Troubleshooting odd network issue

on my on prem cluster, every now and then I run in to an issue where the service just doesn’t load from the pod behind it

image

The pod is online. From all the worker nodes I can reach the ip 10.244.1.33. But the service does not appear to be able to get to it to load.

Not sure what the best way to troubleshoot kube networking is.

thanks!

Network is complicated, because network is different in different environments.

I really don’t know how your network setup looks like, but I’d try to understand if the network overlay is related here or not. And if it is… It really depends on which one you have and how it is configured. But I don’t have much experience with them :-/

Did you check if pod-to-pod communication is possible and that the pod itself can access the network according to the policy you defined?
If you find that the pod can not communicate with other pods (skipping the service) it is probably related to either your networkpolicy or your CNI. If you find that pod-to-pod works, it is related to how you configured your services (probably a weird combination of nodeport and loadbalancer - maybe try clusterip if that fits your needs)