|
Neohapsis is currently accepting applications for employment. For more information, please visit our website www.neohapsis.com or email hr@neohapsis.com |
From: Moy, John (John.Moy
SYCAMORENET.COM)Date: Thu Sep 20 2001 - 13:35:42 CDT
Phil-
I personally wouldn't bother trying to identify the previous
adjacencies. Just send a Link State Update containing the
grace-LSA out all interfaces. If you want to try extra hard, send
it a couple times; if by some chance the (previously) adjacent
neighbors don't get it, the worst that will happen is that you'll
get the normal OSPF restart procedure.
That said, for OSPF Hitless Restart to work, you really only
need to flood grace-LSAs on those adjacencies that are reflected
in LSA contents. (For example, no grace-LSA need be send between
the backup DR and a router in DR-other state). And those adjacencies
*can* be identified just by examining the router-LSAs and network-LSAs
that the restarting node had itself originated.
John
> -----Original Message-----
> From: Phil Ilett [mailto:phililett
YAHOO.CO.UK]
> Sent: Tuesday, September 18, 2001 4:58 PM
> To: OSPF
DISCUSS.MICROSOFT.COM
> Subject: draft-ietf-ospf-hitless-restart-01.txt - determining
> (previous)
> full adjacencies
>
>
> This may appear obvious to some, but can anyone tell me
> whether a complete list
> of FULL adjacencies can be determined given (just) the
> router-LSA (for all
> areas).
>
> I'm trying to work out how much state I need to keep on a
> redundant (stand-by)
> processor in order that it can attempt to perform hitless
> restart in the event
> of its primary cpu failing.
>
> It would be nice if I didn't have to store (the primary's)
> adjacency state on
> the stnd-by and settle for only keeping (duplicates of) the
> primary's router
> LSAs, but I can (presumably) only do this if I can recover a
> list of all
> (pre-restart) full adjacencies so I can flood grace LSAs to
> them when my
> primary cpu fails.
>
> Another question: in a cryptographic environment, where the
> crypto sequence
> numbers need to be kept, am I right in thinking that
> (a) there are really two sequence numbers to consider (those
> that this router
> is using when send to its neighbour and those its neighbour
> is using when
> sending to this router) and,
> (b) that this (pair of) sequence number(s) is maintained
> per-adjacency.
>
>
> Thanks,
> Phil.
>
>
> ____________________________________________________________
> Do You Yahoo!?
> Get your free
yahoo.co.uk address at http://mail.yahoo.co.uk
> or your free
yahoo.ie address at http://mail.yahoo.ie
>
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]