OSPF initialisation and updates

I've looked at the OSPF article on this site, but it doesn't explain something that I'd like to have clarified.

Let's say that I have 5 routers, each attached to a LAN segment. The Fa0/0 addresses are Router A = 10.0.0.1/24 to Router E = 10.0.0.5/24 and the priority of each remains at the default = 1. When the routers all come up, each router establishes an adjacency with the other four on the LAN and this is via the Hello protocol. As far as I can understand, the purpose of these initial adjacencies (2-way communications) is to establish DR and BDR via an election process. I figure that Router E will become the DR and Router D will become the BDR.

When everything is stable, there are no routine updates but, if Router A identifies that a network which is attached to it goes down, it sends a multicast update to 224.0.0.6 which only goes to the DR and BDR. The DR sends a unicast acknowledgement back to Router A then it sends a multicast update to 224.0.0.5 which goes to all routers on the LAN. Each sends a unicast acknowledgement back to the DR.

Questions:

1. Is this simplistic view correct so far?

2. When the DR sends out to 224.0.0.5, that doesn't include the BDR for two reasons, first it's already learned of the problem via the multicast from Router A to 224.0.0.6 and secondly, it's multicast address is 224.0.0.6 rather than 224.0.0.5?

3. I assume that the BDR doesn't do anything actively (other than update it's own records) when it receives the information from Router A because the DR will update the other routers on the LAN?

4. When the DR sends the multicast update out to all routers (224.0.0.5), does that include Router A, which informed the DR of the problem in the first place?

I have a whole load more things going around in my head about how OSPF works and they're starting to crystallise slowly. I'd like to get things absolutely straight so far, before I ask further questions.

Thanks for your time.

When everything is stable, there are no routine updates

Yes, except for hello messages.

but, if Router A identifies that a network which is attached to it goes down, it sends a multicast update to 224.0.0.6 which only goes to the DR and BDR. The DR sends a unicast acknowledgement back to Router A then it sends a multicast update to 224.0.0.5 which goes to all routers on the LAN. Each sends a unicast acknowledgement back to the DR.

Absolutely correct. Moreover, the updating process is called LSA "Link State Advertisement" and the update packet itself is referred to as LSU "Link State Update".

1. Is this simplistic view correct so far?

Yes

2. When the DR sends out to 224.0.0.5, that doesn't include the BDR for two reasons, first it's already learned of the problem via the multicast from Router A to 224.0.0.6 and secondly, it's multicast address is 224.0.0.6 rather than 224.0.0.5?

Correct. However, I'm not sure that the BDR is not also a member of the 224.0.0.5 multicast group as well as the 224.0.0.6. But your probably correct.

3. I assume that the BDR doesn't do anything actively (other than update it's own records) when it receives the information from Router A because the DR will update the other routers on the LAN?

Correct, again except hello packets.

4. When the DR sends the multicast update out to all routers (224.0.0.5), does that include Router A, which informed the DR of the problem in the first place?

Yes, According to the Cisco academy material. The reason why (I think) is because Router A is already registered as a member of the multicast 224.0.0.5 address group.

By the way, for multicast to reach some routers and not others, the LAN switch will have to support muticast groups. Otherwise multicast packets will act just like broadcast. But the whole OSPF process will still work.

Thanks for the general "thumbs up" so far. Now a couple of questions about the initial period leading to convergence:

1. I think that the DR and BDR routers are elected before any exchange of link states occurs. Is this correct?

2. Do all routers form an adjacency with each other in order to exchange their Link States (i.e. Router A <---> Router B, Router B <---> Router C, Router A <---> Router C etc.) or are the adjacencies ONLY formed between the non-DR/BDR routers and the DR/BDR and then ALL the Link State information is exchanged *via* the DR (or BDR if the DR is down)?

I've found conflicting information about this stage. I think that I have the post-convergence situation fairly clear in my head.

Thanks again for your time.

No prob. SteveP.

1. It seams that you are right. When I "debug ip ospf adj" and "debug ip ospf events" in packet tracer, I see the elections happening first. However, it seams that at the start of the network (when some routers are just being started) some elections may be replaced by other newer elections, while some LSA were sent just after the first election and before the newer election.

2. Cisco defines the "adjacency database" as: "A list of all the neighbor routers to which a router has established bidirectional communication. This is unique to each router". I understand from this that a router forms adjacencies with all neighboring routers. Not just the DR and BDR. But this is just my own understanding. Still, anyways, There is no doubt that Link State information is exchanged ONLY via the DR (or BDR if the DR is down) as you said.

Thanks again S0lo

Since posting my follow-up question, I've also played around with Packet Tracer and was confused to see that DR & BDR elections were made early in the process but then it happened again and the subsequent elections were as I'd expected on the basis of Priority & RID. I just wondered if it was a flaw in the Packet Tracer software.

I've spoken with my tutor and she's given me the go-ahead to install Wireshark on a lab PC. I'll have my 3 routers and the PC attached to a hub so I can watch the packets flying everywhere when the routers are turned on. I know it'll generate whole load of information (arp, cdp, stp as well as ospf data) but I'm determined to get to the bottom of this fascinating process.