K8s中Nginx前端连接AspNetCore后端方法及解析错误排查求助

Hey there! Let's break down how to connect your Nginx frontend to an ASP.NET Core backend in Kubernetes, and fix that frustrating net::ERR_NAME_NOT_RESOLVED error you're seeing.

Part 1: Proper Setup for Nginx + ASP.NET Core in Kubernetes #

First, let's make sure your base setup is correct—you're already halfway there with your ClusterIP service for the backend:

1. Backend ASP.NET Core Deployment & Service #

• Ensure your ASP.NET Core app is deployed as a Kubernetes Deployment (with replicas, health checks, etc.).
• Create a ClusterIP service (like your abphost service) to expose the backend internally within the cluster. This service acts as a stable DNS name for your backend pods, which is why you can curl it from the Nginx container.

2. Nginx Frontend Configuration #

The key here is to have Nginx act as a reverse proxy for your backend API calls. This keeps backend traffic within the cluster and avoids browser-side DNS issues. Here's a sample Nginx config:

server {
    listen 80;
    server_name localhost;

    # Serve your static frontend files
    root /usr/share/nginx/html;
    index index.html;

    # Proxy API requests to the backend ClusterIP service
    location /api {
        proxy_pass http://abphost; # Use your backend service name here
        proxy_set_header Host $host;
        proxy_set_header X-Real-IP $remote_addr;
        proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
    }

    # Handle SPA routing (if your frontend is a single-page app like React/Vue)
    location / {
        try_files $uri $uri/ /index.html;
    }
}

• Mount this config into your Nginx pod using a ConfigMap or persistent volume so it's applied when the pod starts.
• Deploy Nginx as a Deployment, then expose it with a NodePort service (which you're already doing) to make it accessible from outside the cluster.

Part 2: Fixing the net::ERR_NAME_NOT_RESOLVED Error #

The issue you're facing is super common—here's why it happens:

When you curl http://abphost from inside the Nginx container, you're using Kubernetes' internal DNS system, which knows how to resolve the abphost ClusterIP service. But your browser runs outside the cluster, so it has no idea what abphost is—hence the DNS resolution error.

The Fix: Adjust Frontend API Requests #

You need to make sure your frontend doesn't try to call the backend's internal service name directly. Here are two reliable solutions:

Option 1: Use Relative Paths (Recommended)

• Update your frontend code to use relative paths for API calls. For example, instead of http://abphost/api/get-data, use /api/get-data.
• With the Nginx config above, any request to /api will be proxied to your abphost backend. The browser only communicates with the Nginx NodePort address, so it never needs to resolve abphost.

Option 2: Expose Backend Externally (Less Ideal)

If you absolutely need to use absolute URLs in your frontend, you'll have to expose the backend to the outside world (via NodePort, LoadBalancer, or Ingress) and update your frontend to use that external address. This is less secure and couples your frontend to the backend's external endpoint, so use it only if necessary.

Quick Validation Steps #

1. After updating your frontend code and Nginx config, restart your Nginx deployment to apply changes:

   kubectl rollout restart deployment <your-nginx-deployment-name>
   

2. Visit your Nginx NodePort address (e.g., http://<cluster-node-ip>:<node-port>) in the browser. Your frontend should now make API requests through Nginx, and the DNS error should be gone.
3. Double-check that your backend service and Nginx are in the same Kubernetes namespace—if not, update the Nginx proxy path to use the full service DNS name: http://abphost.<backend-namespace>.svc.cluster.local.

内容的提问来源于stack exchange，提问作者tnovau