EIGRP/PBR

[jimfixit]

I have a client with a main office and several branch offices. Each office connects to main office via two MPLS networks. One is "primary" the other is backup. They wish to update speed on backup and load balance with failover as part of that. We use bgp across the MPLS and EIGRP on each lan. That works and we have load balancing and fail over from that.

However now they want to expand on this by "forcing" some traffic to always prefer one MPLS net over the other, mainly because that particular provide allows QOS through the cloud where the other does not. But they still want to failover to the less prefered MPLS net in case of an outage.

So, Right now I have two equal cost paths in my route table under EIGRP. I could force one to be the prefered by changing metrics or AD and that would cause it to be prefered and still fail over. BUT they only want it to be prefered for SOME traffic.

It sounds like a job for PBR but...can you do that with EIGRP? When I read up on PBR I'm seeing a way to change next hop but what would happen if that hop went down?

 

[Minue]

Hello
You can give it two next hops!That's the way we have configured on our Routers that use PBR.
Regards

 

[plshlpme]

ive never played with it.. but you can use soem tracking features with pbr

http://www.cisco.com/univercd/cc/td/doc/product/software/ios124/124cg/hirp_c/ch20/h_pbrtrk.htm

 

[jimfixit]

excellent, thank you both. I found the docs on multiple hops and that's a "go".

Next question: is there a way to redistribute ONLY internally leanered EIGRP routes into BGP rather than all of them? I'm getting suboptimal routes with redist from eigrp into bgp but without it; don't get all the routes I need. Can use ACL but hate to do that as I'd have to code a line for every route I don't want and then if the IPs ever change, there I am changing them again. I see little benefit to that over static routes....

is there a code to redist only the routes learned via internal eigrp into bgp? or is that only controllable via the eigrp AS number?

 

[Minue]

Hello
You will not have much options with redistribute eigrp.So I think the way to go is with access-list and route-maps for tweaking the metrics.
The suboptimal routes is in BGP?What is BGP using for routing decision.AS hop,weight,local preference?The problem could be here.
Regards

 

[jimfixit]

I changed nothing from the default settings. But after trying different stuff I found it's better to inject from BGP into eigrp and NOT back the other way. I'm using static routes to get the few routes in EIGRP that I need to go back the other direction. It's cleaner to set up, if harder to maintain.

I have another question about eigrp but, It's best for a new topic as others may not see it burried in here....

 

[Igymann]

I am actually trying to do the same (NO PBR) but have BGP be my primary circuit of communication and my EIGRP (DS3 ppp) be my secondary. Any ideas are welcome.

 

[Minue]

Hello
Be careful how much BGP routes you inject in EIGRP,this protocol could go crazy.
Regards

 

[jimfixit]

I agree with that statement. The main thing is, our provider is not doing this over traditional ISP it's a true PIP so we only see the bgp routes that either they need for the group of PE's we are assigned to or that we put there with our network statements in the bgp config.

I do note on our ISP service when I look at BGP routes it's filled with stuff we don't need to know about. I keep meaning to call them on that.....

 

[plshlpme]

the provider has created your own vrf.. so you will only see your own routes.
the other routes that you see that you shouldnt would most likely be their management ips and wan ips between the other CE and PE routers..

they should be able to put a distrbute list on the redistribution and block those routes from being passed to you.

i dont know why thats not standard on all of these installataions

 

[jimfixit]

well, whah...

no joy...well almost no joy.

I have my PBR working...from the router at least. I set up the policy on int vlan 1 then also globally with the local policy statement...which does not set it universally for all interfaces but does set the policy to be in effect for traffic generated on the router (or in this case switch). As I traceroute to addresses on the switch across the other side of my routers, I can see by the path taken that specific routes in my route-map go through the next-hop specified; stuff not in the route-map is load balance. Terrific.

Next part: from a workstation attached to the switch. It never seems to load balance when it should. The tracerts to the stuff in the map take the next hop as specified but the path to things not in the map seem to just "pick a route" and they use that same route as naseum no matter how many times I trace to it...it never seems to balance....

I thought that the interface I was testing from was perhaps not in vlan 1 where my policy is applied but alas this is not the case...