DC redundancy with BGP

Solved
suneq
Getting noticed

DC redundancy with BGP

Hi,

 

I have a simple topology, as show in the diagram below :

 

suneq_0-1626568677008.png

 

  • 1 spoke with 2 hubs : priority of hub 1 is higher than hub 2
  • eBGP between hub 1 and Core 1
  • eBGP between hub 2 and Core 2
  • iBGP between hubs and spoke

 

I would like to understand how the traffic from the DC to the Spoke will prioritize the Hub 1 in this scenario.

I learned from the article https://documentation.meraki.com/MX/Networks_and_Routing/BGP#Route_Advertisement_Behavior

that routes learned from the Spoke by the Hub2 will have an additional ASN pre-pended but if I understand correctly it does not garantee that the traffic from the DC to the Spoke will prioritize the Hub 1: if the traffic arrives on Core 2 and there is not iBGP between Core 1 and Core 2, it will be sent to the Hub 2 and then to the Spoke (except that there is iBGP between hubs that I am not aware of).

 

Could you please advise?

 

Thanks a lot.

 

1 Accepted Solution
GIdenJoe
Kind of a big deal
Kind of a big deal

Hey you have the logic backwards.

 

The ASN prepending happens from Meraki to the DC router.

So if a certain spoke has hub1 as first prio it will take that path for a certain subnet.

However the spoke network will be advertised via eBGP to both DC's but DC-2 will have a second ASN prepended.

 

Of course this is only important if you have duplicate IP space in both DC's.

So if you would have a scenario that a spoke sends traffic to dc hub 1, but the end host is at dc 2 then that traffic will have to be tromboned using the DC interconnect that would have to be in place.  In that case return traffic will also first return to DC-1 before being received by the dc 1 hub for encapsulation in SD-WAN.

View solution in original post

4 Replies 4
ww
Kind of a big deal
Kind of a big deal

There are active iBGP sessions between the spoke MX and both concentrators.

 

Why there is no ibgp between the as 6500 cores

suneq
Getting noticed

Hi @ww , thanks for your reply, our client does not have any link and therefore no iBGP between Core 1 and Core 2. It's a bit strange but it's their design. 

GIdenJoe
Kind of a big deal
Kind of a big deal

Hey you have the logic backwards.

 

The ASN prepending happens from Meraki to the DC router.

So if a certain spoke has hub1 as first prio it will take that path for a certain subnet.

However the spoke network will be advertised via eBGP to both DC's but DC-2 will have a second ASN prepended.

 

Of course this is only important if you have duplicate IP space in both DC's.

So if you would have a scenario that a spoke sends traffic to dc hub 1, but the end host is at dc 2 then that traffic will have to be tromboned using the DC interconnect that would have to be in place.  In that case return traffic will also first return to DC-1 before being received by the dc 1 hub for encapsulation in SD-WAN.

suneq
Getting noticed

Hi @GIdenJoe ,

Thanks for your explanation. I will ask the client to add a back-to-back link and iBGP between 2 Cores.

I wanted to be sure that I undertood the logic before asking the client to change their design.

Thanks again.

Get notified when there are additional replies to this discussion.