From: Ronnie Royston (RonnieR@xxxxxxxxxxxxxxxxx)
Date: Tue Dec 05 2000 - 14:32:56 GMT-3
My senario looks like,
AS100 <---> AS100
R3--------R4---------R1
| \Frame |ISDN
| \ |
| \__ R7(AS300)
R2(AS200)
R7 is running eBGP w/ R4 and R1. R1 is running the conditional
advertisement feature. There is no dynamic IGP between AS100 and AS300
(all loopbacks are router# twice, e.g. 11, 22, 77, etc. and all links are R#
to R#, e.g., 34 is the link between R3 and R4)
Without the 'non-exist-map' statement, the remote router gets routes from
both peers in his bgp database. See R7's bgp database without the
conditional feautre on R1:
Network Next Hop Metric LocPrf Weight Path
* 11.0.0.0 17.255.255.5 0 0 100 ?
*> 47.255.255.5 0 100 ?
* 14.0.0.0 17.255.255.5 0 0 100 ?
*> 47.255.255.5 0 0 100 ?
*> 17.0.0.0 0.0.0.0 0 32768 ?
***text ommitted***
L7 chooses to send his packets to only one peer in this case. Convergence
is faster, however, after the frame link is brought up, L7 does not return
to L4 as his gateway of choice. See below:
L7#sho ip bgp 11.0.0.0/8
***text ommitted***
47.255.255.5 from 47.255.255.5 (44.255.255.17)
Origin incomplete, localpref 100, valid, external, ref 2
100
17.255.255.5 from 17.255.255.5 (11.255.255.17)
Origin incomplete, metric 0, localpref 100, valid, external, best, ref
2
After a 'clear ip bgp *':
L7#sho ip bgp 11.0.0.0/8
***text ommitted***
17.255.255.5 from 17.255.255.5 (11.255.255.17)
Origin incomplete, metric 0, localpref 100, valid, external, ref 2
100
47.255.255.5 from 47.255.255.5 (44.255.255.17)
Origin incomplete, localpref 100, valid, external, best, ref 2
So, the traffic won't take the frame link after the link is broght back up
until you restart bgp. To fix this, I set the weight of all routes heard
from R4 to 20000 on R7. Now, after a frame link up/down/up L4 preempted
without a 'clear ip bgp'. Basically, I have observed this conditional
advertisement feature to work as the name says, 'conditional advertisement'.
In my case, when link 47.0.0.0 goes down, L1 tells L7 of his routes, when
link 47.0.0.0 is up, L1 stops advertising those routes.
As mentioned above, I also got the backup senario working by adding weight
to routes. Without adding wieght, the backup link continues to be used
until a 'clear ip bgp'.
I like your idea, Simon. Don't run BGP on L1; instead, peer L4 and L7 using
loopbacks w/ 'ebgp multihop 5' on L4 and L7. I hadn't thought of that. You
could run RIP or something over DDR with snapshot routing, all with no
'backup interface' and no floating statics to achieve a backup link.
Thanks.
-----Original Message-----
From: Simon Baxter [mailto:Simon.Baxter@au.logical.com]
Sent: Monday, December 04, 2000 9:11 PM
To: Ronnie Royston
Cc: CCIE Group Study (E-mail)
Subject: RE: BGP - advertise map ... non-exist-map
Can you send me more of the config?
I can't see how this is working. If the 47.255.255.255.4 network goes down,
won't BGP keepalives re-route via the backup? (assuming you're running
multihop and upd lo0 - and running a filtered (so not always connected) IGP
over the link)??
looks good though!!! I'm sure I've got the wrong end of the stick!
Simon
This archive was generated by hypermail 2.1.4 : Thu Jun 13 2002 - 08:25:59 GMT-3