[c-nsp] Strange ARP behaviour

[David Coulson]

I'm having a weird issue with a 7507/RSP4 (Router A) running 12.2(25)S5.
There is a subnet defined, which is redistributed into another router
(Router B) using OSPF. I have a single host within this subnet, which
I'm running tcpdump on. I've also got a Router C, which has the same
problem as Router B.

If I try to ping the host from router B, I don't get any response. A 'sh
ip ro x.x.x.x' shows a normal looking routeing entry:

#sh ip ro 207.166.219.197
Routing entry for 207.166.219.192/28
  Known via "ospf 10", distance 110, metric 2, type intra area
  Last update from 207.166.192.3 on FastEthernet0/0, 00:10:36 ago
  Routing Descriptor Blocks:
  * 207.166.192.3, from 207.166.219.2, 00:10:36 ago, via FastEthernet0/0
      Route metric is 2, traffic share count is 1

If I traceroute to it, I see the correct nexthop, but it dies at that
point. Strange thing is, I NEVER see an ARP request on the host I'm
ping/tracing to.

If I ping direct from Router A, I drop the first packet, probably due to
the ARP lookup, then I can throw 1000 packets at it and not drop any.
More strange, I can now ping/trace to the host from Router B without any
packet loss. As long as the IP is in the ARP cache of Router A, I don't
have any problems - Changing the IP, or dumping the ARP cache, and I'm
back to square one where I can't hit it at all from Router B.

I'm guessing this is probably an IOS bug or something, but I'm having
difficulty tracking it down. Any suggestions? I've got no access-lists
or weird route-maps at all.

David