From: Olopade Olorunloba (lolopade@ipnxnigeria.net)
Date: Thu Oct 27 2005 - 20:23:57 GMT-3
Look at the command as giving you more flexibility with respect to the kind
of adjency you form with each neighbor. More especially that if it is not
necessary to have IPv4 session to the neighbor, then you do not have to.
In the situation, the POP router still has an IPv4 session to other upstream
ISPs, hence Internet is still carried on the global routing table. What was
not desired was that if the Internet link at the POP goes down, for the
Internet based traffic to go through the iBGP neighbour (i.e. the other
POP). By exchanging IPv4 routes with the other POP, this will be the default
behaviour. Depending on the configuration the remote POP could be preferred
as the exit point.
The solution was to filter routes so that IPv4 are not received on the iBGP
session. I'm now saying that it would have been easier to just disable the
IPv4 session.
Another instance you might want to disable the IPv4 is if you are using a
small router as a PE to offer MPLS VPN service. Lets assume the router does
not have enough memory to hold the Internet table. Assuming the PE is
located in a stubby part of your network, it could use a default route in
its IGP to reach the internet, while still holding BGP VPNv4 for MPLS VPN
services.
Really, I just think it is more flexibility and gives you the designer more
options of what you could do.
Regards
Learning from me, you tell me your school and let me come and enroll.
-----Original Message-----
From: Andrew Lissitz (alissitz) [mailto:alissitz@cisco.com]
Sent: 27 October 2005 23:57
To: Olopade Olorunloba; swm@emanon.com; The Great Ryan; Cisco certification
Cc: comserv@groupstudy.com
Subject: RE: Different Command Format on BGP routing process
In your example, how would this stop the pop routers from getting to the
internet? A global route can be present in a vpnv4 update...
I think I see what you are referring though Olopade, from a general
perspective I can agree with what you are proposing.
Tell me if I am correct: stop ipv4 routes (internet routes) and you stop
access to the internet. However, for any mpls vpn customer who needed
access to the global routing table for internet access, how did you
provide this?
Did you provide separate links from and to the customer, one to other
vpn sites and one to internet? The link to internet connected to a diff
PE that needed ipv4 neighbors? Or did you redistribute a global route
into a vrf that pointed to a P router?
If you disable default ipv4, and you already have ipv4 neighbors
configured, this will not stop these neighbors from sharing routes. IOS
will simply modify your config, add a ipv4 address family and keep
going... business as usual.
I am interested in learning from you Olopade, if you can say what you
did, please do. Kindest regards,
-----Original Message-----
From: Olopade Olorunloba [mailto:lolopade@ipnxnigeria.net]
Sent: Thursday, October 27, 2005 5:49 PM
To: Andrew Lissitz (alissitz); swm@emanon.com; 'The Great Ryan'; 'Cisco
certification'
Cc: comserv@groupstudy.com
Subject: RE: Different Command Format on BGP routing process
If the peers will only exchange VPNv4 routes, why spend more time
filtering, when a single line can disable the IPv4 session.
-----Original Message-----
From: Andrew Lissitz (alissitz) [mailto:alissitz@cisco.com]
Sent: 27 October 2005 02:38
To: Olopade Olorunloba; swm@emanon.com; The Great Ryan; Cisco
certification
Cc: comserv@groupstudy.com
Subject: RE: Different Command Format on BGP routing process
I hate to say it, but I am not sure I totally follow what you are
saying.
If you wanted the routers to not learn or advertise particular routes,
then using communities or filtering was definitely the right thing to
do.
If these routers did not need to have ipv4 neighbors, then disabling
default ipv4 would have been ok. You can still propagate a global
internet route into a vrf, and thus populate it to other routers. So
disabling the default behavior of BGP would not have helped here. Some
sort of filtering is still the answer.
Sorry if I do not fully understand the environment you managed and my
comments are not inline with what you saw / thought.
-----Original Message-----
From: Olopade Olorunloba [mailto:lolopade@ipnxnigeria.net]
Sent: Wednesday, October 26, 2005 7:59 PM
To: Andrew Lissitz (alissitz); swm@emanon.com; 'The Great Ryan'; 'Cisco
certification'
Cc: comserv@groupstudy.com
Subject: RE: Different Command Format on BGP routing process
Well, like you will say, some of these things are for theories.
Though, I've seen a situation in which I would have used the command
(unfortunately, I did not know the command then).
I have two POPs dual homing to the Internet. They are in the same AS but
bandwidth on the InterPOP connection is limited, so this should not be
used to carry Internet base traffic whatsoever. The POPs are also used
to deliver MPLS VPNs, hence they are VPNv4 neighbors. To ensure that
Internet access is not carried on the Inter POP site, I used the
no-advertise community to stop the advertisement from going out to the
other POP on the iBGP session.
I think now that disabling the IPv4 session would have been a quicker
and faster way to achieve it.
-----Original Message-----
From: Andrew Lissitz (alissitz) [mailto:alissitz@cisco.com]
Sent: 27 October 2005 00:34
To: Olopade Olorunloba; swm@emanon.com; The Great Ryan; Cisco
certification
Cc: comserv@groupstudy.com
Subject: RE: Different Command Format on BGP routing process
Hey Yall,
Please feel free to show me when you have seen other behavior besides
this.
If you disable ipv4 default behavior with the command you mention, then
you will need to create an address family for ipv4, ipv6, vpnv4, etc...
If you want to have a 'normal' ipv4 session with another bgp router, you
would need to create the ipv4 address family and then add the neighbors
etc... to this address family.
If you use this command and turn off default ipv4 behavior, and you
currently have ipv4 neighbors configured, then the configs will be
modified and placed under a ipv4 address fam for you. You do not need
to modify the configs, just enable the no bgp default ipv4 command and
you will see your config modified correctly for you.
A question for the group ...
In production or lab practices, have you seen a scenario / real world
problem that required you to turn off the default ipv4 behavior? I can
imagine if you have a pure vpnv4 network, and your PEs only speak vpnv4
with route reflectors and PEs then you could turn this off... but again,
how much memory / processing are you saving by not having this left @
the default? Perhaps the bgp scanner will run less...? Pure ipv6
environment perhaps?
What do you think group? Have you see this command needed in either
practice or production?
-----Original Message-----
From: Olopade Olorunloba [mailto:lolopade@ipnxnigeria.net]
Sent: Wednesday, October 26, 2005 6:47 PM
To: Andrew Lissitz (alissitz); swm@emanon.com; 'The Great Ryan'; 'Cisco
certification'
Cc: comserv@groupstudy.com
Subject: RE: Different Command Format on BGP routing process
On Cisco IOS where the IPv4 family does not show by the default, simply
disable default IPv4 session by giving the command no bgp default ipv4.
This will cause the IOS to now show the IPv4 session, and the operation
will be like what Scott has described here, because after adding a
neighbor, you will still need to go to the IPv4 address family to
activate it.
Regards
-----Original Message-----
From: nobody@groupstudy.com [mailto:nobody@groupstudy.com] On Behalf Of
Andrew Lissitz (alissitz)
Sent: 26 October 2005 18:18
To: swm@emanon.com; The Great Ryan; Cisco certification
Cc: comserv@groupstudy.com
Subject: RE: Different Command Format on BGP routing process
Nice Scott ... Thanks for the perspective on how to view the configs /
address families.
Cisco .. Confuses people?
-----Original Message-----
From: Scott Morris [mailto:swm@emanon.com]
Sent: Wednesday, October 26, 2005 1:08 PM
To: Andrew Lissitz (alissitz); 'The Great Ryan'; 'Cisco certification'
Cc: comserv@groupstudy.com
Subject: RE: Different Command Format on BGP routing process
No difference actually. IPv4 is the default address family used. Some
newer versions of the IOS automatically put the "address-family ipv4"
stuff there, I think just to avoid confusion with people using multiple
families.
In the R&S lab, the only one to be concerned with is the IPv6 address
family. In the SP lab, you may need the vpnv4 (MPLS VPNs) and/or
multicast as well.
All the address-family ipv4 part does is help to separate things so that
we
(humans) are less confused. Apparantly this hasn't been entirely
successful! (grin)
Always remember, there are two parts to the BGP configuration pieces.
Thinking of it this was (since BGP is an application not really an IP
routing protocol) will help you think things through!
There's first the "connections" piece which establishes your basic
connection to your neighbor. This will include the neighbor statement,
and update-source, remote-as, ebgp-multihop parameters and things like
that.
Then there's the RIB pieces. Each RIB is different. A single neighbor
may talk multiple RIB/update types. You may have 20 neighbors, and only
10 of them talk about IPv4 information. That's why each neighbor should
be activated in the appropriate RIBs they will share information with.
Then any filtering (RIB-based stuff) or network commands go under the
address families.
It's hard to follow when you're first getting into this MBGP stuff
because in generic, all-purpose BGP where we only run IPv4, it's all
just right there, not separated. But if we start thinking about the
pieces involved, then the address families suddenly aren't all that bad!
:)
Scott
-----Original Message-----
From: nobody@groupstudy.com [mailto:nobody@groupstudy.com] On Behalf Of
Andrew Lissitz (alissitz)
Sent: Wednesday, October 26, 2005 11:42 AM
To: The Great Ryan; Cisco certification
Cc: comserv@groupstudy.com
Subject: RE: Different Command Format on BGP routing process
Hey Ryan,
Afaik this command is used for inter-as mvpns (mpls vpn customers with
links in two or more ISPs).
There are different options for the different address families.
Curiosity... is this needed for a lab you are working on?
Once you type address-fam ipv4 (without a VRF at the end) you will be
working in the same default address family you have when you type router
bgp x. Now ... try getting rid of the command address-fam ipv4 ... once
you typed it, IOS really wants to keep it there. You will need to
either continue to work with this address-fam listed or remove bgp and
re-add bgp.
Strange how you can not get rid of this address-fam since it is the
default address family for bgp.
Does this answer your question Ryan?
-----Original Message-----
From: nobody@groupstudy.com [mailto:nobody@groupstudy.com] On Behalf Of
The Great Ryan
Sent: Wednesday, October 26, 2005 8:20 AM
To: Cisco certification
Subject: Different Command Format on BGP routing process
Hi Group,
There is any different result if I use the following to enable BGP ?
Router(config)# router bgp 65412
Router(config-router)# neighbor 192.168.0.1 remote-as 65412
Router(config-router)# neighbor 192.168.0.1 ebgp-multihop 255
Router(config)# router bgp 65412
Router(config-router)# neighbor 192.168.0.1 remote-as 65412
Router(config-router)# address-family ipv4 Router(config-router-af)#
neighbor 192.168.0.1 activate Router(config-router-af)# neighbor
192.168.0.1 ebgp-multihop 255
I found something very interesting. the command "neighbor 192.168.0.1
next-hop-unchanged" will never display in the first format but it takes
effect when I use some show command. I'm afraid that if I write erase
the configuration and paste the configuration again, it will miss this
command.
Any have this experience before ?
Thanks !
Ryan
This archive was generated by hypermail 2.1.4 : Sun Nov 06 2005 - 22:00:54 GMT-3