Re: Reestablish adjacencies after Graceful-Restart

[Debopam Bandyopadhyay <debopam at ALUMNUX.COM>]

Thanks a lot for the help, Acee. It has cleared a major chunk of my design
requirements.
Finally, I have one question related to a hypothetical case (which may be
required for interoperability).

I will go back to the previous example:

            X ----- lan ----- Y

In the topology, we have two routers, X and Y; and Y happens to be the DR in
the network segment.

Now Router X restarts. After 'T' seconds, Router X becomes fully adjacent
with Router Y.  'T is very much less than X's Grace Period.
Let's say, X is unable to trace its pre-restart Router-LSA in its database
(because Y had not sent it during the DD-exchange). So my question is:
1> Would X now exit Graceful-restart at this point, or
2> Would X wait for its pre-restart Router-LSA to be received from Y (in the
form of LS-Updates), until grace period expires; or
3> Would it be implementation dependent?

Thanks,
Debopam

----- Original Message ----- 
From: "Acee Lindem" <acee at CISCO.COM>
To: <OSPF at PEACH.EASE.LSOFT.COM>
Sent: Monday, May 09, 2005 5:50 PM
Subject: Re: Reestablish adjacencies after Graceful-Restart


Debopam Bandyopadhyay wrote:

>This question is a bit long, backdated and maybe, queer. But it is a bit
critical for me.
>It is regarding RFC 3623 (Graceful OSPF Restart), Section 2.2. "When to
Exit Graceful Restart", Point 1 "Router X has reestablished all its
adjacencies".
>
>I am trying to identify the corresponding event from RFC 2328 for Router X.
Is it one of the following:
>a> Upon installation of an LSA in the database [RFC2328, Sec 13.2] [and
thereby tracing the database for all the relevant LSAs which indicate the
previous adjacencies],
>     or
>b> State of neighbor Y changing to Full; [and thereby tracing the database
...]
>     or
>c> Implementation specific
>
>Let me explain this with a simplified example:
>
>            X ----- lan ----- Y
>
>In the topology, we have two routers, X and Y; and Y happens to be the DR
in the network segment.
>
>Now Router X restarts. After 'T' seconds, Router X becomes fully adjacent
with Router Y.
>
>At this point, three scenarios are possible:
>1> X had received all the relevant LSAs (i.e. router-LSAs of X, Y and
Network-LSA of Y) BEFORE becoming fully adjacent with Y.
>     Hence, the graceful-restart exit event has already been fired BEFORE X
became fully adjacent with Y.
>2> X had received all the relevant LSAs BEFORE becoming fully adjacent with
Y;
>     but graceful-restart exit event gets fired EXACTLY when X became fully
adjacent with Y.
>3> X had completed receiving all the relevant LSAs AFTER becoming fully
adjacent with Y;
>     Hence, the graceful-restart exit event gets fired sometime AFTER X
became fully adjacent with Y.
>
>
Hi Depopam,

As stated in RFC 3623 section 2.2, a restarting router uses its
pre-restart router
LSA to determine whether or not it has re-established all its
pre-restart adjacencies.
Adjacency formation implies database synchronization and reception of
all LSAs within
the flooding domain of the router with whom the adjacency is being
formed. Hence,
 #2 above is the only scenario which makes sense.

Hope this helps,
Acee


>Thanks,
>Debopam
>
>