the doc states “Refrain from adding an Ingress to the local cluster” (https://rancher.com/docs/rancher/v2.x/en/k8s-in-rancher/load-balancers-and-ingress/), but thats exactly what I want to have: A common URL for the whole Rancher web gui, no matter if one of the nodes in the “local” kube (3 nodes bare metal) is down. Changing other clusters should not affect the ingress of local. Surely I don’t want to loose my “Rancher connection”, whatever thats supposed to mean.
Somehow I have the impression of “dirty code” in this case. Would it be possible to distinguish between a “local” ingress controller and the common ingress controller to access all clusters created via Rancher? Can I add a LoadBalancer to local, e.g. via metallb?
It would be very nice if somebody could elaborate.
The docs are referring to the “Ingress” resource, not the “Ingress Controller.”
The helm chart that installs rancher does indeed create an Ingress for rancher.
I believe what we are trying to communicate here is that if you create an ingress resource, the Nginx ingress controller will cause Nginx to reload, which websocket connections. The connection from downstream clusters to Rancher is long-lived websocket connections. So if you cause Nginx to reload, you’ll see a blip where the downstreams disconnect and reconnect.
Sorry for the confusion. I didn’t see the difference between ingress and ingress controller.
https://rancher.com/docs/rancher/v2.x/en/installation/options/chart-options/ mentions to set “hostname” for the FQDN for the Rancher server, but it doesn’t tell how to set the IP address. Is the IP address bound to this hostname a common, independent IP address for the whole cluster, or is it a Round Robin construct listing all controlplane nodes?
Tricky question? Or was I too blind to see?