In an HA configuration, as I understand it, the Rancher Dashboard should be configured as a separate node from the Kubernetes cluster. To save on the server allocation is it even possible to configure the Rancher Dashboard on one of the master node?
Thanks,
Chris
It is, but admins of that cluster are effectively admins of all clusters attached to the server because they can write to resources that manage permissions in the other clusters.
As a general rule, if the cost of doing a HA install is even a consideration, you would probably be better off with a single container and bind-mounting in persistent storage (and/or a good backup strategy). HA is at least triple the cost + additional overhead, and complexity, and bandwidth if you’re paying for it.
“Not HA” sounds scary, but during a failure, existing clusters and their workloads keep running as normal. So you’re talking about a small amount of downtime for managing things on the rare-ish occasion the single container fails until someone fixes it.
If you are an individual or small business, you can probably just fix it when it happens and move on with your life. If you have hundreds of employees that are going to be losing productivity and paging you if it’s down, then you definitely need HA. But typically the cost of 3 VMs is going to be rounding error on the k-cup budget for those companies.
Individual clusters should definitely have at least 3 nodes for etcd, but these can also be worker nodes.