The GUI/shell difference is worth isolating before changing more NAT rules. Your earlier results list 10.2.0.1 as the source for the shell tests, but not the exact command.
On TAC, assuming those addresses are still current, could you compare:
and the GUI ping to the same destination with the LAN source? A capture of ICMP on the WireGuard interface of TAC and of the hub during each test would show which source is used and whether replies come back.
A separate check on the hub:
On TAC, assuming those addresses are still current, could you compare:
Code Select
ping -c 3 10.1.0.11
ping -c 3 -S 10.2.0.1 10.1.0.11
and the GUI ping to the same destination with the LAN source? A capture of ICMP on the WireGuard interface of TAC and of the hub during each test would show which source is used and whether replies come back.
A separate check on the hub:
Code Select
wg show all allowed-ipsIn your first post, peer TAC has 172.19.0.2/24. WireGuard stores that as 172.19.0.0/24, and if several peers list the same prefix, only one peer keeps it. If the other spokes are set the same way, traffic to their tunnel addresses would all go to that one peer. That alone would not explain the GUI/shell difference, so I would start with the ping comparison.
"