From: Nick Griffin (ngriffin@sio.midco.net)
Date: Thu May 25 2006 - 21:29:51 ART
It sounds to me like you were running into next hop recursion
oscillation. Check out the practical bgp book:
"The BGP process will then revert the route back to the bestpath through
router C, and the next time the BGP scanner process runs, the route will
switch the router to the path through router B. This will continue,
every 60 seconds.
The problem here is not that the route to the next hop is learned via
iBGP, but rather the route to the next hop recurses through the route
itself, which is impossible. There are several possible solutions to
this problem:
*
Include the next hop, in this case 10.1.1.1, in the interior
gateway protocol so it is not learned through iBGP. This breaks
the recursion of the route onto itself.
*
Create a static route to the next hop from router A. This also
breaks the recursion by providing another path to the route's next
hop than through the route itself.
*
Configure router C to set the next hop for all routes it is
advertising to itself using the Cisco IOS software command
neighbor <x.x.x.x> next-hop-self. This breaks the recursion of the
route by changing the next hop so it's not part of the route being
learned."
Vishal Patel wrote:
> Hi ,
>
> Got it fixed.
>
> The problem was , I was accepting only default route in IGP.
>
> The BGP's neigbour was also not accepted in the filter.
>
> I added the IBGP neigbour in the filter and the things are looking fine
> now.
>
> But I don't know the reason , why the routes where flip flopping.
>
> If the router was not having the route for the BGP neighbour , then the BGP
> session shouldn't have established at all.
>
> Any ideas ?
>
> Cheers
>
> Vishal
>
> -----Original Message-----
> From: nobody@groupstudy.com [mailto:nobody@groupstudy.com] On Behalf Of Nick
> Griffin
> Sent: Thursday, 25 May 2006 11:15 PM
> To: Vishal Patel
> Cc: ccielab@groupstudy.com
> Subject: Re: Routing issue
>
> I would venture to say that if you did a "show ip bgp x.x.x.x" before the
> routes flip to IGP, that the next hop would be marked as inaccessible. Have
> or can you verify this? It's possible you have a next hop recursion problem
> causing the routes to be installed and removed by the scanner process when
> the next hop is marked inaccessible.
>
> Vishal Patel wrote:
>
>> Hi all,
>>
>> Iam facing a weird problem.
>>
>> I have a router running ospf and ibgp.
>>
>> And i have a distribute-list which allows only a default route from
>> the neigbour in ospf domain.
>>
>> And Iam accepting all the internal routes via IBGP.
>>
>> My routing table should show the internal routes via IBGP with the
>> metric of 200.
>>
>> But the problem happening is the routing updates are continously
>> changing between ospf and bgp.
>>
>> for a movement I get only default route via ospf and next movement I
>> get all the internal routes via IBGP.
>>
>> As soon as remove distribute-list from ospf and accept the whole
>> network via ospf , the things are stable.
>>
>> Well i also have more ospf neigbour , to a different router but it
>> has higher cost and it never appears in the routing table.
>>
>> Any ideas ?
>>
>> Cheers
>>
>> Vishal
>>
>> ______________________________________________________________________
>> _ Subscription information may be found at:
>> http://www.groupstudy.com/list/CCIELab.html
>>
>
> _______________________________________________________________________
> Subscription information may be found at:
> http://www.groupstudy.com/list/CCIELab.html
>
> _______________________________________________________________________
> Subscription information may be found at:
> http://www.groupstudy.com/list/CCIELab.html
This archive was generated by hypermail 2.1.4 : Thu Jun 01 2006 - 06:33:22 ART