Rancher2 + AKS, is it ready for use?


Long time Rancher1/Cattle user, and now jumping to try out Rancher2/K8s.

I’m fortunate to have plenty of Azure credits to work with, and worrying about anything but my services doesn’t really interest me much, so AKS would on the surface seem like a great fit for us.

I followed the instructions to set up Rancher 2.0.6 , a Vnet+subnet, and service principal. Got Rancher to initialize the cluster, and mostly things seemed to work as it should. AKS cluster was created in the correct subnet. So far the only problem seems to be continuous timeouts and slowness in rancher. (rancher and the cluster are in the same region and same vnet, so that should not be to blame)

As a simple test I deployed a hello-world web service, and within the network i can curl it fine. Took ages to deploy again, though. like 10 minutes for nothing but a container. But how do I expose it outside the cluster via somehow configurable load balancer? It says in the documents that Rancher does not support L7 on AKS clusters? So does that mean the whole Load Balancing tab and adding an Ingress is unusable? atleast adding one there just creates a rule, but i haven’t found a way to connect a dns entry to it, as there does not seem to be a public or internal ip of any kind to connect to.

Creating and L4 load balancer as part of the workload creation, generates an L4 ingress and gives me a public ip, but that’s no way to expose any real services. I can’t have everything pointed to different public IP’s directly, without routing control (similar to the rancher 1.6 haproxy LB). Again, though, it took more than 1 hour for the L4 lb to appear after the workload was created.

Another issue seems to be that as I was trying Catalog apps, there was honestly not one that would deploy all the way. Namespace creation might take an hour, and the rest of the deploy would generally fail at some point within the next 2 hours. sometimes it was config map fail, and sometimes secrets would not get deployed. On the positive side, disks seem to get provisioned.

Overall to me it seems like nothing works all the way out of the box, and the Rancher2+AKS combo seems unusable due to slowness, and missing support for features.

Are there any other people experimenting or even using Rancher2+AKS in production that have had a better experience? Any AKS specific solutions or tips&tricks that I should know about before going further?