[c-nsp] VRFs and redirect cache...

[Jeff Kell]

In the process of chasing down an odd problem earlier this week, I ran
up against a grey cloud perhaps someone can clarify.

We had moved an internal NTP-configured interface (loopback) that some
of our gear was configured to use as a reference server.

The disappearance of the /32 route led to taking a default route, which
in our topology generated a redirect to another gateway (FWSM) which was
then denying the connections.

Tracking back to the switches in question and "show ip redirect"
indicated the cached redirect information.

"clear ip redirect" removed the problem.

But there seems to be only one "redirect cache", that's not a VRF-aware
thing on the Catalysts.  Are redirects only done by the global VRF? 
What's up with that?

I can disable redirects and avoid the issue (at some extra-hop cost when
forwarding to a non-routing ASA that can't announce a default route),
but curious how redirects are handled in a multi-VRF scenario.

Jeff