sh ip msdp sa-cache ...

Unanswered Question
Jul 28th, 2008
User Badges:

Hi,


I worked on a network with the following protocols : OSPF, IGMP, PIM sparse dense mode, Anycast RP and MSDP. Topology is made of rings with four RP. On each sites we've got an MRV box to convert FastEthernet to FO. Or it appears that when we disconnected the FO on the MRV, some sites disappears from the RP ( sh ip msdp sa-cache). Have you any ideas ?

Thanks for your help.

  • 1
  • 2
  • 3
  • 4
  • 5
Overall Rating: 0 (0 ratings)
Loading.
Giuseppe Larosa Mon, 07/28/2008 - 09:21
User Badges:
  • Super Silver, 17500 points or more
  • Hall of Fame,

    Founding Member

Hello Frederic,

from your description I would expect that the missing source addresses in the output of sh ip msdp sa-cache are those reachable via the missing fiber link.

May be MSDP takes advantage of PIM capacity of using unicast routing protocols updates to purge its sa-cache.


If so it could be seen as a normal behaviour that you have to take in account when performing activity on the network.


OSPF is fast to detect topology changes so it would update the unicast routing table.


If the missing source addresses are not reachable via the disconnected fiber link that would be strange and would need further investigation.


Hope to help

Giuseppe


fmagen Tue, 07/29/2008 - 00:35
User Badges:

Hello Giuseppe,


Thinks. Indeed OSPF converged, we checked using traceroute before breaking the link and after, so it's ok for unicast. But people on the impacted site didn't receive multicast flows. I join a schema and configs of site site B and one RP. We break the FO on site D and we've got impact at the opposite of the network in site B for example. Thanks for your help.



Giuseppe Larosa Tue, 07/29/2008 - 07:00
User Badges:
  • Super Silver, 17500 points or more
  • Hall of Fame,

    Founding Member

Hello Frederic,

I cannot see your attachment files check if you have uploaded them.


ok people on another side stop to receive multicast frames but where are the sources of these multicast frames? because this is what counts.


Check with your ring topology what happens when you break the ring on site D fiber.


From the point of view of Site B there is any change about the layer3 interface that receives the affected multicast streams ?


If so you could face an RPF check failure that will stop users from receiving multicast flows arriving on the not best interface


Hope to help

Giuseppe



fmagen Tue, 07/29/2008 - 07:55
User Badges:

Hello Giuseppe,

Thanks for your help, i put back the attachments. In fact people on both sites are sources and receivers in the same time (phonie). We've checked RPF on the neighbor router when this happens and effectively we've got RPF check failure :

Group: 224.0.5.233, Source count: 6, Packets forwarded: 3141, Packets received:

3286

RP-tree: Forwarding: 2366/0/45/0, Other: 2511/141/4

What i don't understand is why whereas ospf did converge site B doesn't receive multicast flows. Thanks a lot again for your help.





Giuseppe Larosa Wed, 08/06/2008 - 13:39
User Badges:
  • Super Silver, 17500 points or more
  • Hall of Fame,

    Founding Member

Hello Frederick,

OSPF unicast routing can deal with equal cost parallel links.

The RPF check cannot deal with equal cost parallel links: it will choice one link as best and if the multicast stream arrives on the other one it will be silently discarded for RPF check failure.


Hope to help

Giuseppe

Actions

This Discussion