[c-nsp] IS-IS route separation/filtering

[Oliver Boehmer (oboehmer)]

Jared Gillis <> wrote on Thursday, August 06, 2009 20:48:

> Daniel Verlouw wrote:
>> On Wed, 2009-08-05 at 15:02 -0700, Jared Gillis wrote:
>>> Hm, interesting though. Unfortunately, it doesn't seem to pan out
>>> in the lab. The LSPs don't seem to get flooded, but the routes do
>>> get passed through Router A to all the stub routers, regardless of
>>> how I set up the mesh-groups. 
>> 
>> right. Mesh-groups block only LSPs, CSNPs would still be flooded.
>> 
>>> This is almost what I'm trying to do, there will be very few routes
>>> in IS-IS, but the decree from on high is that each stub router
>>> should be totally stubby =( 
> 
> Mostly due to longevity, planning for the worst case of high growth,
> IPv6 deployment, etc that will make each route in our routers very
> costly over time. Also, given our topology, there's no reason for the
> stub routers to learn anything but default.

Well.. not sure how large you want to grow your L1 area, but you could
investigate "advertise-passive-only" to only adveritse the loopbacks
(all customer routes should be in BGP if you need to plan for growth),
and you'll be fine, even with a 1000 nodes in the area. And if you reach
this number, address summarization (and the implications of it) will
become an issue (even with OSPF)..

> It's looking like we might have to run OSPF on this, but we'd really
> rather stick with IS-IS. It seems that OSPF's ability to put
> individual interfaces into different areas might be the required
> feature that forces us that way. That is, unless anyone knows a way
> to put an IS-IS router into different areas aside from assigning
> multiple NET addresses... 

No, doesn't work with Integrated ISIS (only CLNS allows you to use
different ISIS areas on a single node)..

	oli