Re: EIGRP backdoor link on MPLS VPN and multiple SoO values

From: Pierre-Alex Guanel <paguanel_at_gmail.com>
Date: Thu, 16 Feb 2012 06:48:39 +0000

Sorry, did not hit REPLY ALL so resending - with the typo correction.
_________

Hi Yuri,
>
> Thank you for the hints. I spent quite a bit of time labing based on your
> input.
>
> Indeed, the SoO attribute is propagated in the EIGRP packets as you
> mentioned. Also, as indicated, a CE can filter EIGRP updates based on the
> SoO information.
>
> A question remains however as to the usefulness of SoO filtering. Here is
> why:
>
>
> - 1. I added a switch to CE1 and advertised a loopback in EIGRP.
>
> - 2. When SoO is configured on CE1 on the backdoor interface, the update
> sent by CE2 is NOT recorded in the the EIGRP topology ( as expected)
>
> - 3. When SoO is NOT configured on CE1 on the backdoor interface, the
> update sent by CE2 is recorded in the EIGRP topology database ( as
> expected)
>
> - 4. In either case, CE1 never advertises any update back to PE1 because
> the CE2 is not a feasible successor (basic EIGRP rules)
>
> - 5. I tried many things over the past two days (playing with offsets,
> delays and variance throughout the topology) to make CE2 a feasible
> successor but I never was able to make this happen.
>
> Finally I reasoned myself that if it was mathematically possible to adjust
> the metrics so that CE2 was indeed a feasible successor, then this could
> lead to a massive routing loop.
>
> Thus my doubt as to the usefulness of this feature:
>
> If there can't never be a routing loop in a back door topology (because
> of the feasibility condition), what really is the point of the EIGRP SoO
> filtering mechanism?
>
> Is it just some kind of paranoid filtering ..or am I mistaken in
> thinking that CE2 cannot be made a feasible successor ????
>
> Thank you
>

On Tue, Feb 14, 2012 at 4:47 AM, Yuri Bank <yuribank_at_gmail.com> wrote:

> The CE can inspect the EIGRP updates coming from the backdoor link. You
> configure a sitemap to do this. The CE does not have to be running BGP!
>
> Check your packet capture again. Wireshark will not recognize the SoO
> attribute included in the EIGRP packet. Look for "Type Unknown 0x0104" The
> last 6 bytes of this field is the SoO! So just convert the HEX to Decimal
> :-)
>
> Cheers
>
> -Yuri
>
> On Mon, Feb 13, 2012 at 5:10 AM, Pierre-Alex Guanel <paguanel_at_gmail.com>wrote:
>
>> According to paragraph 2) on page 2545 of INE CCIE R&S Lab Workbook Volume
>> 1 (CCIEv4.0), when configuring PE routers of a multi-homed sites with
>> different SoO values, an additional SoO interface should be configured on
>> the CE routers with the backdoor links " to prevent routing loops.
>>
>> Diagram:
>>
>>
>> PE1 -------------P--------P------------------PE2
>> | |
>> | |
>> SoO=100:1 SoO=100:2
>> | |
>> | |
>> CE1--SoO=100:1----------SoO=100:2-CE2
>>
>>
>>
>> Questions:
>>
>> How can a CE (acting as a backdoor) filter updates based on the site of
>> origin since the CE is not running
>> MBGP ?
>>
>> I have taken a network capture of traffic on the backdoor link, I do not
>> see any extension in EIGRP carrying the SoO attribute ...
>>
>> Any input welcome.
>>
>> Thanks
>>
>> Pierre-Alex
>>
>>
>> Blogs and organic groups at http://www.ccie.net
>>
>> _______________________________________________________________________
>> Subscription information may be found at:
>> http://www.groupstudy.com/list/CCIELab.html

Blogs and organic groups at http://www.ccie.net
Received on Thu Feb 16 2012 - 06:48:39 ART

This archive was generated by hypermail 2.2.0 : Thu Mar 01 2012 - 11:46:56 ART