cancel
Showing results for 
Search instead for 
Did you mean: 
cancel
5507
Views
0
Helpful
12
Replies

EIGRP and BGP routing

MachadoGB
Level 1
Level 1

I am using EIGRP and BGP redistribution into each other protocols.

It is working as designed.

But I am facing a routing issue:

- I have some remote sites using static routes on point to point links. the static routes point's to the remote LAN subnets.

- A routing issue is happening when one of the remotes goes down (the link goes down) for few hours and comes back.

- My end router has static facing the remotes and EIGRP facing the WAN routers.

- The WAN routers are running EIGRP and BGP (eBGP to the PE and IBGP between the two routers).

When the site goes down the IBGP injects the remote subnetwork into the routing table.

When the remote comes back the routing table does not release the IBGP entry and use the EIGRP entry.

So there is a loop for that subnetwork between the two IBGP WAN routers.

Is there any solutin for this?

12 Replies 12

Hi,

   Please post the detailed topology and configurations on related devices.

Toshi

Hi Toshi,

Here is the basic configuration for my primary router. The secondary is exactly the mirror.

The CENTRAL_BRANCH prefixes are the ones in the EIGRP.

My guess is that I have to remove the redistribution of EIGRP into BGP  (at BGP config) and only advertise the prefixes in the iBGP peers.

note the iBGP peers ate to the physica interfaces.

!
router eigrp 1
redistribute bgp 61234 metric 100000 1000 255 10 1500 route-map NO_DEFAULT
network 10.x.a.40 0.0.0.3
network 10.x.y.1 0.0.0.0
network 10.x.y.16 0.0.0.15
maximum-paths 8
distance eigrp 90 201
no auto-summary
!
router bgp 61234
no synchronization
bgp router-id 10.x.y.1
bgp log-neighbor-changes
bgp graceful-restart restart-time 120
bgp graceful-restart stalepath-time 360
bgp graceful-restart
bgp redistribute-internal
bgp suppress-inactive
network 10.x.y.1 mask 255.255.255.255
network 10.x.y.12 mask 255.255.255.252
network 10.x.y.16 mask 255.255.255.240
network 10.x.y.253 mask 255.255.255.255
redistribute eigrp 1 route-map CENTRAL-BRANCH_ROUTES
neighbor 10.x.y.13 remote-as 703
neighbor 10.x.y.13 description eBGP Peer to PROVIDER
neighbor 10.x.y.13 version 4
neighbor 10.x.y.13 timers 7 21
neighbor 10.x.y.13 soft-reconfiguration inbound
neighbor 10.x.y.13 prefix-list CENTRAL-BRANCH_routes_in in
neighbor 10.x.y.13 prefix-list CENTRAL-BRANCH_routes-out out
neighbor 10.x.y.19 remote-as 61234
neighbor 10.x.y.19 description iBGP Peer to CENTRAL-72-B physical interface
neighbor 10.x.y.19 version 4
neighbor 10.x.y.19 timers 7 21
neighbor 10.x.y.19 next-hop-self
neighbor 10.x.y.19 soft-reconfiguration inbound
no auto-summary
!

!
ip prefix-list CENTRAL-BRANCH_routes-out seq 5 deny 0.0.0.0/0
ip prefix-list CENTRAL-BRANCH_routes-out seq 20 permit 10.z.0.0/16 le 32
ip prefix-list CENTRAL-BRANCH_routes-out seq 25 permit 10.x.0.0/16 le 32
ip prefix-list CENTRAL-BRANCH_routes-out seq 30 permit 10.c.0.0/16 le 32
ip prefix-list CENTRAL-BRANCH_routes-out seq 100 deny 0.0.0.0/0 le 32
!
ip prefix-list CENTRAL-BRANCH_routes_in seq 5 deny 0.0.0.0/0
ip prefix-list CENTRAL-BRANCH_routes_in seq 20 deny 10.z.0.0/16 le 32
ip prefix-list CENTRAL-BRANCH_routes_in seq 25 deny 10.x.0.0/16 le 32
ip prefix-list CENTRAL-BRANCH_routes_in seq 95 deny 10.c.0.0/16 le 32
ip prefix-list CENTRAL-BRANCH_routes_in seq 100 permit 0.0.0.0/0 le 32
!
ip prefix-list no-default seq 5 deny 0.0.0.0/0
ip prefix-list no-default seq 10 permit 0.0.0.0/0 le 32
!

!
route-map NO_DEFAULT permit 10
match ip address prefix-list no-default
!
route-map NO_DEFAULT permit 20
!
route-map CENTRAL-BRANCH_ROUTES permit 10
match ip address prefix-list CENTRAL-BRANCH_routes-out
!

Hi Machado,

From what I can see the problem is not with the redistribution .It's got to do with the AD of EIGRP and BGP. Allow me to explain it to you.

From your description, I can see that the remote sites are learned by your WAN routers via  eBGP and also EIGRP(static route to the sites redistributed into EIGRP) via the edge routers.Am I right? , If yes, then when they go down and come back up the WAN routers learn the subnets of remote sites via EIGRP and also eBGP. However, the routing table installs the route learned via eBGP because the AD is 20 which is less than EIGRP.

Also, looking at your eigrp config, you have actually made the AD even worse by giving external EIGRP routes and AD of 201.
When you redis your static on the edge router into eigrp, it will be seen as external eigrp route. So, your WAN routers will see the routes learned via EIGRP with an AD of 201 and eBGP with an AD of 20. Since 20 < 201, it will install that route in the routing table.

You will be able to see the remote subnets in the "sh eigrp topology " table. However, they wont be installed in the routing table

router eigrp 1

distance eigrp 90
201

HTH,

Regards,

Kishore

Okay, I agree.

But what if I remove teh distance command and leave teh EIGRP wiht internal AD of 90 and external AD of 170.

Will the external route be preferred for the routing table than the iBGP (ad = 200)?

Thanks.

hi,

if your  WAN routers are learning the remote sites via eBGP then the AD will be 20 which is less than 170 so the routes will be installed by BGP.

iBGP AD doesnt come into picture here.

HTH

Thank you Kishore,

I did resolve the AD problem by removing the EIGRP distance command. But the main problem still there.

When I loose one of my remote sites the /24 network for that site does not get removed from routing table of the two routers doing redistribution EIGRP into BGP and BGP into EIGRP. So when the site comes back the /24 network for that site gets a loop between the iBGP peers. It does not get the EIGRP update into the routing table and holds the iBGP information into the routing table. The way to stop the loop is by shutting down the iBGP peers.

I am studing a solution to remove redistribution from EIGRP into BGP or doing Site of Origin technique.

let me know if anyone have any idea - I think it is a known issue but I just cannot figure that out.

Hi Machado,

When you loose your remote site lets say ip address is 10.1.1.1/32 and comes back up, your WAN router sees the route via eBGP(AD 20) and eigrp(170) and installs it in the routing table via eBGP and doesn't install the eigrp route because a better route is already there with an AD of 20. Your eigrp-to-bgp redistribution doesn't work because for redistribution to work , it needs to have a route in the routing table learned by eigrp. But in your case the route is already installed by eBGP so if you do " sh ip bgp 10.1.1.1" and sh ip route 10.1.1.1 you will see it learns this from the PE of your ISP. However, this route will be in your eigrp topology table as I already mentioned.

A simple fix for this is to increase the admin distance of eBGP , so  under your router bgp process just put the command distance bgp 200 200 200

and the remote sites will always be learned via EIGRP. and when the link via EIGRP goes down, your WAN routers will learn this via eBGP and when they come back up , they will be learned via EIGRP because AD 170<200

HTH

Regards,

Kishore

Please rate if helpful

Hi Kishore,

I will consider your suggestion.

I will be preparing a solution and posting it soon. I will prepare a change request, going thorugh the explanation and solutions. So it will be at least 3 days before I come back with results.

I did the show commands below and the sh ip route on my CE routers 72-A and 72-B the output indicates the route is learned via EIGRP and redistributed via BGP. Whne I do the show ip bgp for the same prefix I have two entries, one from iBGP and one from Local (best path). When the /24 subnetwork is removed the entry for "Local" disapear and BGP route is the only one in the routing table. When the /24 comes back the "Local" entry does not apear unless I clear the EIGRP neighbors.

So, like I said above I will consider yur suggestion for next change. But I will also work towards elemination of EIGRP-to-BGP redistribution.

Many Thanks.

Hi Machado,

You see two entries in the " sh ip bgp" table on the WAN router because the wan router learned the remote prefix via EIGRP first and redistributed into the BGP table. The weight of the locally injected prefix is 32768 and the one learned via eBGP is 0. Hence, your local prefix wins and the route stays there. Once you shut the link via EIGRP, then the prefix learned via eBGP remains and when you bring the EIGRP link back up, redistribution doesnt work as I explained in my prev post.

HTH,

Kishore

Edit:

Let us know how u go

Hello MachadoGB,

>> I am studing a solution to remove redistribution from EIGRP into BGP

you can use BGP network commands under router bgp, instead of redistributing EIGRP into BGP,  BGP network command looks for the prefix in the IP routing table and works also for prefixes learned by other dynamic routing protocols.

the only limitation is that you cannot have more then 200 network commands under router bgp and if auto-summary is disabled you need an exact match (including prefix length) to have a match.

Your current scenario with two routers performing mutual bidirectional redistribution would require the use of route tags in order to be handled correctly.

Hope to help

Giuseppe

Hi Giuseppe,

I will consider your sugestion since the use of network commands will be a better solution han redstribution.

I am also investigating the possibility of a /24 subnetwokr from EIGRP be injected into BGP on my CE router 72-A and that same /24  subnetwork be added to EIGRP again via redistribution to my CE router 72-B.

Many Thanks

Today I did some tests and came up with the following:

- the remote site is up and running all the remote sites subnetwokrs are well know via EIGRP to the 7200 routers A & B.

- the subnetworks from remote site are removed from EIGRP routing process because of an outage for that particular site.

- that subnetwork is not removed from the routing table of the 7200 routers A & B. You can see that from the show ip route output and from the show ip bgp output of router A pointing to the interface on router B. and vice-versa.

- the remove site comes back and advertise new entries in the EIGRP. teir networks are back to the routing table of all devices except the 7200 routers.

- the 7200 routers still keep the output of routing table and BGP table pointing to each other.

- What is the common part between the two 7200 routers... on witch 3750 running IOS 12.2-46-SE.

have you seen something like that? I see if I can provide outputs tomorrow.

Getting Started

Find answers to your questions by entering keywords or phrases in the Search bar above. New here? Use these resources to familiarize yourself with the community: