A Strategic Approach to Redistribution

Category: By Barooq
I am getting better at redistribution or at least thinking so.

I complete LAB 7 ( yes I am not doing labs in order) and Lab 6 uptil BGP.And though redistibution gave me a headache a i spent like 2 hours on each task, I did them right. What more, I did those in my own way, which is quite different from the Solution Guide.

After sweating a lot , I think I am slowly adopting a strategic approcah towards redistribution.

The way I see it is that we have to have a method before starting redistribution.I used to ignore suboptimal paths and try to tag everything, but now I have realized that going through a step by step process, eliminating loops and suboptimal paths at each step is the way to go.

Here is the methodology that has been working for me

1) Identify one of the routing domains as core routing domain. It doesn't matter which protocol it is. For me its the largest domain, which normally sits in the centre of topology.

2) Identify if there has to be redistrbution between non core or edge domains.If so, do that first. Again making sure that if two domains have more than one rendezvous point, they keep preferring their native routes. ( Doesn't matter if redistrbution is done on a single point or two. Avoid routes to get back in

3) Ensure connectivity on the edge domains.Also ensure that native paths are preferred.

4) Now redistribute between core and edge domains.Single point of redistribution first, and verify connectivity.Again, make sure that native routes are preferred. (AD or tagging)

5) If there are more than onr points of redistribution between core domain and edge domain, tag at one and deny the routes getting back at the other point and vice versa.

6)Make sure that before redistribution, if there was a 'redistribute connected' command used to advertise a loopback or a connected link, you modify it so that no interface that is supposed to be redistributed in left out.

For example R1 is a border router between OSPF and Eigrp. R1 runs EIGRP at serial and OSPF at ethernet. If I advertised R1's loopback in EIGRP through redistribution at an earlier stage, then during OSPF->EIGRP redistribution the ethernet interface won't be advertised, which should be as its running OSPF.

Workaround is simple, go to the route-map and make changes according.

What this approach does?

Well first of all we'll have connectivity between the edge domains.Then we'll inject routes from core into edge domains, making sure that core domain, within itself prefers its own routes and also edge domains prefer their own routes.There will be no loop (hopefully) and routing will be optimal, as each domain is preferring its native routes. Plus if we encounter any loop, it'd be easy to find out after which step the loop is occured.

It'd be easy to troubleshoot and worst case scenario, we can eliminate the step that resulted in a loop maintaining end to end reachability.

There is some problem with my Visio.I'll make a topology later and explain the process, through that. I could use an internetworkexpert topology, but I am not sure that'd be all right due to copy right issues. Though all I want to do here is borrow a diagram :S

0 comments so far.

Something to say?