There is a Cisco ASA connected to the aggregation switches. There is a port-channel L3 subinterface that maps to the interface vlanif in the system context, which allows L3 separation and the absence of a spanning tree loop and at the same time feed the same vlan to the ASA in several contexts, which saves public IP addresses.
In Hedex's documentation on USG and "CloudEngine 12800, 12800E, 8800, 7800, 6800, and 5800 Series Switches M-LAG Best Practices" I found a connection method that only uses L3 interfaces on the USG side, which is logical, because avoids loops in this topology.
Do I understand correctly that in this topology we can use only the L3 interface, Eth-trunk L3, Eth-trunk L3 subinterface, on the AGG side - interface Vlanif (dual active gateway) to have a topology without a loop?
Also, CloudEngine 12800, 12800E, 8800, 7800, 6800, and 5800 Series Switches M-LAG Best Practices in the example for static routing provides an example only when dual active gateway is implemented using VRRP. What is the reason that for static routing between nodes there is no recommendation to use dual active gw with one mac-address? The fact that each interface must have its own physical IP address?
