With Prepending, adding next-hop-self chooses iBGP over eBGP

[Hamed Karimi]

Hey Guys

In the below diagram, I have path prepending between R2 and R3. But still, R2 for reaching the 8.8.8.8/32 chooses R3 as it is eBGP.

But when I add next-hop-self on R1 and R2,  R2 choose R1 to reach 8.8.8.8.  Why?

R1#sh run | sec bg
router bgp 100
bgp log-neighbor-changes
neighbor 10.10.10.2 remote-as 200
neighbor 10.50.50.2 remote-as 100

----------------------

router bgp 100
bgp log-neighbor-changes
neighbor 10.50.50.1 remote-as 100
neighbor 20.20.20.2 remote-as 200

----------------------

R3#sh run | sec bg
router bgp 200
bgp log-neighbor-changes
network 8.8.8.8 mask 255.255.255.255
neighbor 10.10.10.1 remote-as 100
neighbor 20.20.20.1 remote-as 100
neighbor 20.20.20.1 route-map a out

----

route-map a permit 10
match ip address prefix-list a
set as-path prepend 200 200 200

Here is the output:

After adding next-hop-self:

[Alex Pfeil]

The route to 8.8.8.8 through router 1 is not accessible to router 2 until you place the next-hop-self command on the neighbor statements between 1 and 2. This is because router 1 does not have a route to 20.20.20.2 and router 2 does not have a route to 10.10.10.2.

Router 3 is sending a route to router 1 with the next-hop of 10.10.10.2.

Router 3 is sending a router to router 2 with the next-hop of 20.20.20.2.

Router 2 has a route for 20.20.20.2 and it is therefore the best.

When you add next-hop-self, router 1 sends update to router 2 saying to get to 8.8.8.8, go to 10.50.50.1.  Router 2 can get there now, and it updates the routing table.

Hope this explanation helps.

Thanks,

Alex

[paul driver]

Hello

---

@Hamed Karimi wrote:

Hey Guys

In the below diagram, I have path prepending between R2 and R3. But still, R2 for reaching the 8.8.8.8/32 chooses R3 as it is eBGP.

But when I add next-hop-self on R1 and R2,  R2 choose R1 to reach 8.8.8.8.  Why?

The reason looks like being the next-hop to that prefix, As before you added the next-hop self command it was R1's external subnet being advertised and R2 didn't have a valid reachable next-hop for that subnet but it did after next-hop self was added as it then used the connected subnet between R1-R2.

As a valid reachable next-hop is required for a route to be included in the rib this will take precedence of weight in path selection

R2
show ip route

[Hamed Karimi]

After adding next-hop-self, if I remove AS prepending on R3, R2 will choose R3. 

So in brief, With having AS prepending on R3 and next-hop-self on R2 and R1, R2 will choose R1, but if I remove prepending or next-hop-self, it will choose R3. I was expecting to always choose R3 as it is eBGP over iBGP.

[Meheretab Mengistu · Accepted solution]

Hi Hamed,

On BGP best path selection criteria, Weight > Local Preference > Locally originated path > AS Path > Origin Type > MED > eBGP over iBGP > ... In you case, R2 learned about 8.8.8.8/32 from both R1 (iBGP) and R3 (eBGP). When you prepend as-path, the iBGP route is preferred (as far as the next-hop is reachable from R2 -- which is resolved by adding "next-hop-self" command) over the eBGP route.

Please look at the following link for BGP Best Path Selection: https://www.cisco.com/c/en/us/support/docs/ip/border-gateway-protocol-bgp/13753-25.html

HTH,

Meheretab

[paul driver]

Hello

---

@Hamed Karimi wrote:

After adding next-hop-self, if I remove AS prepending on R3, R2 will choose R3. 

So in brief, With having AS prepending on R3 and next-hop-self on R2 and R1, R2 will choose R1, but if I remove prepending or next-hop-self, it will choose R3. I was expecting to always choose R3 as it is eBGP over iBGP.

---

Neighbor type is quite far down the path selection of bgp if that what your asking but it was selected before you added the next-hop-self command because the as-path didn't come into play until then due to a non valid reachable next-hop?

1) removing prepending and no next-hop-self - and all things equal the result will be R2-R3 neighbor type
2) Adding just pre-pending =  R2-R3 neighbor type due to unreachable next-hop
3) Adding just next-hop self = R2-R3 neighbor type
4) Adding pre-pending and next-hop-self  =R2-R1 shortest as-path (as seq) next-hop now reachable

Next hop (reachable)
weight (highest) 
local pref (highest)
locally originated routes
as-path
origin 
med (lowest) 
neighbor type 
igp metric (lowest)