No worries John. It was I who mentioned the devious nature of classless and synch as
well :)
Keep in mind that synch was designed for transit networks that have transit providing
routers which do not run BGP. Back when the internet was smaller I expect some
designs had the IGP in an AS carry the full table, or parts of it and hence it was
relevant to make sure your BGP and IGP were synchronized to ensure you didn't
blackhole routes. Today, BGP is run fully meshed with all transit providing routers
in an AS peering with IBGP and hence synch is a complete non issue.
*********** REPLY SEPARATOR ***********
On 4/2/2001 at 8:24 AM John Neiberger wrote:
>Ah, after checking up on this I see where I was confused.
>Synchronization does not specifically refer to the behavior we were
>talking about. I thought that synchronization meant that the next-hop
>had to be in the routing table before a prefix could be moved from the
>BGP table to the routing table. That's not quite correct. I'll quote a
>portion of Halabi:
>
>"The BGP rule states that a BGP router should not advertise to external
>neighbors destinations learned from iBGP neighbors unless those
>destinations are also known via an IGP. This is known as
>synchronization. If a router knows about these destinations via an IGP,
>it assumes that the route has already been propagated inside the AS, and
>internal reachability is ensured."
>
>Thanks for pointing this out, Peter. Someone on the list recently
>pointed out that BGP synchronization and ip classless seem to be in the
>class of misunderstanding. Just when you think you really understand how
>it operates, you realize you have it wrong. <g> I think I have it now!
>Maybe...
>
>John
>
>>>> "Peter Van Oene" <[EMAIL PROTECTED]> 4/2/01 7:03:16 AM >>>
>Synch is an issue that gets way too much attention in my opinion. It's
>not used at all. It's a legacy feature that is meaningless in todays'
>networks.
>
>What John describes below, the fact that IBGP routers will no post
>routes unless they have reachability to the Next_Hop is not a
>synchronization issue, rather it is a fundamental function of BGP. If
>routers started posting routes that they have no hope of delivering
>traffic to, things would get pretty messy pretty fast. Thankfully,
>there are no nobs to turn this _behavior_ off :)
>
>
>
>*********** REPLY SEPARATOR ***********
>
>On 4/1/2001 at 4:58 PM John Neiberger wrote:
>
>>When an eBGP neighbor forwards routing information to another eBGP
>>neighbor,
>>it changes the next hop to itself. When an iBGP neighbors exchange
>>information they do not, by default, change the next hop. This is
>where
>>the
>>synchronization rule comes in.
>>
>>An iBGP neighbor will not be able to use a route if it does not have
>a
>>valid
>>route to the next hop in its IGP. Having synchronization turned on
>is
>>often
>>unnecessary, so most people turn it off. You still have a problem,
>>though:
>>the receiving iBGP neighbor still might not know how to reach the next
>hop
>>for any of the routes in its BGP table. To solve this, on your iBGP
>peers
>>use the next-hop-self command. Since the peers already know how to
>reach
>>each other, this solves your problem.
>>
>>I hope that helps, and I hope I haven't mischaracterized the issue.
>I
>>haven't really thought through all of this in a while so I may have
>some
>>details wrong.
>>
>>If you really want to understand this stuff, pick up a copy of
>Internet
>>Routing Architectures (2nd Ed.) by Sam Halabi.
>>
>>Another book I really liked is short but sweet. It's BGP4:
>Interdomain
>>Routing in the Internet (or something close to that.) It's very short
>but
>>it's an excellent resource. Perhaps you should read that first and
>then
>>read Halabi.
>>
>>Or you could also get a subscription to Certification Zone and read
>>Howard's
>>papers on BGP, they're quite excellent.
>>
>>HTH,
>>John
>>
>>> I'm really confused about the how Next-hop attribute works for IBGP
>and
>>> EBGP. Can somebody please shed some light on this. Any tips or
>help
>>> would be greatly appreciated.
>>>
>>> Regards,
>>> Hunt
>>_______________________________________________________
>>Send a cool gift with your E-Card
>>http://www.bluemountain.com/giftcenter/
_________________________________
FAQ, list archives, and subscription info: http://www.groupstudy.com/list/cisco.html
Report misconduct and Nondisclosure violations to [EMAIL PROTECTED]
