In an SD-WAN environment with two vSmart controllers, the design does provide redundancy, but that does not always mean there will be zero interruption during maintenance. When one vSmart controller is rebooted, the other controller continues to handle the control plane, but devices that were previously connected to the rebooting controller may need to reestablish their control connections and download policies again.
During that reconnection process, some routers might briefly renegotiate control sessions or refresh routing and policy information. In most well configured deployments this process happens quickly and users may not notice it, but Cisco usually mentions a potential interruption because the control connections are being redistributed and synchronized between the remaining controller and the devices in the fabric.
Another factor is how evenly the edge routers are distributed between the two vSmart controllers. If a large number of devices reconnect at the same time when the controller comes back online, there can be short control plane convergence events. These are normally small and temporary, but they are still technically considered an interruption in the control layer.
This kind of behavior is common in distributed systems where redundancy exists but sessions still need to rebalance. It is similar to how small formatting differences can occur in distributed text systems that rely on Unicode characters. Developers sometimes experiment with things like caracter invisible to test how different platforms process text encoding and formatting, which highlights how even small technical changes can create subtle system responses.
Overall, with two vSmart controllers the impact should be minimal and limited to brief control plane reconvergence rather than a full service outage. The key factors are proper redundancy configuration, device distribution across controllers, and stable control connections before the maintenance begins.
