[bess] Discussions on ARP sync by RT-2 in pure L3 EVPN usecases

[<[e-mail]>]

Hi all,


I have some questions about using RT-2 in pure L3 EVPN usecases.


I haven't find clear explanation in current EVPN RFCs or drafts on these usecases.


I need some help to check my understanding.


Can you help me to check whether this usecase is supported by current EVPN control-plane or not ?


The phrase "pure L3 EVPN usecases" means that there is no IRB or MAC-VRF.





   /---- PE1----\

CE1(ES1)        PE3----CE2

   \---- PE2----/

   


CE1 attaches to IP-VRF x via sub-interface a/b on PE1/PE2.


And I think we can use RT-2 to sync ARP entries between sub-interface a and b.


In order to do this, I configured ESI1 on the main-interface of a and b.


Then I let the EVPN control-plane advertise the ARP entries on a/b to all EVPN PEs using EVPN Route-type 2.


I think the format of such RT-2 routes can be the following:


1) the MPLS label1 field is null (may be implicit-null or explicit-null),because there is no MAC-VRF.


2) the MPLS label2 field is the label of IP-VRF x


3) the RD field is the RD of IP-VRF x


4) the ESI is ESI1, the ETI is 0 


5) the RTs is the export RTs of IP-VRF x 






I also think we can advertise the subnet of sub-interface a/b together with ESI1 in EVPN RT-5 route.


And I haven't find clear explanation about the detailed usage of EVPN RT-5 route with ESI field as active forwarding information. 






Is my understanding right? or Is there something important I have missed?






Best Regards


Bob