Multicast and Rapid-pvst

[CHARLES HEUPEL]

I have a question regarding MCAST and how it works with spanning-tree.  We currently have two 7609 routers that are running multicast for IPTV to our subscribers.  The access equipment we use has redundant 802.1q links to each 7609.  We also have a 802.1q link between the 7609s.  When I look at spanning tree for one of my MCAST vlans, it's forwarding on one 7609 and blocking on the other as I would expect it to be.  However, when I look at the mroute table on my 7609s, it appears the router which is blocking the MCAST vlan is actually forwarding the MCAST stream on that VLAN.  So my question is how does MCAST operate in a spanning-tree environment.  Does it not follow the spanning-tree topology?

[rsimoni]

Hi Charles,

the part you are missing is that SPT works at layer 2 while the mcast traffic you are referring to is layer 3.

So a multilayer switch (as the 7600) can easily block some ports at L2, also for the mcast vlans, while at L3 mcast traffic is still forwarded on other L2 ports not being blocked. Nothing abnormal about that.

Riccardo

[CHARLES HEUPEL]

Riccardo,

Thanks for the reply but I think you misunderstood my post.

I have a VLAN (643), that exist on both 7600s ( lets say A and B).  A and B have a trunk link between them on interface 1/0/1 respectively.  I have an access device (C) that has a trunk port going to 7600A and B on interface 1/0/2 respectively.  VLAN 643 exists on all trunk ports.  When I do a show spanning-tree vlan 643, 7600B is blocking on port 1/0/2 and forwarding on port 1/0/1.  However, when I do a show ip mroute for my MCAST group on 7600B, there is an mroute built and it shows that it's forwarding out vlan 643.  7600A shows no mroute for the MCAST group. I thought since 7600B was blocking the VLAN on port 1/0/2, any IGMP membership reports would come in on 7600A and it would build the mroute table for that group.  From what I can tell, it looks like this is all happening on 7600B.  I will do a packet capture to confirm but I was a little uncertain how MCAST responded to STP topology changes.  Thought there might be something I'm not aware of......

Shane

[rsimoni]

Hi Shane,

I think I did not misunderstand what you meant ... the answer is still the same, SPT refers at L2 while IP multicast refers to L3, you cannot mix the  2 things.

In a nuthshell:

both 7600A and 7600B receives IGMP joins from vlan643 at L2 > 7600A receives them from 1/0/2 (access device) while 7600B receives them from 1/0/1 (7600A) - anyway the physical port does not count as what counts is that both router receives it a L3 on vlan643 interface, as again we care of the L3 topology for mcast not the L2.

So both 7600A and 7600B receive the IGMP join from receiver in vlan643 (or from downstream router) but only one is the DR (router registering to RP if PIM-SP is used). Depending on your mcast tree the DR might also be the router which forwards mcast traffic down, that is the one with the mroute for the mcast group fully populated (OIL being vlan643). The other router will have the mroute with the OIL null. Anyway is not the DR election that determines who is forwarding the traffic in the vlan but the Assert Winner mechanism. There are cases where both routers in a subnet receive the mcast traffic, but only the assert winner forwards it with the tie breaker being the highest IP address.

For some topology the DR is the one which forwards as it is the only router that receive the stream after registering to the RP. In other scenarios both routers receive it and hence the assert mechanism is needed to avoid duplication.

So if 7600B is the Assert winner for vlan 643 (or simply the DR in a topology where 7600A does not receive the traffic from upstream router)  it is expected that it forwards mcast traffic even though the join, FROM A L2 POINT OF VIEW, came in via 7600A.

Riccardo