I am trying to figure out what the best practices are around using Kubernetes Ingresses in Rancher. My initial thought is to use an Ingress per namespace, but I am quickly run out external ips. Is there way in to provide Rancher with additional virtual ip addresses? We are running rancher on our own infrastructure.
For any future searchers: I think this may solve my problem:
“SCHEDULING AGAINST MULTIPLE IPS OF A HOST
By default, Rancher assumes that only the public IP of the host is available when scheduling services that publishes ports or launching load balancers. If your host has multiple IPs that could be used, the host needs to be configured to allow the Rancher scheduler to know which IPs are available.
When there are multiple IPs available on the host for scheduling, Rancher will schedule against all the available scheduler IPs when a port is published through a service or a load balancer. The scheduler will report port conflicts after all available scheduler IPs have been allocated for that port.”
From the read of that, I should be able to create several virtual interfaces on each node and then give those to rancher to do with as it pleases. So “ifconfig en0:1” and so on. Please note: I have not tested this yet, so it may not work.