cancel
Showing results for 
Search instead for 
Did you mean: 
cancel
258
Views
1
Helpful
3
Replies

SDA LACP with two EN

mikhailov.ivan
Level 1
Level 1

Hello colleagues! There is a theoretical (or not) question. We know that in the DC design if we need to use a multihoming connection from endpoints (servers or something) to two independent leaf switches - we can use vPC or multihoming ESI or call it whatever you want for redundancy and speed purposes and it works in EVPN vxlan or ACI , right? Let's imagine that we have the same aim but within the SD-Access fabric with LISP and VXLAN, how can we solve this issue? Conditions: we can't use the stack, we can't use a L3 from the endpoint side and we have to utilize both links. There is a simple server with 2 NICs which supports either LACP or static LAG and we want to plug it into 2 independent SDA edge nodes (like c9300 something). For the server it has to look like a simple LACP. What can we do in this case? After couple of days googling I didn't find neither confirmation or denial of this design. Let's be prepared for the worst case if Cisco answers that "sorry it isn't supported", and what will heppen underhood if we plug a server anyway into 2 independent EN ports with the same L2 VNI ? Thanks!

 
 

 

 

1 Accepted Solution

Accepted Solutions

Hello @mikhailov.ivan  ,

A single port-channel stretched across two independent SD-Access Fabric Edge Nodes is not supported today, our LISP implementation will need enhancements to make this use case possible. We do have some roadmap which you can discuss under NDA with your sales team. Right now all I can suggest is to use StackWise, or StackWise Virtual, or active/standby uplinks, or connect the server outside of the SD-Access Fabric.

 

 

 

View solution in original post

3 Replies 3

jedolphi
Cisco Employee
Cisco Employee

Hi @mikhailov.ivan , it is supported to dual-home an endpoint to two Fabric Edge Nodes if the endpoint presents traffic on only one port at a time. It's not possible to have the same MAC/IP address concurrently active on two different Fabric Edge Node ports. Does that help? Best regards, Jerome

 

So, it means that "sorry it isn't supported"

Is there any reason except of it wasn't implemented by a software team?I mean in xvlan-evpn we can use ESI with route type 1 and 4 for this case, can't we use something with LISP ?

Thanks!

Hello @mikhailov.ivan  ,

A single port-channel stretched across two independent SD-Access Fabric Edge Nodes is not supported today, our LISP implementation will need enhancements to make this use case possible. We do have some roadmap which you can discuss under NDA with your sales team. Right now all I can suggest is to use StackWise, or StackWise Virtual, or active/standby uplinks, or connect the server outside of the SD-Access Fabric.