Same OSPF area different physical locations

[dan hale]

Hellos All,

Need some OSPF advice...I inherited a transitioning network...originally per my attached drawing we had Site A (area 0 which is is our datacenter) and Site B Area 1.

We acquired a second site that we are brining into our datacenter and its a small network...I know design wise its good to have multiple area's however, the site is small and we are looking to bring a few more /24 and possible smaller /25 into our datacenter.

For ease can we keep it the same area as we bring in the next 3 locations?

I might add the third location we plan to bring in is doing EIGRP so I will have to redistribute EIGRP into OSPF this site is a bit larger...and my plan was to redistribute and slowly try and transition IDF's to OSPF....my thought was having the same Area might ease that pain.

Thanks,

Dan

[Georg Pauwen]

Hello,

your design is perfectly valid. The only area that preferably needs to be contiguous is area 0, the backbone (although there are workarounds).

I found a pretty good explanation on why multiple areas can have the same area ID (1, in your case):

https://lpmazariegos.com/2016/02/07/interconnecting-ospf-discontiguous-areas/#:~:text=Having%20discontiguous%20non%2Dzero%20areas,the%20neighbor%20relationship%20in%20OSPF.

[paul driver]

Hello
You shouldn’t have discontinuous ospf area's as in your the design above having area 1 in two locations separated by the backbone area may work but it is not recommended long term but you can have differing areas, And having multiple areas is a form of summarization for your network and can provide flexibility to incorporate other IGPs from those area's as your network grows

[Joseph W. Doherty]

I believe what you propose will work provided you're not using summarization of (the "separate", but same number) area addresses that overlap.

Whether it's a good idea might depend on whether you ever intend to interconnect those same numbered areas.  If not, I don't know if reusing the same area number makes it "better" for your transitioning.  In fact, since it's unexpected, it might further engender confusion.  For example, we have a problem in area 1.  Oh, which area 1?

[paul driver]

Hello 

please review this -

https://tools.ietf.org/html/rfc2328#page-38

[Richard Burts]

@paul driver Thanks for an interesting link. It reinforces that there is not any inherent problem for the original poster to have multiple instances of area 1. There are considerations, as mentioned in some responses in this discussion, about summarization and/or overlapping IP address ranges. From a management perspective it might be desirable to have unique area numbers for each area so it might be easier to keep track of what is where. But from a functionality perspective it could work without problem.

[paul driver]

Hello 

@Richard Burts TBH i was quite surprised with this rfc i have been always on the understanding that it was  a definite no no to do a discontinuous ospf design but the rfc states different however i would say some routing anomalies could arise if this was a prolonged design 

This i why i like these forums always learning !

[Joseph W. Doherty]

Paul, yea, but the RFC doesn't recommend it and it notes, I believe, the same possible issue I did, although I'm unsure it's clear, i.e. "However, in order to maintain full routing after the partition, an address range must not be split across multiple components of the area partition."

To be clear, by splitting an address range, I believe they are referring to summarization.

For example, if you split 192.168.3.0/24 into 192.168.3.0/25 and 192.168.3.128/25, those two networks could be in different numbered areas, or the same numbered area, partitioned or not, the issue is, other than being in the same non-partitioned area, you couldn't use any summary of them such as 192.168.3.0/24.

BTW, Georg's reference explains why this works, simple because the ABRs don't pass the non-backbone area routes, to the backbone, with an area ID.  Basically, they just advertise their area's routes (or summaries).

Also BTW, on the subject of always learning, and being surprised, I was quite surprised when I learned OSPF's cost, as defined by the RFC, isn't tied to bandwidth. 

Like you, learning is one of the reasons, I too, like these forums.

[Richard Burts]

Paul a lot of people share your original opinion that a partitioned area 1 (or any other non-zero area partitioned) would be a no no. Joseph makes a good point that the rfc says that this is permitted but does not say that it is recommended. I believe that we would all agree that it is better if each OSPF area has a unique area ID, but that in circumstances where that is not the case it is not a problem. 

I certainly agree with you and Joseph that we continue to learn new things through our participation in the community and that is one of the main reasons why we continue to participate.