration to see if the DNS policy is misconfigured:

   kubectl get pod <your-pod-name> -o yaml | grep -A5 dnsPolicy
   

   • The default dnsPolicy is ClusterFirst, which automatically points the Pod's nameserver to kube-dns's Cluster IP
   • If it's set to None or a custom value, that could override the correct DNS settings

3. Verify DNS Component Health
   Make sure kube-dns (or CoreDNS, which Minikube uses in newer versions) is running properly:

   kubectl get pods -n kube-system | grep -E 'kube-dns|coredns'
   

   If the DNS Pod is in CrashLoopBackOff or NotReady state, it might have restarted and picked up a new IP, or be unable to serve requests.

4. Check Kubelet's DNS Configuration
   Minikube's kubelet might be configured with the wrong DNS IP. SSH into the Minikube node to check:

   minikube ssh
   cat /var/lib/kubelet/config.yaml | grep clusterDNS
   

   The clusterDNS value here should match kube-dns's Cluster IP (10.0.0.10). If it's 10.96.0.10, that's the source of the misconfiguration.

Permanent Fixes #

Pick the fix that matches your troubleshooting findings:

• If kubelet's clusterDNS is wrong:
  Delete and restart Minikube with the correct DNS IP specified:

  minikube delete
  minikube start --extra-config=kubelet.cluster-dns=10.0.0.10
  

• If Pod's dnsPolicy is misconfigured:
  Update your Pod's YAML to use the default ClusterFirst policy, then redeploy:

  apiVersion: v1
  kind: Pod
  metadata:
    name: your-pod-name
  spec:
    dnsPolicy: ClusterFirst
    # ... rest of your Pod spec
  

• If DNS Pod is unhealthy:
  Delete the faulty DNS Pod—Kubernetes will automatically spin up a new healthy one:

  kubectl delete pod <dns-pod-name> -n kube-system
  

Verify the Fix #

After applying the fix, check the Pod's DNS config again to confirm:

kubectl exec -it <your-pod-name> -- cat /etc/resolv.conf

You should see 10.0.0.10 as the nameserver. Then test the service name resolution:

kubectl exec -it <your-pod-name> -- wget web-server:8081/endpoint

This should now work as expected!

内容的提问来源于stack exchange，提问作者Sergii Bishyr