use of ipv4 mdt AF, multicast-intact and static rpf

Hello,

I attached my lab scheme. The questions are related to use of ipv4 mdt AF, multicast intact and static rpf.

 

As you can see all routers are PE, the PIM-SM (not SSM) with RP is multicast protocol for CORE some TE have static route and are configured on core to reach some PE and RP.

 

questions:

 

1) Do we have to configure ipv4-mdt AF between PE and RR if the multicast protocol for CORE is PIM-SM with RP (and not PIM-SSM) ?

2) Do we have to configure multicast-intact if all TE have static route and not autoroute announce ?

3) How many static RPF do we have to configure if we can't use multicast-intact (so one only to reach the RP or one for every TE with static route (two on this case) ?image

 

Isacco

 

Isacco

Comments

  • Hi,

     

    1) Do we have to configure ipv4-mdt AF between PE and RR if the multicast protocol for CORE is PIM-SM with RP (and not PIM-SSM) ?

    IPv4 MDT AF is needed for PEs to discover each other. In case of core PIM-SM rather than PIM-SSM, all PEs will register with the RP, join the RPT and consequently join the SPT to each of the sources. As you can full mesh of SPTs will be build with each PE as a root and the rest of PEs as leafs without involving IPv4 MDT for PE discovery. So the question is NO.

    2) Do we have to configure multicast-intact if all TE have static route and not autoroute announce ?

    Multicast intact only heps to resolve RPF issues with TE tunnels when autoroute announce is used, i.e. when destionations will be installed int the table by means of IGP. In case of static routes you need to define the same set of static mroutes that point to next-hop rather the tunnel interface. 

    3) 3) How many static RPF do we have to configure if we can't use multicast-intact (so one only to reach the RP or one for every TE with static route (two on this case) ?

    You need a static mroute to build RPT and SPT for each of PEs in case RPF was broken by unicast static route. So for each PE that is in the headend of TE tunnel and that has a static route pointing to RP and other PEs via tunnel you need a static mroute with next-hop specified to fix the RPF for them. I hope I expressed myself clearly enough.

  • Hello,

    exactly on last two days I tried many time ad all is correct.

     

    1) ipv4 mdt ins't request if core multicast protocol is PIM-SM

    2) we can use multicast-intact only with autoroute announce, RPF static on all cases (static, FA)

    3) so with static route to TE we need static RPF for PE and RP (if present) routed on TE

     

    Isacco


  • I'll just add...

    When running PIM-SM with no BGP MDT AF the PE routers can discover directly connected Default MDT PE neighbours even when no RP services the Default MDT group address over the mGRE tunnels....

    You do need either a RP servicing the default MDT group address or in case of PIM-SSM then a BGP MDT AF normally a RR...for non direct PE routers to discover each other to build the RPT/ SPT

    Cisco does recommend configuring BGP MDT AF in any case just to get into good practice, however that does not mean it's technically required...

    For you're other question you have to take it step by step, main principle is RPF cannot occur on TE tunnels so you could have autoroute announce, forwarding adjacency or static unicast routes which point to the TE tunnel interface and for any unicast route that is being RPF'd here then you need multicast-intact under the IGP process this then tells the process for any multicast RPF's then use the parallel IGP RIB prior to running MPLS TE for RPF'ing

    Static mroutes are another way around the problem...

    On 1 Oct 2014, at 19:26, danilzhig <[email protected]> wrote:

    Hi,

     

    1) Do we have to configure ipv4-mdt AF between PE and RR if the multicast protocol for CORE is PIM-SM with RP (and not PIM-SSM) ?

    IPv4 MDT AF is needed for PEs to discover each other. In case of core PIM-SM rather than PIM-SSM, all PEs will register with the RP, join the RPT and consequently join the SPT to each of the sources. As you can full mesh of SPTs will be build with each PE as a root and the rest of PEs as leafs without involving IPv4 MDT for PE discovery. So the question is NO.

    2) Do we have to configure multicast-intact if all TE have static route and not autoroute announce ?

    Multicast intact only heps to resolve RPF issues with TE tunnels when autoroute announce is used, i.e. when destionations will be installed int the table by means of IGP. In case of static routes you need to define the same set of static mroutes that point to next-hop rather the tunnel interface. 

    3) 3) How many static RPF do we have to configure if we can't use multicast-intact (so one only to reach the RP or one for every TE with static route (two on this case) ?

    You need a static mroute to build RPT and SPT for each of PEs in case RPF was broken by unicast static route. So for each PE that is in the headend of TE tunnel and that has a static route pointing to RP and other PEs via tunnel you need a static mroute with next-hop specified to fix the RPF for them. I hope I expressed myself clearly enough.




    INE - The Industry Leader in CCIE Preparation

    http://www.INE.com



    Subscription information may be found at:

    http://www.ieoc.com/forums/ForumSubscriptions.aspx
  • Thanks a lot !!!

    Now all issues are clear for me.

Sign In or Register to comment.