Curtis,Erblich,

Could we focus on the key question of
cascaded restarts, how to make sure the topology
is stable before and after each restart ?. Automatically.

How to make a deterministic planning, so we can know
where the holes are and what is the impact.

Has anyone thought on these lines ?

Thanks,
Abhay




----- Original Message -----
From: "Curtis Villamizar" <[EMAIL PROTECTED]>
To: "Erblichs" <[EMAIL PROTECTED]>
Cc: "Abhay D.S" <[EMAIL PROTECTED]>; <[email protected]>
Sent: Tuesday, October 23, 2007 12:00 PM
Subject: Re: [OSPF] Cascaded Graceful Restarts with NSF for
OSPFprotocolQuestion


>
> In message <[EMAIL PROTECTED]>
> Erblichs writes:
> >
> > Abhay D.S.,
> >
> >
> > Actually it is just a bit more
> > complicated.
> >
> > With OSPF, we should start with a case
> > of two routers.
> >
> > How does the restarting router know
> > that the restart is done and how does
> > the other router know that the restart
> > is done?
> >
> > The restarting router can determine
> > whether NSF and restart is done. Whether
> > restart required a a standard resync of
> > the LSDB is almost irrelevant.
> >
> > So, the question is why must Gracefull
> > be successful? It basicly just "reduces"
> > overhead while NOT distrupting forwarding.
>
> More important it does not slosh all of the BGP routes (hundreds of
> thousands of them) away from the restarting router and then back.
> In an IGP only transient routing nasties due to inconsistent notion of
> whether that router is up or not is something to be avoided (you know,
> transient loops, transient blackholes, and all that).
>
> > Thus, It is done when it re-starts sending HELLOs
> > and re-extablishes full adjs from the perspective
> > of the 2nd router.  The hellos are the aliveness
> > criteria from my earlier post.
> >
> > ` So, you MIGHT want to track the nbr FSM
> > if you have access to the OSPF source code.
> >
> > Mitchell Erblich
> > ----------------
> >
> >
> > "Abhay D.S" wrote:
> > >
> > > Mitchell,
> > >
> > > 2) Their is no GUARANTEE that any Restart will
> > >    be sucessful, that the LSDB will be stable and/or
> > >    that helper(s) exist.
> > >
> > > I wanted an automated and deterministic number 2.
> > >
> > > Any ideas or any attempts in the direction of 2 ?
> > >
> > > With Regards,
> > > Abhay
>
>
> Looks to me to be more like a windmill than a dragon.
>
> Curtis


_______________________________________________
OSPF mailing list
[email protected]
https://www1.ietf.org/mailman/listinfo/ospf

Reply via email to