Re: ISIS and connected networks

From: Carlos G Mendioroz (tron@xxxxxxxxxxx)
Date: Thu May 02 2002 - 22:40:40 GMT-3


   
Baety Wayne 18 CS/SCBX wrote:
>
> Routing protocols won't advertise the same network on to the source
> network... see below...
>
> R1----10.0.1/24----R2
>
> R1 will _never_ advertise 10.0.1/24 to R2, likewise R2 will never advertise
> the same to R1. You can think of this as internal split horizon, except
> there's no knob to turn this off. This obviously doesn't apply to LSP type
> routing protocols which share complete information about network topology
> with each node. In any case, this would rarely be useful to do, except to
> inform routers that happen to have more than one network configured for a
> particular segment. In practice, even in the case of secondary networks, R1
> and R2 do not advertise information about connected networks unto the
> connected network, even to other secondary networks configured on the
> interface. Even if they did, the information would never be installed in
> the routing table since connected networks have a higher administrative
> distance than next hop style advertisements. But let's play around with
> this scenario just to make sure you completely understand the implications.
> Let's say R1's address is 10.0.1.1 and R2 is 10.0.1.2 and a host on that
> segment has an address of 10.0.1.3. Let's say that R1 advertises the fact
> that it can reach 10.0.1/24 to R2 and R2 advertises the same to R1 and they
> both install the route into their respective routing tables. Let's also
> suppose a packet arrives with a destination of the host 10.0.1.3. R2 (or
> R1) would look in its routing table and realize that network 10.0.1/24 is
> reachable by R1 (or R2). It would send the packet to R1 which would make
> the same determination about R2. They would bounce the packet back and
> forth amongst themselves until a comet from outer space smashes into them or
> you pull the plug. Clearly, not a useful thing to do =)
>
> But to make things even weirder for you... see below...
>
> R1-----10.0.1/24
> Ospf
> Rip
> IS-IS
> EIGRP
> etc...
>
> R1 will never advertise the connected 10.0.1/24 to any other routing process
> via redistribution <routing protocol x>. You __ALWAYS__ must use
> redistribute connected, no exceptions. For the same reasons noted above,
> connected networks are treated special.

Wayne,
I'm not sure I understand what you are saying here...

Is that "no protocol will advertise local connected networks to other
protocols via redistribution" ?

If that's what you are saying, I think it is not true.
In fact, I'm doing this just fine from ospf into isis with no trouble.
(and it is also an example situation of indirect connected
redistribution in
http://www.cisco.com/warp/public/105/redist.html)

There seems to be an issue with isis advertising this kind of networks.
Someone in this list attributed it to some kind of loop prevention in
isis
implementation, but who knows ? (Fact is that isis is not doing it, but
ospf, for example, is).

Hope this adds to the confusion, out of which we can drive conclusions
:-)

>
> These are common misconceptions and well talked about in the archives.
>
> WAYNE BAETY, MCSE, A1C, USAF
> Network Systems Trainer
>
> > -----Original Message-----
> > From: Jason [mailto:jgraun@attbi.com]
> > Sent: Thursday, April 25, 2002 1:01 PM
> > To: ccielab@groupstudy.com
> > Subject: ISIS and connected networks
> >
> > I was wondering if anybody has found a way to get the connected network
> > between two routers to enter the ISIS database and get sent to other
> > routers. For example
> >
> >
> > R1---12net---R2----23net----R3
> > ISIS L2 OSPF
> >
> > When I redistribute ISIS into OSPF on R2 I see any other ISIS routes
> > behind R1 but I don't see the 12 network which is connecting the two of
> > them. I have seen this on both Frame Relay and Ethernet, I have used a
> > redistributed connected matching a route-map to get the network into the
> > table but is there any other way?
> >
> >
> > Thanks
> >
> >
> > Jason



This archive was generated by hypermail 2.1.4 : Thu Jun 13 2002 - 10:58:49 GMT-3