From: Peter van Oene (pvo@xxxxxxxxxxxx)
Date: Mon Jun 10 2002 - 09:06:41 GMT-3
inline
At 08:49 AM 6/10/2002 -0300, Carlos G Mendioroz wrote:
>Hmm,
>synch does not address non full-mesh, it just prevents
>full-mesh with internal non BGP routers from advertising
>a route too soon. iBGP has to be full mesh, always.
synch does address non full mesh transit networks. it was designed to
support networks where igp only routers might exist in the transit path
between bgp speakers. it adds no value to full mesh ibgp networks
>RRs just "collapse" the full mesh to some routers (the reflectors)
>but this should not change anything of the standard working of
>iBGP (clients should not notice, almost :-)
this is correct
>I mantain that the problem lies in using the same IGP across
>AS boundaries, and if you fix that, everything is ok.
i maintain that you'd never run an IGP between AS's. they weren't designed
for this and I've never seen anyone argue that you should.
>I don't think that breaking an OSPF backbone that is inter-AS
>is that bad. I feel ok with having each AS run its own IGP.
there shouldn't be any inter-as ospf domains. ospf is an IGP, not an EGP
and really quite dangerous to run between networks that are under different
control.
>Peter van Oene wrote:
> >
> > Why is it that synch and route reflection aren't intended to work
> > together? Possibly because the former solves some potential routing issues
> > in non full ibgp mesh transit networks while the latter helps scaling
> > properties of full mesh ibgp network. ie, they are designed to be applying
> > to different networks (full mesh vs non full mesh)
> >
> > When the answer to the question of dealing with route reflection in ospf
> > based networked is to partition the ospf backbone or fully reconfigure the
> > bgp topology I tend to doubt the validity of the question as these seem to
> > be pretty stretching solutions.
> >
> > I'll look up your 6/6 answer to see if i'm missing anything.
> >
> > At 08:30 AM 6/10/2002 -0300, Carlos G Mendioroz wrote:
> > >Why is that ???
> > >
> > >I've just (thursday 6/6) posted a way of "fixing" this using OSPF &
> > >BGP,
> > >namelly making the OSPF routing domain coincide with the BGP AS, by
> > >breaking it at the BGP borders (running 2 OSPF processes at the borders,
> > >one for intra-AS and the other for inter-AS).
> > >
> > >Other ways that have been talked about are turning into BGP confeds
> > >(thus
> > >getting rid of the iBGP sync issue) and tweeking the router IDs...
> > >
> > >Peter van Oene wrote:
> > > >
> > > > Synch and route reflection are not intended to work together.
> > > >
> > > > At 08:19 PM 6/9/2002 -0400, Yagnesh Patel wrote:
> > > > >Hi,
> > > > >I know the following problem is frequently discussed in the group
> study
> > > > >but couldn't find any definite answers. Can someone please direct
> me to
> > > > >the possible solutions to this problem
> > > > >
> > > > >R2 ------IBGP-----R3 (RR)--------IBGP-------- R1
> -----<EBGP>---------R4
> > > > >
> > > > >
> > > > >Sync is enabled in all router. R1's reflected routes are not sync
> in R2
> > > > >router.
> > > > >
> > > > >Thanks
This archive was generated by hypermail 2.1.4 : Tue Jul 02 2002 - 08:12:30 GMT-3