All your nodes and master should have communication with each other, without this you are going to have problems on cluster functionalities.

The ingress creates a load balancer pointing to your nodes machines.

Why your master cannot connect to your nodes?

Give a check on: https://kubernetes.io/docs/concepts/architecture/master-node-communication/

If you are deployed in a cloud environment, the most likely cause is incorrect firewall settings or route configurations. However, ingress configuration errors also may appear to look like infrastructure problems at times.

The Ingress will redirect your requests to the different services that it is registered with. The endpoint health is also monitored and requests will only be sent to active and healthy endpoints. My troubleshooting flow is as follows:

1. Hit an unregistered path on your url and check if you get the default backend response. If no, then your ingress controller may not be correctly set up (whether it be domain name, access rules, or just configuration). If yes, then your ingress controller should be correctly set up, and this is a problem with the Ingress definition or backend.

2. Try hitting your registered path on your url. If you get a 504 gateway timeout, then your endpoint is accepting the request, but not responding correctly. You can follow the target pod logs to figure out whether it is behaving properly.

If you get a 503 Service Unavailable, then your service might be down or deemed unhealthy by the ingress. In this case, you should definitely verify that your pods are running properly.

3. Check your nginx-ingress-controller logs to see how the requests are being redirected and what the internal responses are.