Here is a discussion of problems in supporting
multiple OSPF interfaces to a single subnet.
I thought that it might be of general interest.
John

-----Original Message-----
From: Moy, John
Sent: Tuesday, March 06, 2001 2:50 PM
To: 'Hari Krishna G'
Subject: RE: A doubt abt Appeindix F RFC2328

Hari Krishna G-

Thanks for the interesting message. First let me say that
Appendix F is very sketchy -- there isn't really enough
detail to get an implementation working.

You have to make sure that:

a) One DR election is performed, taking both interfaces
(c1 and c2) into account. In this respect your
implementation produces the correct result, c2 as DR and
c1 as BDR. If the customer doesn't like this, they should
change DR priorities accordingly.

b) c1 and c2 should *not* accept packets from each other.
I always code my implementations to discard OSPF packets
from their own Router ID. However, C's router-LSA
should *always* be originated as if c1 and c2 were fully adjacent.
There should be two links in the router-LSA to the
network-LSA with Link State ID c2, one link having c1 in
the Link Data field, and the other c2.

John

-----Original Message-----
From: Hari Krishna G [mailto:harikrishna_gpyahoo.com]
Sent: Friday, March 02, 2001 7:45 AM
To: John.MoySYCAMORENET.COM
Subject: A doubt abt Appeindix F RFC2328

Hi John,

   I've a few doubts regarding the proper behaviour
of OSPF for the case described in RFC2328 Appendix F.

The implementation under test took Method 1 suggested
in the spec. I've the following configuration.

               a1 b1
            A-----------B
        a2 | / b2
            | /
            | /
            | /
            | /
            | /
            | /
            | /
            | /

         c2 C c1
S1 = Subnet1
S2 = Subnet2

a1 is in S2
a2 is in S1
c1 is in S1
c2 is in S1
b1 is in S2
b2 is in S1

I configured c1 and c2 to have hieghest priority(c2's
priority is greater than that of c1 )
and C's router id is the hieghest among all the
routers

1. After DR election C2 has become DR
                     C1 has become BDR
         Effectively i've the same router as DR and
BDR of the N/W

2. C1 accepted c2 as a neighbor and vice versa
   Adjacencies started forming.
             Since c2 is DR it has become fully adjacent to
a2,b2 and similar thing happened with c1

3. Now I've a case where DR and BDR of the N/W are not
adjacent. The reason is there is no tie braker for the
event negotiation done to fire. (Router id of two
neighbors is the same if c1 and c2 are to be fully
adjacent)

  I assume its a bug in the implementation. But where
should we have to introduce a break so that we never
run into situation like this

I've the following doubts

First is the DR election correct? If it is correct
then we have a scenario where DR and BDR of the
N/W is same. Will it not create problems if C (DR &
BDR) goes down?

Secondly should c1 accept c2 as neighbor?
  If it should not then where should that get stopped?
Is it at hello processing?
  If it should then what is the tie braker for
Master/Slave decision?
   What is the point in establishing ajacency bet'n
tow interfaces of the same router.

Could you please help me in solving out this. Please
let me know if I'm not clear in my description.

Thanks alot for any help
Hari Krishna G

             Actual addresses of the configuration for
more clarity
        Intf Address Mask Priority
         a1 192.168.104.1 255.255.255.0 1
         a2 192.168.103.1 255.255.255.0 1
         b1 192.168.104.2 255.255.255.0 1
         b2 192.168.103.2 255.255.255.0 1
         c1 192.168.103.3 255.255.255.0 5
         c2 192.168.103.4 255.255.255.0 6

        Router Router ID
        A 10
        B 11
        C 12

__________________________________________________
Do You Yahoo!?
Get email at your own domain with Yahoo! Mail.
http://personal.mail.yahoo.com/

• Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]