We use something different - but it works for us
We have Usermanager on a central server, we export a copy of the database and run a backup usermanager on each hotspot as well. I place both servers's IPs into the Radius client on the MT Hotspot. I put the central server as the first entry with a 3 second timeout, local copy as the second entry.
In this way, if the central server dies, it falls back to the local copy, which is usually no more than about 15 minutes old. It's not perfect, but it allows existing customers to connect without them even knowing the central server has fallen over.
It has flaws, in that while the central server is down, no new customers can create an account and also existing customers cannot top up their credit, also we lose any usage data for when the central server is down as it will now connect to and update the local copy, which will eventually get overwritten with old data once the main server is back up....
But then, this is a cheap and dirty fix to keep the existing customers online long enough for us to figure out why the central server has died, without them knowing anything has gone wrong. Nothing more than that.
It's worked really well for us for the last few years and we've had at least three or four outages of the central server.