OSPF Process ID need not match. Network Warrior sorta implies otherwise.

CodeBloxCodeBlox Member Posts: 1,363 ■■■■□□□□□□
In this book, Network Warrior, the author says that you can redistribute routes between the multiple instances of spanning tree. I know that it is NOT a requirement for process IDs to match. Regardless of the PID's configured, all routers still learn all OSPF routes. Why does he say this about redistributing between OSPF processes? Was this the way things were with OSPF in the olden days?
Currently reading: Network Warrior, Unix Network Programming by Richard Stevens

Comments

  • fsanyeefsanyee Member Posts: 171
    different instances can have different neighbors.
  • CodeBloxCodeBlox Member Posts: 1,363 ■■■■□□□□□□
    Yeah, but they still learn all the same routes... Two routers that are not adjacent still eventually learn each others routes is what I'm saying. In the book, he's got a diagram with several OSPF running routers with matching PIDs like it's required that they match.
    Currently reading: Network Warrior, Unix Network Programming by Richard Stevens
  • networker050184networker050184 Mod Posts: 11,962 Mod
    Two OSPF processes on a router do not necessarily learn the same routes. I'm not familiar with the example you are referring to, but I'm assuming its a simplified approach to show the basics. The real world isn't always so clean cut.

    As far as do they need to match, no but its a good idea to use the same number on all your routers. It makes mass configuration pushes much easier. You don't want to log in a router go to router ospf 10 and add a network then after 20 minutes of troubleshooting realize this router uses process 20.
    An expert is a man who has made all the mistakes which can be made.
  • CodeBloxCodeBlox Member Posts: 1,363 ■■■■□□□□□□
    Two OSPF processes on a router do not necessarily learn the same routes. I'm not familiar with the example you are referring to, but I'm assuming its a simplified approach to show the basics. The real world isn't always so clean cut.

    As far as do they need to match, no but its a good idea to use the same number on all your routers. It makes mass configuration pushes much easier. You don't want to log in a router go to router ospf 10 and add a network then after 20 minutes of troubleshooting realize this router uses process 20.
    Seems like they would. I can't think of a case where two OSPF processes don't but this could just be the newbie in me. I haven't seen this at work in the real world. His diagram has got PID 100 using routers on the left and PID 200 routers on the right. There is a sorta "Border Router" (no better term for it) that has both PIDs configured for OSPF. I fail to see how this would prevent the routers on the left from learning internal routes to those on the right and how redistributing would be required.
    Currently reading: Network Warrior, Unix Network Programming by Richard Stevens
  • NetworkVeteranNetworkVeteran Member Posts: 2,338 ■■■■■■■■□□
    CodeBlox, the CCNP Routing exam would probably clarify all this. :)

    If you have:

    --A---B---C--

    If:
    + All routers are running OSPF.
    + B is running OSPF process 100 on network {B-left} and OSPF process 200 on network {B-right}.

    Then:
    + B will have network {A-Left} in its RIB via OSPF process 100.
    + B will have network {B-Left} in its RIB because it's directly connected.
    + B's OSPF-200 LSDB will not have any info about networks {A-Left} or {B-Left}.
    + C's OSPF LSDB is identical to B's OSPF-200-LSDB. Thus, neither {A-Left} nor {B-Left} will be in its RIB.

    EDIT - Note, I didn't mention A's or C's process-ids, because they are functionally irrelevant. However, I would strongly encourage using process-id 100 for router A and process-id 200 for router C to simplify troubleshooting.
  • fsanyeefsanyee Member Posts: 171
    Which diagram are you talking about CodeBlox? Give us a page no.
  • CodeBloxCodeBlox Member Posts: 1,363 ■■■■□□□□□□
    Thanks for the explanation. I hadn't considered the fact that the network statements aren't configured in both processes so yeah... I see the need to redistribute between the two in this case as they don't learn each others routes (A and C). Don't see why one would run multiple OSPF processes though. CCNP is where I'm going next actually, looks like it could answer all of my questions.

    As for the page, it's in the beginning of the routing protocol chapter. Let me get the page no.
    Currently reading: Network Warrior, Unix Network Programming by Richard Stevens
  • CodeBloxCodeBlox Member Posts: 1,363 ■■■■□□□□□□
    fsanyee wrote: »
    Which diagram are you talking about CodeBlox? Give us a page no.

    Page 103 and 104
    Currently reading: Network Warrior, Unix Network Programming by Richard Stevens
  • NetworkVeteranNetworkVeteran Member Posts: 2,338 ■■■■■■■■□□
    CodeBlox wrote: »
    Don't see why one would run multiple OSPF processes though
    Suppose you work for company Alpha, who just acquired company Omega, and both companies run OSPF across their internal networks. Your long-term goal may be to integrate the two network infrastructures, but your short-term goal may be to simply get them talking. This would be one of several interim solutions.
Sign In or Register to comment.