Mutual Redistribution

Netstudent

Okay so I'm trying to brush up on my advanced routing protocol topics like redistribution. Anyways I'm reading through this cisco article and it is talking about mutual redistribution problems when more than one router is performing redestribution.

In this example there are 2 routers doing mutual redestribution and they are redistr. RIP and IGRP. It is trying to explain problems that can occur and how to solve them. Anyways to answer this question, you will probably have to skim through the article real quick.


So basically what is happening here is IGRP is redis. RIP into the IGRP domain. IGRP has a better AD than RIP. What the article is saying here is once the RIP routes are redis. into IGRP, then the updates bounce back to the routers performing mutual redis. and those routers believe that the new route through the IGRP domain is the correct path to get to the RIP domain because IGRP has a better AD. Thus creating a routing loop. Then they go on to talk about how to use a distribution list to stop those updates from going back to the routers performing redis. so that the correct routes can stick in the routing table. Again this is going to seem vague untill you see the diagram and skim through the config.

So here is where my question comes in. Why in this example did Split-horizon NOT do the exact same thing the distribute-list is doing??? Shouldn't split-horizon stop those IGRP updates from going back to those routers doing redis.?

So is split-horizon aware of the route administrative distances coming in? So if an external route came in on an interface, updates can still be sent back out that interface from which it was learned because the routes being sent back out are a different AD than the external routes that came in?



http://www.cisco.com/warp/public/105/redist.html

Here is the link, it's actually a pretty good article. The diagram and config is exactly half way down the page. the title is "Avoiding Problems Due to Redistribution"

This should be a good topic for you guys who are tired of answering "What does /24 mean?"

dtlokee

The concepts of mutual redistribution can be complex and require a lot of thought to understand what's going on.

First, split horizion prevents a router from re-advertising a route on the same interface it was learned from, if the route is being redistributed from one protocol to another split horizion would not apply.

The problem generally starts because some routing protocols make no distinction between internal vs external (redistributed) routes with regards to their administrative distance. For example, say you have 2 routers connecting part of your network that is using rip and part of your network using ospf. Onc the routes are redistributed from rip into ospf on R1 they will assume a new administrative on all the ospf routers of 110. When this route is advertised to R2 though ospf it will now accept these routes as the best route (because RIP has an AD of 120) so R2 will now route all traffic for the RIP part of the network through OSPF (sub optimal routing)

You can fix this through applying a distribute list to filter updates, a route-map to apply a higher AD or route tags that can be filtered on later, or modifying the AD of routes.

Netstudent

Okay I get that. Thanks man. I was waiting on you.

r_durant

Net...we are doing something similar here at work, but it's kind of unavoidable (is that a word?)...

We have a RIP network and we are migrating to EIGRP and the first site we migrated, did cause some routing loops, I wish I had see this article earlier...

So basically we had to create some access lists so that some routes won't be distributed to some of our sites and those that are, be distributed where they are supposed to until the full migration to EIGRP is complete....

We did one site and basically learnt our lessons the hard way, the other sites should be ok from here...but as dtlokee says, it seems pretty complex, but it's interesting because of the point i'm at in my CCNA study...

RD

Netstudent

Cool! I wish i was doing stuff like that. As a system admin I can't even look at our gear. Not literally, but all routers are owned by our ISP, nuvox. Our WAN infrastructure is so simple I don't even think we need a routing protocol. before we went to frame-relay, we had nothing but P-T-P connecting our sites. Even then, we only had static routes pointing to the places data needed to go.

r_durant

Well, we still outsource support & implementation for "major" projects, but I'm trying to get my boss to allow us (well, me) to do that stuff...It's only me now that deals with the networking stuff, the other guy is on a 15-month project...

It's no fun being out at 1 and 2 am and being a spectator...while an engineer is playing with all the "toys", i like to get my hands dirty...

We're on frame-relay now, a main site and a DR site, with 8 other sites connected, kind of a hub and spoke setup, redundant links at all sites...if the main link fails it should switch to the DR and be routed back to the main site...pretty cool....

The way i do it, i troubleshoot until it's above my head, then i call in support...i figure there's more than one way to skin a cat!!

Netstudent

there you go...But I totally understand what you mean by having to be a spectator after hours. Experienced that myself a couple times.

Do yall have redundant PVC's on subinterfaces? Or do yall like switch over to DDR for redundancy? wasn't sure what DR was

tech-airman

Netstudent wrote:

there you go...But I totally understand what you mean by having to be a spectator after hours. Experienced that myself a couple times.

Do yall have redundant PVC's on subinterfaces? Or do yall like switch over to DDR for redundancy? wasn't sure what DR was

Netstudent,

I think r_durant was referring to "Disaster Recovery." It's basically a fully operational backup of the main site.

Netstudent

ohhhhh ok.

r_durant

Tech-airman is correct...DR = Disaster Recovery

Basically 2 PVC's...one to the main site and one to the DR site

Primary link runs RIP and the DR link has static routes with a higher cost, like 150...so if the main link drops, the traffic is routed thru the DR site to the main site...

Netstudent

oh ok i gatcha. Thats pretty cool.