From: Fred Nielsen (fred_nielsen@xxxxxxxxxxx)
Date: Tue Aug 01 2000 - 18:12:24 GMT-3
Thanks for everyone's input, good stuff all. Phillip mentions a point
below that I am was a little stumped on for a while also, not only
will a router not forward a route but the route won't even make it to
the router's local routing table without "no sync" or an IGP route in
place.
------
Fred Nielsen [fred_nielsen@hotmail.com]
------
----- Original Message -----
From: pkm@calweb.com
To: Geatti
Cc: ccielab@groupstudy.com ; asafayan@msdinc.com
Sent: Sunday, July 30, 2000 2:55 PM
Subject: Re: BGP Synchronization rule
Based on my lab experiences with BGP, I do not think that fully IBGP
speakers is a requirement. Let's say that al your router inside your
transit AS are only BGP speakers and there is no IGP runnning. You
will still have to use the NO SYNCH command on the IBGP speakers,
regardless of having full IBGP mesh topology. If no synch is used, the
network will not appear in the route table even if it in the BGP route
table. There is a good example on the Cisco web site at :
http://www.cisco.com/warp/public/459/14.html (BGP Case study Section
2): http://www.cisco.com/warp/public/459/14.html#A15.0. Also, the book
All in oneCCIE lab study guide has good examples of the no synch rule.
Also, for the CCIE lab, it is most likely that you will have to apply
the next-hop-self statement to the transit AS. The combination of
these two commands (next-hop-self and no synch) allows the BGP route
table to show up in the IP route table of the IBGP speakers correctly.
See all in one CCI Elab study guide. http://www.mentortlabs.com has a
good lab demonstrated the issues with no synch and next-hop-self in a
NBMA network.
Sincerely,
Phillip
Geatti wrote:
Fred,The sync rule basically says in order to advertise a route it
needs to be in the RIB (Routing information base) and the routing
table. Turning sync off via the "no sync" command will allow you to
advertise a given route without it being learned via IGP first. Now
most people would say, why on earth would you want to redistribute
all the routes learned via EBGP into IGP, that would be crazy yeah?
And you would be right. Running no sync with IBGP routers is the
way to go.However to run no sync you must meet one of the following
criteria....a. you must be fully meshed IBGP within your AS.
ORb. you are a stub network - not transit. The reason for the sync
rule is to prevent a packet arriving into your AS destined for
another AS (meaning you are transit) getting to a router within
your AS that does not know what to do with it. If you are running
sync with IGP this would not matter as you would have a IGP route
to the external destination. If you are not in sync as soon as
your packet hits the router it won't know how to handle it, doesn't
have an IGP route to the destination nor is it running IBGP, it
won't have a route to that external AS, the packet is dropped.The
sync rule says that IGP must be synchronized with BGP routes, this
is not practical in most cases. Therefore make sure that when you
use no sync on all routers within you AS that you are fully meshed
via IBGP or using something like route reflectors to make appear
so. If you are fully meshed via IBGP there is no need to be
sync.Sync does require you to have an exact match I believe,
10.0.0.0 /8 does not catch 10.10.10.0 /24 and 10.5.0.0 /16. An
exact match is necessary.Hope this is of some
helpMarco-----Original Message-----
From: nobody@groupstudy.com [mailto:nobody@groupstudy.com]On Behalf
Of Fred Nielsen
Sent: Saturday, July 29, 2000 8:51 PM
To: ccielab@groupstudy.com
Subject: BGP Synchronization rule
I would like to hear the opinions of the group on the
synchronization rule, which states *something* like: a BGP router
will not forward an externally learned route to another external
peer until the route is also present in that router's IGP as
well. The Halabi book touches on this, but didn't spend enough time
for me to really understand the intent behind the rule, other than
to prevent routing loops inside an AS. Because many typical
configurations out there do not redist BGP routes into IGP's, you
see the "no synchronization" command employed fairly often. Why is
sync turned on by default in the IOS? Is it part of the
specification perhaps? Also, referring to "external peer" above,
this really means separate router entities running IBGP within an
AS, right? Not between EBGP peers, where the rule doesn't
apply.. And one more question, does the IGP route have to match
precisely, or can a less specific route do the trick? In other
words, can the presence of 10.0.0.0/8 in the IGP allow BGP to
forward a 10.1.0.0/16 route? Hoping all this makes sense.------
Fred Nielsen [fred_nielsen@hotmail.com]
------
This archive was generated by hypermail 2.1.4 : Thu Jun 13 2002 - 08:24:20 GMT-3