[j-nsp] ng-mvpn problem

Antonio Sanchez-Monge amonge at juniper.net
Wed Oct 23 03:43:44 EDT 2013


Brilliant! I think the reason why it worked with rpt-spt, even for non-DR, is that the upstream PE was receiving type 6 route, so that triggered the signaling without the need of a pre-existing type 5 route.

From: Mihai Gabriel <mihaigabriel at gmail.com<mailto:mihaigabriel at gmail.com>>
Date: Wednesday, October 23, 2013 8:49 AM
To: Antonio Sanchez Monge <amonge at juniper.net<mailto:amonge at juniper.net>>
Cc: "Stacy W. Smith" <stacy at acm.org<mailto:stacy at acm.org>>, "juniper-nsp at puck.nether.net<mailto:juniper-nsp at puck.nether.net>" <juniper-nsp at puck.nether.net<mailto:juniper-nsp at puck.nether.net>>
Subject: Re: [j-nsp] ng-mvpn problem

Hello,

 You are always right :) Everything works as expected now.

Thanks,
Mihai


On Wed, Oct 23, 2013 at 2:04 AM, Antonio Sanchez-Monge <amonge at juniper.net<mailto:amonge at juniper.net>> wrote:
Agreed, that should do the trick too :)


On 10/23/13 1:01 AM, "Stacy W. Smith" <stacy at acm.org<mailto:stacy at acm.org>> wrote:

>Agreed.
>
>The lt-1/1/10.770 interface which is in VRF mvpn on logical-system x must
>have PIM enabled (or multicast forwarding enabled). If running PIM, it
>must be the DR.
>
>I wasn't suggesting disabling PIM on the lt-1/1/10.770 interface which is
>in VRF mvpn on logical-system x, just disabling PIM on the remote end of
>the a-x link (in logical-system a).
>
>--Stacy
>
>On Oct 22, 2013, at 4:49 PM, Antonio Sanchez-Monge <amonge at juniper.net<mailto:amonge at juniper.net>>
>wrote:
>> You need PIM in the interface towards the source IMHO
>>
>>
>>
>> On 10/23/13 12:47 AM, "Stacy W. Smith" <stacy at acm.org<mailto:stacy at acm.org>> wrote:
>>
>>> Yes, that would also work, but since logical-system a is really just
>>> emulating a multicast source, there's really no need for it to run
>>>PIM. A
>>> typical multicast source would not be running PIM.
>>>
>>> --Stacy
>>>
>>> On Oct 22, 2013, at 4:44 PM, Antonio Sanchez-Monge <amonge at juniper.net<mailto:amonge at juniper.net>>
>>> wrote:
>>>> Solution would be setting a higher PIM priority in lt-1/1/10.770, so
>>>> that
>>>> it becomes the DR
>>>>
>>>>
>>>>
>>>>
>>>> On 10/23/13 12:40 AM, "Antonio Sanchez-Monge" <amonge at juniper.net<mailto:amonge at juniper.net>>
>>>> wrote:
>>>>
>>>>> That's a brilliant analysis Stacy, I think you nailed it (awaiting
>>>>> Mihai's
>>>>> confirmation).
>>>>>
>>>>>
>>>>> On 10/22/13 11:59 PM, "Stacy W. Smith" <stacy at acm.org<mailto:stacy at acm.org>> wrote:
>>>>>
>>>>>> On Oct 22, 2013, at 2:44 PM, Mihai <mihaigabriel at gmail.com<mailto:mihaigabriel at gmail.com>> wrote:
>>>>>>> Removing PIM fromlt-1/1/10.770 is not a solution because the PE
>>>>>>>will
>>>>>>> not learn about the source and the multicast group.
>>>>>>
>>>>>> Actually, removing lt-1/1/10.770 from PIM would allow the source and
>>>>>> multicast group to be learned, and fix the problem (as long as
>>>>>> multicast
>>>>>> routing was still enabled on the lt-1/1/10.770 interface).
>>>>>>
>>>>>> The problem is that there's a PIM neighbor relationship between a
>>>>>>and
>>>>>> x.
>>>>>> Because of your IP addressing, a is the DR for the a-x LAN.
>>>>>>
>>>>>> Because you are injecting traffic with ping and "bypass-routing
>>>>>> interface
>>>>>> lt-1/1/10.771" logical-system a is NOT the first-hop router. It's
>>>>>> simply
>>>>>> acting as a multicast source that's pumping traffic with destination
>>>>>> IP
>>>>>> 225.10.10.10 out the lt-1/1/10.771 interface.
>>>>>>
>>>>>> Logical-system x instance mvpn receives this traffic on
>>>>>>lt-1/1/10.770
>>>>>> and
>>>>>> does not forward it because it is not the DR.
>>>>>>
>>>>>> Therefore, the logical-system x instance mvpn doesn't learn about
>>>>>>the
>>>>>> active (S,G).
>>>>>>
>>>>>> Another way to solve this problem is disabling PIM on logical-system
>>>>>> a.
>>>>>> This will make lt-1/1/10.770 on logical-system x instance mvpn the
>>>>>>DR,
>>>>>> and cause it to learn about the active S,G (and therefore generate
>>>>>>the
>>>>>> NG-MVPN Type 5 route).
>>>>>>
>>>>>> I have mocked up your configuration in the lab and confirmed that
>>>>>> removing PIM from logical-system a fixes the issue.
>>>>>>
>>>>>> --Stacy
>>>>>>
>>>>>>
>>>>>>
>>>>>
>>>>
>>>>
>>>
>>>
>>>
>>
>>
>
>
>





More information about the juniper-nsp mailing list