redistribution BGP into EIGRP (metric working weird)

[ecouto]

Hi,

I'm having some problems to understand an output from the EIGRP topology. I attach the config of router 1 and 2 also the "show ip eigrp topology..." output

Basically, I got a router peering eBGP with an MPLS provider and EIGRP internally, I got the same in a Secondary location but using a different MPLS, internally between my site I have direct links so I can send traffic from site A and B internally.

To avoid asymmetric routing I need to tune the EIGRP metric of the routes that I receive from BGP, for this I increase the bandwidth and decreases the delay. In BR1 and I can see the delay/bandwidth change but in BR2 is doesn't make any difference.

Someone did ever cross something like this?

Both router the same hardware and running, c3845-advipservicesk9-mz.124-15.T7.bin

Cheers,

Emilio

[johnspaulding]

Can you please provide just a simple diagram.

Does it look something like this

LAN---EIGRP--BR1--BGP----MPLS-----BGP---BR2---EIGRP--LAN

When you mean direct link are you talking about you also have a connection (EIGRP) between them as well?

In you config I see a few things that might look like mistakes when you made the notes but if you could clarify

router eigrp 103

redistribute bgp 64201 route-map BGP_TO_EIGRP

network 10.20.30.0 0.0.0.255

no auto-summary

router bgp 65201

no synchronization

bgp log-neighbor-changes

redistribute eigrp 104 route-map EIGRP_TO_BGP

neighbor 10.10.11.2 remote-as 3400

neighbor 10.10.11.2 timers 10 30

neighbor 10.10.11.2 soft-reconfiguration inbound

no auto-summary

The redistribute eigrp statement in bgp says AS 104 not 103? Im taking this as a yes from you show ip eigrp topology but just wanted to make sure.

Is your end goal to make the EIGRP routes the preferred? since you have a direct link internal as well? If you could please clarify I would be glad to help you.

[ecouto]

Hi John,

See my ASCII art below :)

LAN--EIGRP(101)--BR1--BGP--MPLS1-BGP/EIGRP--NET_PRI

| |

EIGRP(105) EIGRP(xx)

| |

LAN2--EIGRP(103)-BR2--BGP--MPLS2-BGP/EIGRP--NET_SEC

Yes, is a typo, is not 104 is AS103

My end goal, is to reach NET_PRI using BR1 from either LAN1 or LAN2 and reach NET_SEC using BR2 in normal state, if MPLS1/BR1 goes down NET_PRI should be reached using BR2 and vice-versa.

The main issue is either in the route-map and metric parameters (delay/bandwidth) but the show eigrp topology display different results.

Let me know if isn't clear

Cheers,

Emilio

[ecouto]

My ASCII art didn't work....

I attach a .jpg

[johnspaulding]

Thanks that helps alot.If you dont mind could you just draw out a quick layout thats more visual please..I think I have the right layout from your post but just want to be 100% so I dont give you the wrong information. thanks

[ecouto]

what about something like this?

[johnspaulding]

Thats alot better :) give me a few minutes to look over

[ecouto]

Thanks.

I forgot to say that LAN1 and LAN2 are advertised in both router (BR1 and BR2) for redundancy

Also NET_PRI/SEC will manually tune the metric to make the return traffic for LAN1/2 symmetric.

[johnspaulding]

Inside LAN1 and LAN 2 network device inside EIGRP 105. Is it load balancing between the two path? which routes are showing preferred IE: to reach NET_SEC which MPLS router does it go out to?

[johnspaulding]

Here is a better idea than trying to deal with EIGRP metrics for path selection:

1.) On the NET_PRI MPLS router you would advertise the individual network prefixes out to BR1

- 10.100.0.0/16, 10.102.0.0/16 and advertise the subnets inside of this (ex.10.100.1.0/24,10.100.2.0/24)

- When this gets advertised out to BR1 and redistributed into EIGRP 105 it will pick the longer match route via BR1 for NET_PRI networks

2.) On the NET_SEC MPLS router you would than advertise the individual networks inside 10.101.0.0/16 and 10.103.0.0/16 networks out to BR2

- When this gets redistributed into EIGRP 105 the longer match route for those networks will be through BR2 for NET_SEC networks

3.) If there was every a failure the advertisments would be deleted from EIGRP and it would pick the summary route that you have here

NET_PRI 10.100.0.0/16

NET_PRI 10.102.0.0/16

NET_SEC 10.101.0.0/16

NET_SEC 10.103.0.0/16

If you implement this you would get the following results

- From LAN 1 to route to destination NET_PRI ---> BR1

- From LAN 1 to route to destination NET_SEC ---> BR2

- From LAN 2 to route to destination NET_PRI ---> BR1

- From LAN 2 to route to destination NET_SEC ---> BR2

- If there was a failure in the MPLS cloud the networks would be match based off the summary routes you are adverting on the least prefered MPLS path.

Example

- You would still advertise the summaries for NET_PRI from MPLS provider 2

- You would still advertise the summaries for NET_SEC from MPLS provider 1

This would be better off in the long run since it will allow you to have more control over the routing and control over your traffic.

[johnspaulding]

sorry if went off track but by making a recommendation but thats not what you looking for. If you could paste the show ip eigrp topology for the networks behind the BR1 and 2 devices so I can see which direction the traffic flow is going. What I think might be happening and why your not seeing the metric change for the route is because is preferring the eigrp route being redistributed from BR1. But thats just an assumption at this point because I need to see more of a view from the other devices. thanks