From: Jongsoo.Kim@Intelsat.com
Date: Mon Aug 09 2004 - 19:21:56 GMT-3
Below your email said, "If one of the interfaces is being used as the BGP
source and it goes down, say it is the other provider link and not the
directly connected link to you, your BGP session goes down. You still have
connectivity, but no BGP." I can't understand how this is possible. If you
didn't use update source, then the only way to establish BGP is to use the
directly connected interface, which shouldn't be affected by the other
provider's link. What am I missing here? ^^
I just don't agree of using eBGP update source always in CCIE LAB exam.
I believe it should depend upon the requirement.
If it says something like " eBGP connection should be up as long as routers
are reachable", eBGP update source using Loopback interface should be used
plus the loopback ip address should be announced via IGP.
But if it said " eBGP connection should be down when directly connected link
is down", most likely eBGP update source shouldn't be used.
Also per my small experience( 1 st attempt failed and 2nd one in three weeks
if re grade request failed), the eBGP question did very clearly say if eBGP
update source is required or not .
Jongsoo
-----Original Message-----
From: Tony Schaffran [mailto:groupstudy@cconlinelabs.com]
Sent: Sunday, 08 August, 2004 1:53 PM
To: Kim, Jongsoo; ccielab@groupstudy.com
Subject: RE: RE: BGP Update Source
I work for an ISP and the preferred method for multi-homed BGP is
update-source loopback with the multi-hop set to the minimal hop count to
reach the peer. You should not be using multi-hop=255, I agree. In our
experience, it is actually the reverse, we have had bad situations where
update-source loopback is not used. If one of the interfaces is being used
as the BGP source and it goes down, say it is the other provider link and
not the directly connected link to you, your BGP session goes down. You
still have connectivity, but no BGP. Yes, there are ways to ensure this
does not happen, but the most simple straight forward way is to use
update-source loopback.
Anyway, this string has gone way off the point of the LAB. For lab
purposes, it is best to use update-source loopback. Do not try to
rationalize the lab with what is done in real life. If you do that, you
will be scheduling you re-test.
Just my two cents.
Cheers,
Tony Schaffran
Network Analyst
CCIE #11071
CCNP, CCNA, CCDA,
NNCDS, NNCSS, CNE, MCSE
www.cconlinelabs.com
Your #1 choice for online Cisco rack rentals.
-----Original Message-----
From: nobody@groupstudy.com [mailto:nobody@groupstudy.com] On Behalf Of
jongsoo.kim@intelsat.com
Sent: Sunday, August 08, 2004 9:26 AM
To: ccielab@groupstudy.com
Subject: Re: RE: BGP Update Source
Just my two cents on why EBGP peering w/ multihop, update source is not
recommended.
In real ISP life, EBGP session with multihop = 255 and update source
loopback can be really "dangerous" config as this can create all kind of
problems( black hole, BGP session never being down when it supposed to be)
specially if the customer's( the other bgp peer) loopback is reachable
without bringing up the physical connection ( so customer has multi-homed
and loopback address is reachable via the other ISP link as well).
It is only recommended for load-balance with ttl= 2 or whatever necessary
minimum number. However, in a simplex satellite VSAT network commonly use
eBGP w/ multihop=255 with a great caution.
Jongsoo
This archive was generated by hypermail 2.1.4 : Fri Sep 03 2004 - 07:02:36 GMT-3