Nginx (external) with fastcgi_pass to PHP (Kubernetes): upstream sent unsupported FastCGI protocol version: 72

Question

I am new in Kubernetes.
I had a working setup of separate servers for frontend and backend, all before Kubernetes.

A frontend webserver which served a static build of a React App, and a backend server which had docker containers for PHP APIs.

The URL of the backend server was for example my-backend-docker-server.foo

The backend server exposed a PHP API container, where the PHP port 9000 was exposed to 8087.

The frontend server used the PHP from that backend server like:

    location ~ \.php {
        # ...
        fastcgi_pass my-backend-docker-server.foo:8087;
    }

So this all worked.

Now I want to move all docker containers to a Kubernetes cluster.

I was able to do a Pod -> Service -> Ingress setup, I was able to have a new URL like php-api-on-kubernetes.foo, and when I call the URL in the browser, the output of the index.php from the PHP container shows up.

The Idea is now, that I just change the fastcgi_pass in the nginx configuration of the frontend webserver like:

    location ~ \.php {
        # ...
        # fastcgi_pass my-backend-docker-server.foo:8087;
        fastcgi_pass php-api-on-kubernetes.foo:80;
    }

But the frontend call receives a 502 (Bad Gateway) error, and the nginx log on the frontend server logs:
upstream sent unsupported FastCGI protocol version: 72 while reading response header from upstream

I tried 2 variants of ingress

First, which I am able to see the output of the PHP container over the new URL.

# see https://kubernetes.github.io/ingress-nginx/user-guide/fcgi-services/

apiVersion: v1
kind: ConfigMap
metadata:
  name: {{ .Release.Name }}
data:
  SCRIPT_FILENAME: "/home/api/public/index.php"

---

apiVersion: networking.k8s.io/v1
kind: Ingress
metadata:
  name: {{ .Release.Name }}
  annotations:
    nginx.ingress.kubernetes.io/backend-protocol: "FCGI"
    nginx.ingress.kubernetes.io/fastcgi-index: "index.php"
    nginx.ingress.kubernetes.io/fastcgi-params-configmap: {{ .Release.Name }}
spec:
  ingressClassName: nginx
  rules:
    - host: {{ .Values.host }}
      http:
        paths:
          - path: /
            pathType: Prefix
            backend:
              service:
                name: {{ .Release.Name }}
                port:
                  name: fastcgi

Since it logged that nginx error, I compared it more with the old backend server. I saw that the content of the PHP container was not accessible from a browser. Because there is also no webserver serving the content. So I tried to match, and did another ingress without nginx.

apiVersion: networking.k8s.io/v1
kind: Ingress
metadata:
  name: {{ .Release.Name }}
spec:
  rules:
    - host: {{ .Values.host }}
      http:
        paths:
          - path: /
            pathType: Prefix
            backend:
              service:
                name: {{ .Release.Name }}
                port:
                  name: fastcgi

This is the Service by the way

apiVersion: v1
kind: Service
metadata:
  name: {{ .Release.Name }}
spec:
  selector:
    app: {{ .Release.Name }}
  ports:
    - port: 9000
      targetPort: 9000
      name: fastcgi

Now when calling the new backend URL, I receive 502 Bad Gateway nginx output.
When I call the old backend URL, I receive ERR_CONNECTION_RESET (It should have always been like this, but without any problems).

So there is still a difference. Maybe it has something to do with a magic global ingress controller? As I said, I am still new into Kubernetes.

But still, I dont know how to fix the upstream error. I googled, asked ChatGPT, I dont know how to fix this error. I don't see/understand where I have to change a FastCGI protocol version on Kubernetes side.

The PHP image is just the same.

Any ideas?