OSPF does not really have a holddown period. It does have maximum
LSA advertisement rates: any LSA can be updated at most
once every 5 seconds (MinLSInterval), and this is enforced by
the LSA originator's neighbors, who won't accept updates for a
given LSA more than once a second (MinLSArrival).

John

-----Original Message-----
From: Mike Yeung [mailto:myeungINNOVANCE.COM]
Sent: Tuesday, May 15, 2001 7:27 PM
To: OSPFDISCUSS.MICROSOFT.COM
Subject: Route flaps and Holddown period

Hi all,
    I have a few question related to OSPF operation after a failure, please
give me your opinion on them.

1) Hold down period - After a router received an LSA stating a neighbor is
no longer reachable, the router would perform SPF and advertise of this
changes to its neighbor. Then the router enters a hold down period where no
additional updates (LSA) are accepted. Does this hold down period applies
only to update on the neighbor that was previously claimed unreachable or
does the hold down applies to update to all neighbors?

2) Route flap can happen when the hold down period is shorter than that
required for network convergence. Does anyone else know of other scenarios
where route flap can occur as result of LSA updates?

3) Could the following be an example of a route flap scenario as result of
OSPF advertisement.
- After a network failure, as part of network convergence operation a router
receives two LSAs from two different neighbors. The first LSA advertises a
route that is better than what is in the LS database for a given
destination. The second LSA advertises a route that is better than what was
advertised in the first LSA to the same destination. Therefore, triggering
the router itself to perform two SPF calculations and two LSA advertisements
for a given destination. Personally, I don't think it is possible given
every router know about all networks within the same OSPF area.

4) What is a typical hold down period in an OSPF network? I saw something
like 10 seconds.

5) I know this is highly topology dependent, but say if your network is
partially meshed with a single OSPF area and 250 routers. Would 10 seconds
be enough with the assumption of negligible propagation delay? My feeling
is more like 20 seconds for full network convergence.

thanx in advance,
Michael

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