Catch-22 problem with health monitor

This is not to hard I think, you need to copy your original pool but do not put any monitors on it or put a simple monitor like a TCP or TCP Half Open just to ensure the servers are up.

Then you need an iRule to check default pools active member status:

when CLIENT_ACCEPTED {
     Set backup pool name to variable
    set backupPool "backup_pool"

    if {[active_members [LB::server pool]] < 1} {
        if {[active_members $backupPool] > 0} {
            log local0.alert "Default pool [LB::server pool] as not active members, forwarding to $backupPool"
            pool $backupPool

        } else {
            log local0.alert "Both default pool [LB::server pool] and $backupPool have no active members. Closing TCP connection"
            TCP::close
        }
    }
}

Backup pool name is set to a variable at the start of the iRule and you can also change the final TCP::close if you wanted to do some other response like a sorry page etc.