RE: Cluster-ID and Originator-ID of group RR

A data plane loop means that an actual user's packets are looping around the
network indefinitely until they are dropped by mechanisms like TTL expired.
An example of this would be if you do a traceroute and you see the same couple
of hops over and over and over. Loops in the data plane are caused by bad
information being fed from the control plane. For example if two routers have
static default routes pointing at each other that would cause a loop in the
data plane.

A loop in the control plane is when routing or other control information keeps
getting passed back and forth between devices. An example of this would be a
routing loop due to errors in redistribution and administrative distance.
Loops in the control plane don't always cause problems though. For example
assume you have 3 routers connected in a triangle, R1, R2, and R3, and they
are all EBGP peers with each other. When R1 originates a route into BGP, it
will advertise it to R2 and R3. R2 and R3 will then advertise the same route
to each other. If R3 chooses to route to R1 via R2, then R3 will advertise
R1's route back to R1. This is technically a loop in the control plane
because the routing advertisement went from R1 > R2 > R3 > R1, however it
shouldn't cause any negative effects because BGP has mechanisms to prevent
this. Mainly in this case it would be loop prevention via the AS-Path
attribute.

The same could be said about other protocols, like OSPF for example. When you
flood an LSA into OSPF, there are many cases where your own locally originated
LSA will be flooded back to you. This is because OSPF doesn't implement
split-horizon, since it's not a distance vector protocol. This is okay
though, because OSPF has loop prevention mechanisms built in to figure out
that a self-originated LSA should be dropped instead of flooded when it is
received inbound.

In general loops in the data plane are bad, but (finite) loops in the control
plane are normal.

Brian McGahan, CCIE #8593 (R&S/SP/Security)
bmcgahan_at_INE.com<mailto:bmcgahan_at_INE.com>

Internetwork Expert, Inc.
http://www.INE.com

From: CCIE KID [mailto:eliteccie_at_gmail.com]
Sent: Monday, July 23, 2012 8:04 AM
To: Brian McGahan
Cc: jeremy co; Cisco certification
Subject: Re: Cluster-ID and Originator-ID of group RR

Hi Brian

U always use these terms Control Plane Loop and Data Plane loop

Can u brief me about which are the features avoid control plane loops and data
plane loops

Data plane loops are avoided by TTL only.. Is there any other way to avoid
data plane loops

Control Plane loops: Multicast RPF, Cluster ID, Originator ID, Duplicate LSA
from Sequence Number, ADv Router ID and Link ID in OSPF,

Add more to this..
On Mon, Jul 23, 2012 at 8:04 PM, Brian McGahan
<bmcgahan_at_ine.com<mailto:bmcgahan_at_ine.com>> wrote:
Even if you don't configure a cluster-id it is inherited from your router-id.
The only case where you really need to configure the cluster-id manually is
when you have multiple route reflectors in the same cluster. Specifically
this means that you would have two or more route reflectors that are servicing
the same set of clients, and are also peering with each other as either
clients or non-clients. When the route reflectors reflect the same set routes
to each other, they will drop the incoming updates if they share the same
cluster-id.

Really at the end of the day it doesn't matter how you configure it, because
BGP will eventually stop the looping of updates either based on the
originator-id or the cluster-id. The only difference is how far the update
message is passed around before it is dropped. Also remember this is just a
loop of control plane information, not of the data plane. In other words you
could basically configure a full mesh of iBGP peers with every router as a
route reflector, and each of their peers are clients, and the network would
eventually work itself out.

Brian McGahan, CCIE #8593 (R&S/SP/Security)
bmcgahan_at_INE.com<mailto:bmcgahan_at_INE.com>

Internetwork Expert, Inc.
http://www.INE.com

-----Original Message-----
From: nobody_at_groupstudy.com<mailto:nobody_at_groupstudy.com>
[mailto:nobody_at_groupstudy.com<mailto:nobody_at_groupstudy.com>] On Behalf Of
jeremy co
Sent: Monday, July 23, 2012 3:57 AM
To: Cisco certification
Subject: Cluster-ID and Originator-ID of group RR

Hi,

I understand that Originator-ID is a new way of detecing loop in group of
Route reflectors insteadof cluster-ID

So if I have a scenario of group of RRs, to my understanding there is no need
for cluster-ID to be configured under BGP process and originator-ID will take
are of it.

Is that a correct assumption ?

thanks.

Blogs and organic groups at http://www.ccie.net
Received on Mon Jul 23 2012 - 10:16:31 ART