WAN interface is eth2. It set to DHCP
LAN interface is eth3. It set to static address

@Viacheslav we need to port this to 1.2 too

What is on the eth3 interface?

Considered while implementing T3642 to have a general LE availability.

Tested on VyOS 1.3.0-epa1

Tested in VyOS 1.2.8, VyOS 1.3.0-epa1, VyOS 1.4-rolling-202109190558

The router does not automatically configure the prefix route that the client received. If you use the client's link-local address for the prefix route then everything works. The client also uses the dhcpv6-server link-local address for the default route.

Tested in VyOS 1.3.0-epa1 & VyOS 1.4-rolling-202109190558

https://github.com/vyos/vyos-1x/pull/1028 — though probably needs some tests developed?

Tested on VyOS 1.3.0-epa1.
Confirm IBGP reflection to non-RR-Client
Lab Topology:

@artooro Will be available in the next rolling release
Let us know, if you want some other capabilities

Sorry for the late reply, I've been waiting for a maintenance window to test this in. Again as this is a production device I only have limited debug info before having to roll back to the working version. I installed 1.4-rolling-202110150613 and on booting show interfaces does indeed show all the interfaces. The igb driver interfaces were showing as up (u/u). The i40e interfaces were showing as down (u/D).

PR https://github.com/vyos/vyos-1x/pull/1027

In the real-world to avoid it they used cluster-id / a session BGP between them , it's the idea of RR :

Maybe be added to gether with T1229

@francis Sorry, I don't understand the problem.
Agree that route received from one IBGP peer should not be forwarded to another IBGP peer. Except for the RR client.

@NikolayP The concern here is solely this: IBGP neighbors that do not have route-reflector-client set should not received learned routes. Currently, they do. If this in intentional, then the docs should be updated to clarify this.

@maznu Can you create a PR?

Initial bug was Fixed, VyOS 1.4-rolling-202110130217

vyos@r1-roll# compare 
[edit container]
+name dns02 {
+    image ubuntu:focal
+    network dnsnet {
+        address 10.0.72.253
+    }
+}
+network dnsnet {
+    prefix 10.0.72.0/24
+}
-network net01 {
-    prefix 10.0.72.0/24
-}

If Cluster ID is not used, full IBGP mesh must be used. Exception is RR client, they should only have peering with RR.
Router 10.0.0.21 has no peering with 10.0.0.3.
This is incorrect IBGP design.

Why not always move to decimal output and detect on demand if colons are present or not and adjust the string? Ne need for the user to take any action at all?