The common practice for the SPs is to do the L2->L1 Route-Leaking of the
loopback
address(es) .. so that the remote L1-Routers will have a better picture to
reach the
eBGP next-hops and in the process avoid sub-optimal routing.
See additional comments below -
At 07:46 AM 8/12/2001, Elijah Kagan wrote:
>A question regarding IS-IS and BGP interoperability.
>
>A network runs IS-IS and consists of two areas. Each area has one L1L2
>router and a couple of L1 routers. L1L2 routers from the two areas are
>connected to one another.
>Each L1L2 router has a BGP session with each L1 router in its area and
>acts as a route reflector for them. Both L1L2 talk BGP to each other as
>well.
>
> +--------------------------------+
> | Area1 Area2 |
> | ~~~~~ ~~~~~ |
> | +------+ +------+ |
> | | L1L2 |-----| L1L2 | |
> | +------+ +------+ |
> | ||| ||| |
> | +----+ +----+ |
> | | L1 |+ | L1 |+ |
> | +----+|+ +----+|+ |
> | +----+| +----+| |
> | +----+ +----+ |
> +--------------------------------+
>
>
>Suppose, an L1 router in Area 1 advertises some network via BGP.
>An L1 router in Area 2 receives this advertisement with next-hop set to
>a loopback IP of the Area 1 router, and the only way for it to reach this
>IP is via default gateway (because it comes from another area....). Hence,
>it will not accept the advertisement.
Not sure about the above statements -
You mean "reach this IP is via default gateway" - .... because of the
"ATT" bit set by the L1L2 Router in the L1-Database ... and in the process
L1 Router will inject the default route to the (closest) L1L2 router ?
When L1L2 router is RR to the L1-router (which are in a different area) -
and they have a BGP session - the L1 router does accept the advertisement -
Here R9 is the remote L1-Router in a different area is getting the route
all the way from the *other* remote L1-Router -
R9#sh ip bgp
BGP table version is 2, local router ID is 172.16.9.9
Status codes: s suppressed, d damped, h history, * valid, > best, i - internal
Origin codes: i - IGP, e - EGP, ? - incomplete
Network Next Hop Metric LocPrf Weight Path
* i6.6.6.0/24 172.16.126.6 0 100 0 6 i
*> 172.16.0.0 0.0.0.0 0 32768 i
R9#sh ip ro bgp
6.0.0.0/24 is subnetted, 1 subnets
B 6.6.6.0 [200/0] via 172.16.126.6, 00:01:16
R9#sh ip ro 6.6.6.0
Routing entry for 6.6.6.0/24
Known via "bgp 1", distance 200, metric 0
Tag 6, type internal
Last update from 172.16.126.6 00:01:31 ago
Routing Descriptor Blocks:
* 172.16.126.6, from 172.16.79.7, 00:01:31 ago
Route metric is 0, traffic share count is 1
AS Hops 1
And when changed the Next-hop on the remote L1-Router to itself - who
is forming/learning the eBGP session/routes -
And seeing on the L1-Router (R9) which is on the other side of the world
in a different area...
R9#sh ip ro 6.6.6.6
Routing entry for 6.6.6.0/24
Known via "bgp 1", distance 200, metric 0
Tag 6, type internal
Last update from 172.16.18.1 00:01:43 ago
Routing Descriptor Blocks:
* 172.16.18.1, from 172.16.79.7, 00:01:43 ago
Route metric is 0, traffic share count is 1
AS Hops 1
R9#sh ip bgp 6.6.6.6
BGP routing table entry for 6.6.6.0/24, version 5
Paths: (1 available, best #1, table Default-IP-Routing-Table)
Not advertised to any peer
6
172.16.18.1 (metric 158) from 172.16.79.7 (172.16.1.1)
Origin IGP, metric 0, localpref 100, valid, internal, best
Originator: 172.16.1.1, Cluster list: 172.16.7.7, 172.16.8.8
/Shankar
>Possible solutions I've thought about:
>1. Leak Area 1 aggregate into Area 2 and vice versa.
>2. Manually set the next-hop to L1L2 router's loopback for all prefixes
> learned from another area.
>
>What is the common practice for such configurations, that is, multiple
>IS-IS areas with BGP? How the next-hop issue should be dealt with?
>
>I would also appreciate pointing me to some resources regarding IS-IS and
>BGP interoperability.
>
>Thanks,
>
>Elijah
>
>
>
>
>
>
>
>On Sun, 17 Jun 2001, Elijah Kagan wrote:
>
> >
> > Suppose there is a router (a Cisco router, of course) that learns nothing
> > but a default gateway from its IGP. It also has an iBGP session with its
> > default router and receives the full Internet routing table. The default
> > router acts as a route-reflector.
> >
> > Now here is the problem. The next-hop of every BGP prefix it learns is
> > reachable via IGP's default gateway, but for some reason it is not good
> > enough, BGP marks it as inaccessible and disregards the prefix.
> >
> > This situation could appear in L1 router that learns a prefix originating
> > form a different IS-IS area.
> >
> > Please advise....
> >
> > -- elijah
> >
> >
This archive was generated by hypermail 2b29 : Sun Aug 04 2002 - 04:12:49 EDT