07/01/99 6Bone Routing Report

[Masaki Hirabaru]

>> I think It was occured by 'NLRI-Length attribute' probles with
>> CISCO(AS 109). CISCO have announced and received routes without
>> NLRI_length suddenly, so CISCO mis-recognizes our routes and announce
>> them to other pTLA. And frequently termination the BGP connection
>> cause route flapping.

I've thought that the version difference just causes immediate
BGP session closing since the code should check the packet
format.  If this happens, I think that we'd better stop using the
old version as soon as possible. When I proposed this a long time
ago, it was too early. But, now should we make a schedule to
forget the obsolete BGP4+ version?

Masaki

>> From: SUMIKAWA Munechika ($B3Q@n=!6a(B) <sumikawa@ebina.hitachi.co.jp>
>> Subject: Re: 07/01/99 6Bone Routing Report
>> Date: Sat, 24 Jul 1999 16:28:46 +0900 (JST)
>> Message-ID: <19990724162846Y.sumikawa@ebina.hitachi.co.jp>
>>
>> Sorry for delay resoponse. I needed a week for investigation the
>> problem.
>> 
>> masaki> BTW, there are also similar flapping routes that look from WIDE
>> masaki> Project and contribute great increase of IPv6 traffic on 6bone.
>> 
>> masaki>   0000::/0 path 109 2500 2500 2500 2500 2500 (WIDE)
>> masaki>   1800::/4 path 109 2500 2500 2500 2500 2500 (WIDE)
>> masaki> 1.  (0000::/0) had 93371 BGP+ updates (18 unique aspaths)
>> masaki> 2.  (1800::/4) had 87912 BGP+ updates (16 unique aspaths)
>> 
>> I think It was occured by 'NLRI-Length attribute' probles with
>> CISCO(AS 109). CISCO have announced and received routes without
>> NLRI_length suddenly, so CISCO mis-recognizes our routes and announce
>> them to other pTLA. And frequently termination the BGP connection
>> cause route flapping.
>> 
>> Now I changed my router configuration, it seems that no curious routes
>> annouce to 6bone.
>> 
>> Thanks,
>> 
>> ---
>> Munechika SUMIKAWA @ WIDE Project