When the UNIStim endpoints detect a server failure, the UNIStim endpoints restart and try to re-register with the primary server (S1) – the load balancer of cluster 1. Typically, the load balancer assigns the registration to one of the Device Adapter nodes and registration completes. If S1 does not respond, the endpoints wait briefly and retry the registration.

If the cluster is configured for N+1 along with geo-redundancy and if S1 does not respond, the UNIStim endpoints try to register with the secondary server (S2). If S2 does not respond, the endpoints wait briefly and retry the registration to S1.

In an N+1 model, the UNIStim endpoint try to fail over to the primary server first.

If the primary cluster is sized to have enough capacity to support the endpoints within that cluster, the outage is usually short and may not be detected by an end user, unless the user tries to use the endpoint during the fail over.

Depending on factors such as network speed and Session Manager resiliency, the fail-over time for the endpoint to register to a Device Adapter node is typically five minutes or less after an endpoint detects a registration failure.

However, even a short interval for detecting registration failure and failing over may be noticed by the users. Hence, even a very fast switch-over may be noticed by the users.