BGP Multiple exit point via MPLS

[MEDIAOS]

R4 has multiple ways to reach R1. How to make the patch through R3 the prefer path?

[Flavio Miranda]

BGP is all about that. 

Use MED - Multi exit discriminator 

[MEDIAOS]

I do not control the PE routers (This is the ISP - MPLS).

[Georg Pauwen]

Hello,

I think for any sort of custom route selection within the MPLS cloud, you need to work together with the ISP. As you said yourself, you do not control the PE routers, so the way it is right now, you cannot influence what happens within the MPLS cloud...

[paul driver]

Hello @Georg Pauwen 
MPLS cloud is irrelevant, that is just a transit path between sites, All TE can be accomplished from the respective sites , just need to understand the current setup in terms of bgp peering and redistribution 

[paul driver]

Hello
You dont show it has multiple ways, We just see a single link into the mpls cloud, can you elaborate on what R4 is actually peering with and how many adjacencies does it has active?

sh ip protocols
sh ip bgp summary
sh ip ospf neighbors

[MEDIAOS]

This is something I am trying to lab before moving into production. I make the diagram clearer

*R4 is a CE and it is peering with VZ MPLS network.

*R3 is has a point to point-to-point connection with R1.

We are trying to avoid a loop condition. 

Traffic from R4 to R1 has multiple paths.

1) MPLS - R3 - R1

2) MPLS - R1

3) MPLS - R2 - R1

The question becomes:

*How to I make sure that there is no asymmetric routing when routing traffic from R4 to R1?

*How do I control the path.

[paul driver]

Hello

---

@MEDIAOS wrote:

We are trying to avoid a loop condition. 

Traffic from R4 to R1 has multiple paths.

---

Can you confirm.
Is R1-R3  ibgp or ebgp peers to each ohter?
Are you mutually redistributing between ospf/bgp on R1/R2?

[MEDIAOS]

* R1/R3 are running eBGP (Sorry diagram was not clear)

* Yes, I am mutually redistributing between ospf/bgp on R1/R2?

[paul driver]

Hello

Thanks for the feedback ,So it seems R3  is connected to R1 for resiliency, And R1/ R4 IGP is also, I assume the ospf is just the one area ?

R3 
Advertise just local routes to both of its ebgp peers, this will negate that rtr from becoming a transit path for mpls and R1, you can use an aspath access-list for this.

As for egress traffic, you could use two conditional defaults with the mpls being the primary path, if this is not applicable then utlize bgp weight preference for the preferred egress path

R1-R4
Advertise just local routes their ebgp peers, again this will negate those rts from becoming a transit path for mpls
Also to avoid any race condition incase of mpls outage to either rtr, suggest append a higher weight percentage on their ebgp peering so the ebgp path is always preferred upon convergence of bgp/ospf

As for R1-R4 igp you could create an access-list for each others local routes and advertise them into the mpls cloud with a less preferred preference value (as-prepend) 

R4 - not applicable

[MEDIAOS]

Hi, can you shed more light on the 2nd point - I don't understand what you propose here. You said R1-R4 did you mean R1-R2?

[paul driver]

Hello

Yes sorry R1-R2.

---

@MEDIAOS wrote:

Hi, can you shed more light on the 2nd point - I don't understand what you propose here. You said R1-R4 did you mean R1-R2?

---

When you have redistribution on a rtr that has external and internal routing process running such as bgp/ospf, then its possible that after an outage to the bgp peering the redistributed ospf routes will still be preferred by the rtr over its received external routes even though your ebgp peering is up external traffic would route via the rtrs igp link (path)

This is due to redistributed routes into bgp have a default weight value appended to them (32768) and the same prefixes coming in over the bgp peering wouldn't and as bgp weight value is a part of its best path selection process its possible the redistributed routes will be preferred.

So to negate this you can set a weight value towards the rtrs ebgp peer to be higher than 32768

Example:
route-map MPLSROUTES
set weight 40000

router bgp x
neighbor x.x.x.x route-map MPLSROUTES in

[Georg Pauwen]

Hello,

unless you have this already resolved, can you post the full running configs of R1 through R4, in order to be able to lab this up ?

[MEDIAOS]

I should have provided a thorough view of what I am trying to accomplish. I attached a new diagram (All links in red are new/no config yet).

*in the new diagram, I will be connecting all my through a MPLS network for redundancy (I do not have the config yet).

*For the legacy or current network:

R1/R3 are running eBGP

R3/R3/R5 are running OSPF area 0

R4/R6 are running eBGP

R5/R7/R8 are running OSPF area 1

*When I connect the sites through the MPLS network, I plan to do redistribution BGP <-- OSPF vice versa in R3/R5/R7/R8

*The question becomes:

1) how do I put this network in place for full redundancy without creating loops and asymmetric routing.

2) How do I keep Asia traffic and Asia and Europe traffic in Europe (unless a link is down).

For example, each site on the MPLS network has 4 ways to get to Asia – How do I make sure that the incoming and outgoing traffic are taking the same path.