BGP setup for spine and leaf

[carl.townshend]

Hi Guys

What is the typical and recommended BGP setup for typical spine and leaf networks, would it be EBGP between the spine and leafs? would each leaf be in the same or a different ASN than the other leaf? 

would you peer with loopacks?

would it load share the traffic across all the spines ?

does anyone have a simple diagram to show it?

Cheers

[M02@rt37 · Accepted solution]

Hello @carl.townshend 

As I concerned Spine and Leaf  are all in the same AS! So iBGP neighborship! --> Spine as Route-Reflector...

Multipathing activated too ! I use loopacks routed by my UNDERLAY control-plane...

You UNDERLAY Control Plane is also another question to answer...

[carl.townshend]

Hi, this is really helpful, thankyou.

So this is the best practice design then?

I assume the route reflector will be part of the forwarding path between each leaf then ? and it will use OSPF to get to the bgp next hop ?

[M02@rt37]

In that context, spines act as route reflectors for Leaf, but they are still in the forwarding path only because of the physical topology, not because the are RRs. Underlay still provides next hop reachability, while iBGP carries the overlay routes !

Some admin prefer eBGP context because iBGP add complexity (RR design, path visibility).. depend of your skills, your context ... I read a lot before work on my own design with my own purpose.

[M02@rt37]

-> https://nwktimes.blogspot.com/2018/05/vxlan-part-vii-vxlan-bgp-evpn-control.html

[paul driver]

Hello
evpn vxlan is obviously the overlay control plane function this can use either eBGP or IBGP but i would say mostly the latter is used within a single site and eBGP via border gateways (BGW) for multisite interconnection.
As for the underlay - bgp-ospf-isis can be used to provide the end-end reachability of all L3 nodes consisting within the spine-leaf architecture 
We support aruba vxlans and for the underlay in those ospf provides the underlay