Harvest can support a fail over scenario but the switching of servers is outside of Harvest control. In order to be transparent to the Harvest users, the fail over server must use the RT_FORCE_NODE_NAME=alias
environment variable where the alias is the name of the main Harvest broker machine and not the name of the fail over server host name.
So when a disaster occurs and the main Harvest Broker server is down, then external controls must start the fail over server so that Harvest activity can resume, or if the fail over server must be running at the same time as the main Harvest Broker server then the DNS administrator must set the fail over server offline to prevent host name contention on the network.
If there is any active Harvest transactions occurring and suddenly the main Harvest server is brought down then those activites will not be committed to the database and will have to rerun. If the end user gets an error message indicating connectivity to the Broker is no longer valid, then he/she will have to restart Workbench and reconnect to the Broker again and resume their SCM work.