The /32 entry in RIB when running OSPF on point-to-multipoint network

[khoivht]

Dear the Cisco Networking Community,

I have just read the document: https://www.cisco.com/c/en/us/support/docs/ip/open-shortest-path-first-ospf/47862-ospfdb3.html

I noticed that in Router 1.1.1.1, there is an entry 5.0.0.2/32 via 5.0.0.2. I am pretty confused by this. I wonder if the entry has any effect on the routing table lookup process. I also would like to know if that entry is discarded, will it have any impact?

Thank you!

[Richard Burts]

Not sure that I  fully understand your question. But will offer this in hopes that it might address your issue. The connection between routers is a serial link and is defined for ospf as point to multipoint. So conceptually there is the possibility of multiple neighbors on the link. The entry for 5.0.0.2/32 via 5.0.0.2 identifies one of the potentially multiple neighbors. If that entry were to be discarded then the entry for 6.0.0.0 would become ambiguous - which of the neighbors on the serial link would ospf use?

[khoivht]

Thank you for your reply. I would like to know if I use Ethernet cables, and I define as point to multipoint, does the entry matter?

[khoivht]

I am just confused about the necessity of that entry.

[paul driver · Accepted solution]

Hello

I think your referring as to why host routes are created for end hosts when using ospf point-to-multipoint network type.
My understanding this stems from a traditional hub-spoke frame-relay ospf non-broadcast network where each spoke would see the router that originated route as the next hop but it didn’t have reachability (mapping) to that next hop address, So to overcome that you would have reate a additional dlci mapping for each spoke rtr to provide that reachability.

Now with changing the network type to point-multipoint the hub rtr would be seen as having p2p links to each spoke, Each spoke would advertise as having connection to the hub and it stub networks and hosts routes would be advertise automatically so each rtr would be able to lookup the next-hop for that host route and as such obtain reachability.

[Giuseppe Larosa · Accepted solution]

Hello @khoivht ,

in the document that you have mentioned there is a detailed analysis of Router LSA

>>

!--- This router's (1.1.1.1) interface on the !--- point-to-multipoint network. 

     (Link ID) Network/subnet number: 5.0.0.1 
     (Link Data) Network Mask: 255.255.255.255
     
!--- Notice the mask. Only the interface is advertised, !--- not the whole subnet. 

      Number of TOS metrics: 0 
       TOS 0 Metrics: 0

Being the network type point to multipoint an OSPF device connecting to it cannot advertise the whole subnet but only its interface address.

This triggers underlying OSI layer 2 mapping methods for example a specific FR DLCI  or a specific ATM PVC that is used to reach that neighbor.,

This is why you see the /32 entry 5.0.0.2/32 on the neighbor .

Simply OSPF does not suppose that each neighbor is directly reachable even if they are in the same subnet.

The ones that are direct neighbors tell the router sending in their own Router LSA their interface IP address.

This IP address becomes the IP next-hop for prefix 6.0.0.0/8 in your example.

As noted by Rick if the 5.0.0.2/32 would be removed the prefix 6.0.0.0/8 would have an "unknown" next-hop.

Hope to help

Giuseppe

[Richard Burts · Accepted solution]

Following up on the original post the original poster has asked 2 questions. Here are my responses:

- "I would like to know if I use Ethernet cables, and I define as point to multipoint, does the entry matter?" Even if you use Ethernet cables if you define the connection as point to multi point then you still need the entry. Let me say that having an Ethernet connection and defining it as point to multi point is highly unusual. The default behavior of Ethernet is to be multi point broadcast connection. OSPF point to multipoint was implemented for multipoint non-broadcast networks.

- "I am just confused about the necessity of that entry." As I tried to explain ospf point to multipoint was implemented for interfaces/networks where there are potentially multiple neighbors and when there is not a broadcast mechanism to discover the neighbor and so  the layer 2 relationship of the neighbor may not be clear (think of Frame Relay networks or ATM networks - what DLCI should I use to get to 5.0.0.2?). This seems very odd in todays networks because we are so used to operating with networks that do have broadcast capability. If you are studying point to multipoint or doing a lab for point to multipoint you need to think of it in very different terms from what we are used to now.

Perhaps it might help to think of it this way. With a normal Ethernet network you have an arp table that tells us which layer 2 address to use if you want to get to 5.0.0.2. But what if there were no arp table entry for 5.0.0.2? Think of Frame Relay or ATM networks where there is no arp table. Advertising the /32 host address of the neighbor gives you a way to identify the correct layer 2 path to the neighbor.

[khoivht]

Thank you for all of your replies. They are very helpful to me!

[Richard Burts]

You are very welcome. I am glad that our explanations have been helpful. Thank you for marking this question as solved. This will help other participants in the community to identify discussions which have helpful information. This community is an excellent place to ask questions and to learn about networking. I hope to see you continue to be active in the community.