Re: EIGRP - Split Horizon

Howdy All,

I am signing off for the night but I've received a couple of unicasts
pointing out what can be perceived as weaknesses in my below post, so
I thought I would take one last stab to clear up my intention. The
below quote (never advertise a route out of the interface through
which you learned it) applies to a simple FR topology (depicted in the
link) - which is different than the topology originally in question -
and is thus a technically correct definition of Split Horizon. I
wouldn't argue that. The problem I have with it is that it comes off
as an >exclusive< definition of EIGRP Split Horizon. And one that's
difficult to square - if interpreted literally - with a slightly more
complex topology such as was depicted in the original post of this
thread. I like the definition found in the Command Ref better: "The
split-horizon rule prohibits a router from advertising a route through
an interface that the router itself uses to reach the destination."
Much less ambiguous, IMHO, and it allows for the behavior observed by
the OP.

Nighty night (MST) all ye GSers!

Scott

On Jan 8, 2011, at 5:19 , Scott M Vermillion wrote:

> Sorry about that Jules! Hate to cause rather than to clear confusion.
>
> The below quote of my earlier post was relative to a more
> "traditional" (and not entirely technically correct in this particular
> context!) definition of Split Horizon that a lot of us probably grew
> up with. For example, check out this old Cisco EIGRP link:
>
> http://www.cisco.com/en/US/tech/tk365/technologies_white_paper09186a0080094cb7.shtml#splithorizon
>
> "Never advertise a route out of the interface through which you
> learned it."
>
> I think the OP was expecting that R3 and R4 would take a literal
> interpretation of this and one would wind up advertising the R1 Lo0
> route to the other and then the "learning router" (my term) wouldn't
> be able to be advertised back in the other direction towards the
> advertising router. That's not, in fact, what happens, as you point
> out in your post. The rationale for my having said that in the first
> place was to attempt to sum up what I believed to be the point of
> confusion. And in doing so apparently generated even more! ;-)
>
> Apologies,
>
> Scott
>
>
> On Jan 8, 2011, at 5:07 , jules NYA BAWEU wrote:
>
>> Scott you got me all confused with this Split Horizon thing - please
>> feel
>> free to weight in with more details if you could. Here is my take on
>> this:
>> What Dave is seeing is a normal behavior of Split Horizon. R3 and R4
>> would
>> advertise R1's Lo0 to each other as long as those routes pertain to
>> their
>> path that connect them to R2 - you will only see one route in your
>> routing
>> table unless you have a variance set up or they are exactly equal
>> path
>> routes, but you would see all those routes in your EIGRP topology
>> table. Now
>> with EIGRP, they would not be able to advertise the route back to
>> their
>> interface that they have picked as best path direction - even though
>> that
>> router was learned from a different interface - in this case,
>> assuming that
>> R3 picked R2 as its best path, R3 would not advertise the path via
>> R4 to to
>> R2.
>>
>> My issue in your prior is: "and then the learning router would not
>> advertise
>> R1's Lo0 back to the other neighbor"
>>
>> Please correct me if I'm wrong.
>>
>> Thx
>>
>>
>> Blogs and organic groups at http://www.ccie.net
>>
>> _______________________________________________________________________
>> Subscription information may be found at:
>> http://www.groupstudy.com/list/CCIELab.html
>
>
> Blogs and organic groups at http://www.ccie.net
>
> _______________________________________________________________________
> Subscription information may be found at:
> http://www.groupstudy.com/list/CCIELab.html

Blogs and organic groups at http://www.ccie.net
Received on Sat Jan 08 2011 - 19:09:21 ART