From: Pamela Forsyth (pforsyth@xxxxxxxxx)
Date: Wed Jan 03 2001 - 10:33:01 GMT-3
Chuck,
It makes sense that the virtual links would re-establish themselves in case
of a failure within the area along the best path. The virtual links are
defined between border routers by router ID and are entirely dependent on
the best paths in the transit area. If every router in the area has all
the link states they all know which routers are attached to which subnets,
so as long as there is a path to the opposite router the virtual link will
remain up. The path between them is whatever best path the Dijkstra
algorithm is able to determine at the time.
What doesn't make sense is that the path doesn't seem to revert to the
original best path when the failed link is restored. When the link was
restored, the router(s) experiencing the link change should have sent out a
new router LSA with new information which should have gotten flooded
throughout the area, the Dijkstra algorithm should have run on all the
routers, and a new path between the border-routers should take effect. The
path of the virtual link MUST be entirely dependent on whatever the routers
have in their routing tables at the time--the best path--which should have
been the same as it was originally. Otherwise, in a more complex topology
than the one you have drawn out, a loop might potentially occur. Just out
of curiosity, were the alternate paths through area 1 of equal cost?
Another curious question--which IOS versions were you using on the
routers? I've seen OSPF do some really strange things with IOS version
12.0(x). ;-)
Pamela
At 12:18 AM 1/3/01 -0800, you wrote:
>For lack of a better term.
>
>Got to fool around with a few more routers than normal. So I set up a
>contrived OSPF topology for the sole purpose of seeing if virtual links can
>be dynamic.
>
>The short answer - yes they can
>
>The long answer - in a situation where the transit area terminates to the
>same ABR via more than one path, the virtual link can be dynamic, in that if
>one path to the ABR goes down then the link reestablishes itself via the
>other path.
>
>Router---area_1-----router-------area1
>| | \
>area_3 area_1 \
> | --------router_ABR----area_0
>
> pretty neet, albeit a bit impractical. But it is a question I've not been
>able to test empirically until now. The area 0 ABR router will apparently
>reestablish the virtual link within just a couple of seconds ( in this lab,
>anyway ) The cost of the virtual link was adjusted to reflect the cost of
>the underlying OSPF path. By issuing shut and no shut commands on an
>appropriate interface I was able to look at the results. The original
>virtual link path would not reassert itself when the blocked interface came
>up. I would have to issue a shut on the interface of the replacement path.
>So I conclude that a virtual link is about as permanent as it gets. The path
>remains static, until such time as it disappears, at which point the link
>will reestablish itself using the alternative path.
>
>Whelp, back to the books.
>
>Chuck
This archive was generated by hypermail 2.1.4 : Thu Jun 13 2002 - 10:27:21 GMT-3