[prev in list] [next in list] [prev in thread] [next in thread] 

List:       ms-ospf
Subject:    Re: Aging out entries..
From:       "John T. Moy" <jmoy () CASC ! COM>
Date:       1998-09-22 20:29:39
[Download RAW message or body]


Suresh-

In OSPF you NEVER remove an LSA from the database unless
it ages out. The flooding process is always independent
of the routing calculation. If you try to remove LSAs
that are not used in the routing calculation, one of two
bad things may happen, depending on whether you inform
your neighbors (through premature aging) that you are
removing the LSA.

If you don't inform your neighbors, your database may not
get back the LSA when it is usable in the routing calculation
- there is no guarantee that LSAs will be reflooded just
because the routing calculation has changed.

If you do inform your neighbors that you are deleting the LSA,
then you may get into a flooding war with the LSA's originator.
If the routing calculation can't reach a router, it
does NOT necessarily mean that you can't flood LSAs to that router.

John

skatukam@hypercom.com wrote:

>      If a router R1 receives 'N' number of updates from an adjacent router
>      R2, what happens if R1 and R2 are disconnected??
>
>      Does R1 remove all entries learned from R2?? OSPF does not specify
>      anything about it.
>
>      I think a database entry should be removed only if the entry is not
>      used in SPF computation at all. Currently, as per specification, an
>      entry is removed only when it is aged out. The age value is 1 hour.
>      The advantage of keeping that entry is that it may come back in which
>      case it does not have to be exchanged, if there is no change. The
>      disadvantage of it is, it takes up memory even though it may not be
>      necessary and probably gives a wrong impression if you are trying to
>      debug a problem.
>
>      Thanks,
>      Suresh

[prev in list] [next in list] [prev in thread] [next in thread] 

Configure | About | News | Add a list | Sponsored by KoreLogic