Bright will release an upgrade script about a month after Bright 9.1 was released. You can use that script to upgrade your cluster in place.
Alternatively, you can do a parallel upgrade. A parallel upgrade is where you deploy a new head node, in parallel with your existing head node(s) - meaning that this new head node has an interface on both the external and internal networks. Then you deploy a fresh copy of Bright on that new (parallel) head node. You can "lock down" the DHCP servers on both the old and new head nodes so that each only responds to the nodes it is managing.