Currently, when Rancher tries to provision a Kubernetes cluster on vSphere, it needs to initiate API calls to the vSphere endpoint. In a hybrid cloud environment this often means that the Rancher server is not in the same network as the vSphere endpoint. Therefore inbound access is required to be added to a firewall so Rancher can reach the vSphere system. This naturally poses a security concern and creates administrative burden on our users who have to go through a security review to get this approved.

If instead of requiring direct API access, an agent could exist inside the network where the vSphere API lived, then this agent could broker the communication between the Rancher server and the downstream API. The agent would simply initiate an outbound API connection to the Rancher server (much like any node agent or cluster agent currently) and simultaneously proxy any API calls that Rancher needs to make to vSphere. This would also have the benefit of being able to be run through a HTTP proxy, which many security teams will appreciate as a less risky connectivity model.