Generally, it is recommended to use as much as possible the vsx-sync configuration for all feature-group (aaa, time, ospf, bgp ... to name a few). In case of BGP, if you have iBGP session between VSX nodes, a command has been introduced in 10.3 or 10.4 for this purpose in BGP section:
neighbor <ip-address> vsx-sync-exclude; so that the neighbor IP address of the secondary is not pushed to the secondary itself.
This is probably what you look for.
------------------------------
Vincent Giles
------------------------------
Original Message:
Sent: Feb 26, 2021 10:07 AM
From: Mark Hettleman
Subject: L3 routing within VSX pair
To setup iBGP between switches vxs-sync for BGP needs to be disabled. Is this correct? When would you ever enable vxs-sync for BGP.
------------------------------
Mark Hettleman
------------------------------
Original Message:
Sent: Feb 25, 2019 10:08 AM
From: Vincent Giles
Subject: L3 routing within VSX pair
yes, you can set-up iBGP between your VSX nodes inside the VSX cluster, and if there are not BGP route-reflector clients, it is actually a must for your iBGP continuity.
To configure iBGP between VSX nodes, this is exactly the same as configuring iBGP between any other network nodes.
Best practice is generally to use Loopback address for iBGP peering. The reachability of loopbacks is guaranted by the IGP (OSPF). To constuct OSPF peering between VSX nodes, simply set-up a dedicated transit VLAN, transported over the ISL, on which OSPF neighbors can establish adjacency.