Data Center


This community is currently under full moderation, meaning  all posts will be reviewed before appearing in the community. Please expect a brief delay—there is no need to post multiple times. If your post is rejected, you'll receive an email outlining the reason(s). We've implemented full moderation to control spam. Thank you for your patience and participation.

Expand all | Collapse all

VRF.inet.0 route into bgp.evpn.0 - CBR Design

Jump to Best Answer
  • 1.  VRF.inet.0 route into bgp.evpn.0 - CBR Design

    Posted 01-25-2021 15:19
    I have a scenario where I need to get a /32 host route that exist in a instance-type VRF (vrf.inet.0) but this same vrf serves as the L3GW vrf for the evpn-vxlan instance(instance-type virtual-switch...(this is a vlan-aware instance)).  I have all of the vxlan connectivity working inside the evpn-vxlan and the connectivity within the VRF is working, but I can not get this /32 host in the VRF to be accessible to the host inside the evpn-vxlan virtual-switch instance.  The irb interface (when sourced) can ping host in each instance but  not /32 host(vrf.inet.0) to host in evpn-vxlan instance(bgp.evpn.0).  Is there a "knob" i'm missing?  I've tried playing around w/ route-targets as well, but to no avail.  Any insight/thoughts/help would be greatly appreciated!!

  • 2.  RE: VRF.inet.0 route into bgp.evpn.0 - CBR Design

    Posted 01-25-2021 23:13
    We would need more information about your setup and configs to understand what you're trying to accomplish and help you resolve this issue, but did you have a look at rib-groups? They're meant to leak routes between routing tables

    Ridha Hamidi

  • 3.  RE: VRF.inet.0 route into bgp.evpn.0 - CBR Design

    Posted 01-26-2021 21:51
    Ridha...Thanks for the reply....The issue when I try using rib-groups is I'm taking moreless placeing L3-routes into an L2-table(and vice-versa).  I'm not sure if in my EVE-NG test environment that the vQFXs and vMXs are not behaving properly, but it doesn't appear rib-groups will help in this scenario.

  • 4.  RE: VRF.inet.0 route into bgp.evpn.0 - CBR Design

    Posted 01-25-2021 23:17
    Do you have vrf-table-label configured? 


  • 5.  RE: VRF.inet.0 route into bgp.evpn.0 - CBR Design

    Posted 01-26-2021 22:31
    Yasmin...thanks for the reply....I do have vrf-table-label configured for the vrf routing-instance.  I'm questioning if this is even possible to do as I've tried numerous different ways but to no avail.

  • 6.  RE: VRF.inet.0 route into bgp.evpn.0 - CBR Design

    Posted 01-27-2021 15:42
    Well, at a high level we have something like this: 

    Communication is possible between the hosts in the EVPN and the host behind R2, but the BGP session you see in the picture terminates inside the VRF instance.  

    We had to configure these commands under the routing instance, to allow route advertisements from the VRF to R2: 

    set protocols evpn ip-prefix-routes advertise direct-nexthop
    set protocols evpn ip-prefix-routes encapsulation vxlan
    set protocols evpn ip-prefix-routes vni 100
    set protocols evpn ip-prefix-routes export EVPN-host-routes

    Your setup , if I am understanding well,  is more like this: 
    Where the remote hosts network is learned as l3vpn routes.  I have not tested this, but I am wondering if some of the commands I am using under protocols evpn in the routing instance might help. 


  • 7.  RE: VRF.inet.0 route into bgp.evpn.0 - CBR Design

    Posted 01-27-2021 23:54
      |   view attached
    Yasmin...this is great stuff, thank you!!  I've tried to update the scenario drawing as best as I could  using your drawing.  The .150 host IP resides in the vrf (instance-type vrf) all while all others reside in the evpn-vxlan instance(instance-type Virt Sw.).  I'm needing to make the .150 host accessible from the evpn-vxlan hosts.  I have tried playing around rib-groups, import/export policies and had no luck.  I've started playing around w/ some of the evpn ip-prefix-routes statement and have been able to create and populate the vrf.evpn.0 table and populate with the Type 5 .150 route, as well as see the evpn-vxlan hosts show up as Type 5s.  In the vrf.inet.0 table I see the evpn-vxlan hosts(.2 and .4) as [EVPN/7]  routes.  I however can not seem to get the evpn-vxlan.evpn.0 table populated with the .150 host.  I feel I may be making progress .  Is there anything you have to do to get the Type 5s to appear as Type 2s or even just appear in the evpn-vlan virtual-switch instance?

  • 8.  RE: VRF.inet.0 route into bgp.evpn.0 - CBR Design

    Posted 01-30-2021 16:23
    I have been banging on this all week and have made some progress although I'm unsure if its correct or not.  I've managed to be able to get the learned .150 address from within the VRF instances placed into EVPN as a type 5 route and learned on the leaf.  I however can not ping the .150 from a host that's on the same subnet but in the EVPN-VXLAN instance.  Each host whether in VRF or EVPN-VXLAN can ping the gateway on the irb interface.

  • 9.  RE: VRF.inet.0 route into bgp.evpn.0 - CBR Design

    Posted 01-30-2021 22:51
    I'll take a look tomorrow, see if I can help. 

    I was thinking that it might be an encapsulation problem or a bug! I've had some issues with EVPN on QFX. 


  • 10.  RE: VRF.inet.0 route into bgp.evpn.0 - CBR Design
    Best Answer

    Posted 02-01-2021 10:56
    From your configuration this is what it seems you are doing, which honestly seems  a little strange  because the server is in the same subnet but is advertised by  a router via BGP? 

    In this scenario, wouldn't the hosts assume that is directly connected and ARP for the MAC address of Do the hosts have an ARP entry to reach  Or do the hosts have a route to pointing to 
    proxy arp on the spine might help here. 

    Also, try this: set routing-options forwarding-table chained-composite-next-hop ingress evpn (just a wild guess). This command seems to does some magic with pure type 5 routes - it broke a few things in my setup the other day, because suddenly I had a bunch of /32 routes that I was not expecting and longest match took over ;-)   


  • 11.  RE: VRF.inet.0 route into bgp.evpn.0 - CBR Design

    Posted 02-01-2021 11:00
      |   view attached
    diagrams attached, in case you want to edit it. 



    EVPN diagrams.pptx   49 KB 1 version

  • 12.  RE: VRF.inet.0 route into bgp.evpn.0 - CBR Design

    Posted 02-01-2021 11:15
    HOLY SMOKES!!!  It's working!!  PROXY-ARP, PROXY-ARP, PROXY-ARP!!!!  

    ping routing-instance S4-10 source
    PING ( 56 data bytes
    64 bytes from icmp_seq=0 ttl=64 time=209.060 ms
    64 bytes from icmp_seq=1 ttl=64 time=209.872 ms
    64 bytes from icmp_seq=2 ttl=64 time=218.420 ms
    64 bytes from icmp_seq=3 ttl=64 time=210.427 ms
    --- ping statistics ---
    4 packets transmitted, 4 packets received, 0% packet loss
    round-trip min/avg/max/stddev = 209.060/211.945/218.420/3.770 ms

    Now I got to figure out what configuration is unnecessary as I've configured so many different things in an attempt to make this work!! 

    YASMIN'S THE BEST!!!  :)

  • 13.  RE: VRF.inet.0 route into bgp.evpn.0 - CBR Design

    Posted 02-02-2021 17:09

    Glad I could help! 

  • 14.  RE: VRF.inet.0 route into bgp.evpn.0 - CBR Design

    Posted 02-01-2021 03:47

    Would you be able to clarify a little bit more what you are trying to achieve?

    I have looked over your config and it seems to me that you have everything pretty much there.
    You have your irb.10 in a VRF.  This is for a /24 route.
    Yet you seem to want a /32 route from that subnet installed into EVPN.  The question is "As what type of EVPN route?" Are you working on VTMO ?

    You should have a Type 2 route/s for the Host, but I cant see it listed.
    So if the hosts cannot speak to each other inside the same broadcast domain, then you may have an issue somewhere else.

    As far as I know you should have EVPN type 2 routes for the MAC and the MAC/IP combo of hosts inside that subnet.

    William Jackson