BGP routing

[zulqurnain]

Hi Experts

I have a scenario in which I am kind of stuck and wondering if anyone can give me the best advice on how to accomplish it

Basically I have following scenario and what I am trying to achieve is create a high availability in case one physical connection goes down. 

Site 'A' and 'B' connected over a Fiber connection on 6500 series switches on each side (primary link)

Site 'A' and 'B' has secondary connection from different ISP with BGP 

users in site 'A' connects over fiber to users on 'B' and if this link goes down I would want users to connect via secondary connection and only than 

so what would be the best way to accomplish this.

Internally we have EIGRP between both switches and network 

I was thinking of configuring conditional advertisment in BGP to make it work but I get stuck in a thought that I would have to convert the internal routing from EIGRP to iBGP and than make it work

Also I cannot use any other protocol because we have other sites connected to Site A users via this secondary ISP with BGP and I cannot advertise the same subnet on both side 'A' and 'B' 

Thanks in Advance

[jj27]

There might be a better alternative so I will let the rest of the community speak up, but you could change the BGP distance to something greater than 90 for the neighbors.  Your routing table would include EIGRP routes with administrative distance of 90 and eBGP routes of a higher administrative distance.  If the EIGRP (fiber) connection goes down, the routing table would only consist of eBGP routes and flow as expected.

[e.ciollaro]

Another way to change administrative distance of BGP route is the following command:

network ... mask ... backdoor 

It change the admin dis of the route to 200; this way the router will not use BGP route but EIGRP route (obviously it's useful if you have few network and  change rarely)

Enrico

PS: please rate if useful

[zulqurnain]

Thanks johnston and ciollaro 

I think your input is valuable but I as I said that in this scenario site 'A' which is also connecting other sites via ISP with BGP and currently is the only means for other sites to connect to site 'B' over through the Fiber link is using EIGRP

Therefore these remote sites connect to site 'B' via Site 'A' ISP with BGP 

So having said that I also have EIGRP redistribution into BGP at Site B secondary circuit so that in a case of fiber failure , Site 'A' and remote sites can still connect to site B networks.

hope that makes sense 

[Robert Falconer]

Do you have a basic drawing that shows the connections and routing protocols? That might help us with suggestions.

[zulqurnain]

Hi

I found this post on the forum and its literally the same scenario of what I have . But from the posts I cannot understand what did the person at the time really did to solve this. 

https://supportforums.cisco.com/discussion/12322816/redistribution-bgp-and-eigrp

In any case I am experiencing the same problem that eigrp route after the failover take place in route but once the link comes back online , even than eigrp route remains in route table. 

unless I manually clear the ip eigrp neigbour   

[e.ciollaro]

Hi zulqurnain

I think the problem is due to a mutual redistribution and/or a routing feedback but it's hard to say without the information I asked in the previous post. Probably when BGP takeover somewhere an eBGP route is installed in the routing table and later, when the EIGRP route it's received, it's preferred to the original EIGRP route  because of the admin distance.

Let me know,

enrico

[zulqurnain]

after all the time , everything seems to be going in vain !!! 

any advise 

if the DF is cut , everything flow over the backup link (MPLS) but when DF is back , site B retains the EIGRP routes in RIB and it mess up the site A and other sites connectivity i.e. via MPLS.

when they want to reach Site B .. instead of using Site A and DF to site B . They start coming over this backup link MPLS to Site B 

if I clear the bgp route and eigrp route than everything comes back to regular state 

[Jon Marshall]

-

[zulqurnain]

Hi Jon

I hear you what you saying but what I am trying to convey is that I cannot seems to control eigrp which being redistributed on site B meaning that once eigrp route has been redistributed into bgp and advertised. Once the primary DF comes back up and Site A starts to re-advertise all routes correctly than site B once receiving the same route via BGP does not install/override that eigrp installed routes in RIB.. 

which is by nature of these protocol correct ..

[zulqurnain]

finally solved it 

thank you everyone 

[Robert Falconer]

Can you post 'show ip route' output for each of the following:

- Normal traffic flow

- Dark fiber down

- Dark fiber back up before resetting any routing protocols

[e.ciollaro]

The scenario is not very clear, I think your network is something like the one in attached draft but:

1. which kind of network have you got between A and B ? MPLS ?  IPSec VPN ?
2. redistribution between EIGRP and BGP is just in site B ?  Is it bidirectional or just EIGRP to BGP ?
3. Which are the BGP neighbourship  between A, B and branch ? eBGP or iBGP ? How is neighbour to how

enrico

[zulqurnain]

Hi E.Ciollaro

1. Between A and B , primary is Dark Fiber and back is MPLS (BGP with ISP)

2. Redistribution of EIGRP into BGP at site A because all remote sites connect via it and also reach to site B through it using primary connection of Dark Fiber. Now doing redistribution of EIGRP into BGP at site B because if DF is down than site B can only be reached via MPLS (BGP with ISP)

3. eBGP (not sure what you mean but 'how is neighbor to how)

Also I am neither on site A and B doing BGP into EIGRP . on Site B I am using route-map to set EIGRP local-preference 10 and weight 40 when EIGRP is redistributed into BGP and in BGP config it's weight is 40000

Still as i said BGP route has EIGRP redistributed routes from Site A redistribution but if I clear bgp or disconnect DF than EIGRP route gets inserted in table but once connection is back up or BGP neighour comes up that EIGRP doesn't get removed until I clear EIGRP 

The AD's for both protocols is at default 

[e.ciollaro]

Hi

 

item 1,2 and 3 are clear but I didn't understand what I mean with "Also I am neither on site A and B doing BGP into EIGRP . on Site B I am using route-map to set EIGRP local-preference 10 and weight 40 when EIGRP is redistributed into BGP and in BGP config it's weight is 40000"

Let me know if I understood correctly:

1) no redistribution from BGP to EIGRP (neither in site A nor in site B)

2) redistribution of EIGRP to BGP both in A and B using a route-map that set Local Preference to  10 and weight to 40 (question: in both same LP 40 and weight 10 or did you configure differente values in A and B ?)

3) what do you mean with "in BGP config it's weight is 40000" ? which routes or has this weight  ?

I suppose that the problem is due to timing in the routing advertisement, something like this:

• router receive EIGRP route and inject in BGP before it receives form MPLS
• router receives BGP advertisement form MPLS network but prefer local inject route
• when fibre is disconnect EIGRP route is removed and router put BGP route form MPLS in routing Table
• when EIGRP route comes back it is not put in routing table because of worst administrative distance

This is just an idea, I'm waiting your feedback to  check if it is correct or not

Bye,

enrico