Re: iBGP based network. [9:2300] posted 12/10/2002
- Subject: Re: iBGP based network. [9:2300]
- From: "alok" <alok.dube@xxxxxxxxx>
- Date: Tue, 10 Dec 2002 12:25:49 GMT
my email admin has put some stuff which keeps giving that ACI bit! no idea
why he has done so... sorry for that bit...
----- Original Message -----
From: Guy Davies
To: 'alok' ;
Sent: Tuesday, December 10, 2002 5:29 PM
Subject: RE: iBGP based network. [9:2300]
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
> First of all, what is +ACI- ? A simple quote mark would appear to have
> Anyway, in answer to your question, you have several issues.
> If you plan to use the directly connected interfaces, you have a big
> when any of those interfaces goes down. Destinations attached to each of
> the routers linked by the failed circuit are unreachable to sources
> to the router at the other end. You have no means of doing multihop
> your peerings are setup between the physical interfaces. You can't use
> loopbacks and set next-hop self because the loopbacks are not reachable
> other routers in your network. You could try injecting loopbacks into
> also, but then you rely on recursive lookups using BGP to resolve BGP
> next-hops (IIRC, some vendors don't allow this).
okie no multihop, i agreee........that doesnt matter to me right..no IGP
> In the network you describe, your main problem is that A would be totally
> unknown to B (or x, depending upon which are the BGP routers). Since A is
> directly connected to z and z redistributes directly connected routes into
> BGP, z will announce them with iBGP to y. Unless y is a route reflector,
> will not announce any routes learned from z to x. The only way to even
> start thinking about this is if you have a completely physically fully
> meshed network (yugh!) or a network in which there are route reflectors
> which are physically connected to every client in their cluster and
> physically connected to every other route reflector in the network. Oh,
> they have to set next-hop self towards all their neighbours (clients and
> non-clients) where self is the address of the physical interface with
> they are attached to that specific neighbour!
...hmm so full mesh is a nice problem isnt it :o).......i wont next-hop
self, ill next-hop adjacent neighbour..but i still need a full mesh
> Why would you want to do this, what benefit does it bring?
just.. :o) ....wondering wht "iBGP not bouncing to another iBGP" ever gives
me actually....was it made with the intention of allowing "smaller routing
tables"......or it has a reason...u still have a "source" field in BGP to
recognise the source of the update dont u? so that "RIP equivalent" of
"split horizon" cud be knocked off with that....
> I'm not even going to bother with the question regarding duplicate
> in the network! Suffice to say that you've got a single subnet appearing
> two interfaces on router y. Which one do you choose to send the BGP
> What have you been smoking? ;-)
there was more than that i missed... i had the same IP appearing on one side
of the x router and the same was on its other interface....
that was the bad bit.......sorry abt that.
i realised that immediately after i sent the mail and the "pot" wore
off....that ..its A-x-w-y-z-B ;o)
now can i have the same addresses on the interfaces between x-w and y-z
...yup not with all those issues u mentioned.....
actually here its linux-1 , linux 2, linux 3 , linux 4 linux 5 linux 6
...that i use... and i just relaised i dropped one linux when i sent that
anything else that fails?
> > -----Original Message-----
> > From: alok [mailto:alok.dube@xxxxxxxxx]
> > Sent: Tuesday, December 10, 2002 8:09 AM
> > To: juniper@xxxxxxxxxxxxxx
> > Subject: iBGP based network. [9:2300]
> > Hi,
> > this is not a +ACI-juniper specific+ACI- question, but
> > decided to bounce
> > this off
> > the list:
> > I want to know if this is possible:
> > I want to build a network without OSPF/ISIS as IGP. And I
> > dont want to use
> > static. I plan to use iBGP as my routing protocol and the rest via
> > +ACI-directly
> > connected+ACI-
> > the essential idea being that all +ACI-directly conncted+ACI-
> > next-hops will
> > run
> > iBGP among them ......no routing needed as these are directly
> > conncted+ACI-
> > and
> > all +ACI-host end points+ACI- in my network/their
> > networks/customer networks
> > are
> > distributed into my iBGP (relying on split horizon/ non iBGP to iBGP
> > reflection/as you call it..... to prevent too much route
> > distribution to
> > each node).
> > each iBGP node puts itself as the +ACI-next-hop+ACI- for the
> > routes it is
> > advertising.
> > so now, i know to get to network A from network B, lets say
> > the path is
> > B-x-y-z-A, then B network A is known to x via iBGP with next-hop as y
> > interface, y knows to get to z he has to choose that next-hop
> > interface on z
> > and z has A directly connected.
> > my question is this:
> > do the links between x-y and y-z, or rather the interfaces
> > between x-y and
> > y-z need +ACI-unique+ACI- IP addresses?
> > cant i have interface-x-to-y as 10.1.1.1 and interface-y-to-x
> > as 10.1.1.2
> > and interface-y-to-z as 10.1.1.1 and interface-z-to-y as 101.1.2 too?
> > is there a problem with this? rememeber there is no IGP.
> > -rgds
> > Alok
> -----BEGIN PGP SIGNATURE-----
> Version: PGP 8.0
> -----END PGP SIGNATURE-----
> This e-mail is private and may be confidential and is for the intended
> recipient only. If misdirected, please notify us by telephone and confirm
> that it has been deleted from your system and any copies destroyed. If
> are not the intended recipient you are strictly prohibited from using,
> printing, copying, distributing or disseminating this e-mail or any
> information contained in it. We use reasonable endeavors to virus scan
> e-mails leaving the Company but no warranty is given that this e-mail and
> any attachments are virus free. You should undertake your own virus
> checking. The right to monitor e-mail communications through our network
> reserved by us.
Message Posted at:
FAQ, list archives, and subscription info: http://www.groupstudy.com/list/juniper.html