EIGRP conversion to OSPF enterprise wide- Need your advice

itdaddy

Hi guys,

I have never done this but I understand how to program OSPF though. Okay. I have seen many models for pushing out OSPF for networks but they are the perfect toplogies I only see. Many networks like mine at my works has this.

We have 2 WAN clouds. VPLS cloud and an MPLS cloud.
at each branch site and main site we VPLS and MPLS as 2 choices to get to each site.
The MPLS cloud has CER and PERs and are on one private AS number for BGP.
The VPLS cloud is layer 2 access. If you run command show arp you see all the routers MAC address in the cloud.

I have been tasked to make lab that mimics our setup for conversion of all sites that run only IGP EIGRP to OSPF

Area 0 is for our Main site IGP and the rest are on different areas.

I have questions:

1. can you run OSPF along side by side of EIGRP for the conversion and when you cut over just say do it slowly one site at a time until perfected? at each site. I think you can but wanted to ask best practice what people do???

2. I want to have each site with totally stubby areas I think that is wise to reduce route tables sizes. Is this good???

3. I can make the entire OSPF network across the VPLS and MPLS clouds right??? Right over top of BGP???? is this common?
with no Route redistribution right? just straight OSPF neighbor routing right?

At each site i have a path selection mechanism
for EIGRP presently the delay value to prioritize the metric for best path first. VPLS is delay 100 and MPLS is delay 120. works great. Each site employs this.

4. with ospf i use the ip ospf cost command to do the same effect on path selection at each branch site off the core L3 switch.each branch has a Core switch setup L3 to VPS or MPLS wan options.

I am going to speak to our CCIE route switch guy. I am sure he will tell me everything but wanted your guys's advice am I on the right track. I am labbing in GNS3. I feel I got a handle on it.

5. VPLS looks easy one L2 cloud but what about MPLS with CER and PER Same AS cloud for everything it seems to be iBGP in cloud but across Cer to Per it is eBGP. so do I have to make virtual links across on ABRs at each branch across ABR to from Cer thru per to other edge routers? this part confuses me the connection from eBGP to iBGP how to connect to area 0?

shortstop20

1. Yes, but you have to keep in mind that the most specific route (regardless of AD) will be used.

2. Yes, stubs are a good choice if these are hub sites.

3. BGP must have a route to its Neighbor whether it be via static route(not default) or another IGP. So don't build your OSPF with the idea that it should be over the top of BGP. OSPF needs to be able to stand on its own. BGP can then use OSPF to build its Neighbor relationships and exchange routes.

daveyb

Firstly, I would do away with the idea of multiple areas, and stub areas. With modern equipment (5+ years old is sufficiently modern) reducing your routing table has very few benefits. A simple flat area



Second, it sounds like you are going to run into trouble with your MPLS link. OSPF must be running on every router along a path in order for it to work. It sounds like there is a CPE that you may be running BGP with? It sounds like you would need to distribute between OSPF/BGP to get that bit working. Virtual links are used to connect a none area 0 network to the backbone across another none area 0 network. IE [Area 1] <-> [Area 2] <-> [Area 0]. The vitual link would run across area 2.



Based on your limited information, I would probably do something like the following:

Run a distinct OSPF instance at each site. In each site only install loopbacks and point-to-points into OSPF.

Run a full iBGP mesh at each site to carry all other routes

Run a BGP mesh (or maybe assign routers at 2 different sites as route reflectors - depending on number of sites) across all the routers that terminate the VPLS connection. You may want to use a distinct AS from your site AS numbers.

Use BGP metrics (MED/Local pref) to steer traffic across each link as you see fit.

fredrikjj

why? what are you hoping to achieve?

itdaddy

convert our site IGP from EIGRP to OSPF over the enterprise. run it both over vpls and mpls wan routers
one big OSPF network

OfWolfAndMan

If you're using MPLS L3VPN, there is actually an extended BGP community that can pass along OSPF routes as type 3 LSAs (Regardless if there's an area 0 on point a and point b), while the MPLS infrastructure acts as a super backbone. Here is a decent link to it.
https://www.racf.bnl.gov/Facility/TechnologyMeeting/Archive/06-30-04-CISCO/Using-OSPF-in-MPLS-VPN-Environment.pdf

As for the part about converting, you have two options. Use a moving boundary (Mutual redistribution with some tag filtering or sometimes single direction redistribution with some route modifications) or running two IGPs and raising AD of EIGRP after both routing protocols in place.

Since moving boundaries tend to require a bit more complexity, depending on current setup and topology, if your hardware can handle it, just run dual IGP and pull out EIGRP after raising AD. I don't know your node/link count, but anything greater than 1 OSPF area is questionable. You need a good chunk of nodes/links before considering adding a bunch of areas.

itdaddy

or running two IGPs and raising AD of EIGRP after both routing protocols in place.




thanks man this is the best option. my brain hurt when you mentioned the later. hahahah
but you the man. thank you for stopping by and helping me man!