3-Node-Cluster no connection if node with VIP goes down

[liszca]

On my 3-Node-Cluster there is no connection possible if the node with VIP goes down.

While the node is up I can send my curl request to all nodes and have an answer, except X-Forwarded-For & X-Real-Ip is only correct when hitting the node with the VIP.

Is there something that can be done?

Or is it Traefik related?

[brandond]

If you have only a single pod, then you are dependent on the node where the pod is running being up. Have you tried scaling up your deployment for multiple replicas?

This is just basic Kubernetes stuff.

[liszca]

After scaling its fine:
kubectl scale deployment -n kube-system traefik --replicas 3

But how do I make soure my internet router knows which of the nodes is up?
Lets say one is down for maintenance how does the system work then?
Until now I had one IP address to the nginx reverseproxy with a port forwarding rule in my internet router and that is it.

But now with Kubernetes how is this part configured.

[brandond]

This is where having a real LB controller would be necessary. MetalLB, Kube-VIP, or something like that. As the K3s docs say, ServiceLB is "just enough" of a load balancer implementation for things to work at a very basic level.