My two cents...

It is possible that with database overflow, that some routers
will actually have more LSAs in their database than others.

I am thinking about Opaque LSAs (RFC 2370) allowing opaque
capable routers to contain opaque LSAs.

Assuming that these three LSA types are enough to force a
database overflow, it is concieveable to drop all the adjacencies
with the opaque bit set and to re-establish the adjacency without
having the opaque bit set... It may make sense to keep a
counter of opaque LSAs in the lsdb.. Other LSA counters probably
couldn't hurt either..

Be careful if the adjacency is dropped for less than router
dead interval that the change in the opaque bits are seen. A
cleaner change is drop the adjacency for enough time (greater
than routerdeadinterval) and re-establish the adjacency.

To properly do this, it also concievable to wipe away just
the opaque LSAs, before the opaque bit is set, and to keep
the other ones.. Less time to re-establish the full adjs..

Once all the full adjs are re-established, then run the spf
calculation...

Mitchell Erblich
===================

Hi Amit,

If you note Section 1. in the RFC it clearly states that the value of
ospfExtLsdbLimit should be the same in all routers in
the routing domain. So the ASBR which is actually attempting to put the
external LSA in the domain will be the first one
to overflow, so it will not originate the LSA itself(the situation you
describe will not occur itself).

This RFC is useful, especially cuz we have the entire BGP routing table
acually pumped in into the OSPF domain
inadvertently(which supposedly happens a few times, on the live
networks, a year).

I guess if such a state exists when there are routers which cannot keep
the entire database, the network topology/route
leaking/routers need a change instead.

Thanks,
Vishwas

-----Original Message-----
From: Amit Srivastava [mailto:ospfisfunYAHOO.COM]
Sent: Tuesday, June 18, 2002 7:35 PM
To: OSPFDISCUSS.MICROSOFT.COM
Subject: Re: OSPF Retransmission Doubt

     Hi Vishwas!

               Actually after reading the RFC 1765 only i got this
doubt.Here i feel once the
     OSPFEXTLSDBLinit variable is exceeded the router should indicate
other routers that its database is
     full and they should not keep on retransmitting the LSA which will
clog the n/w bandwidth.Once the
     router comes out of the database overflow state it should indicate
to the other router that now they
     should send the LSA.

           What do you say about this??

     Regards

     Amit

       "Manral, Vishwas" <VishwasMNETPLANE.COM> wrote:

            Hi Amit,

            You may like to look at the RFC1765 - OSPF Database
Overflow.
            http://www.ietf.org/rfc/rfc1765.txt This RFC tells a way to
deal with such
            overflow when it occurs in OSPF.

            Also as the RFC lists as enhancements, if the number of
summary LSA's
            increases we could as well flush type3/4 also and originate
default routes
            in the domain instead, which could lead to suboptimal
routing or blackholing
            however.

            However in ur case if a router does not accept and
acknowledge an LSA, it
            will be continuously retransmitted as long as the adjacency
is up. If this
            is not done, the databases would have no way to come back to
synch as the
            adjacency itself would not be broken.

            Thanks,
            Vishwas

            -----Original Message-----
            From: Amit Srivastava [mailto:ospfisfunYAHOO.COM]
            Sent: Tuesday, June 18, 2002 5:26 PM
            To: OSPFDISCUSS.MICROSOFT.COM!
            Subject: OSPF Retransmission Doubt

            Hi All,
            I have the doubt which is described below:-
            Suppose a router is there whose database is full.Now
            it can send and receive hello packets but discards any
            new LSA coming to it(as database is full) and hence it
            will not ack that LSA.Now my doubt is that RFC 2328
            says the the retransmission will continue till the
            adjacency is broken.
            Does that mean if the database does not have space
            for accomodating the new LSA infinite number of
            retransmission would result???
            Regards
            Amit

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