[j-nsp] Rate limit ARP per interface (or JUNOS bug)?

[Chris Adams]

I had a problem with a metroE circuit today where the provider screwed
up the link and had it looped back to me (so every packet I sent came
right back).  The link connects to a switch and comes into my Juniper
M10i as a VLAN on an 802.1q trunk.

The problem was that my router was sending ARP requests out, getting
them back, and sending them out again, looping madly.  On my monitoring
system, it looks like we hit 1500 packets per second doing this.

While this doesn't appear to have had any effect on the CFEB or RE CPU,
it appeared to have caused problems with other ARP requests (I'm
assuming as other ARP entries expired).  I had various servers (on other
ports on the same PIC) have traffic problems that cleared themselves up
(and the problems stopped when I killed the problem metroE link).

Is this behavior a JUNOS bug or am I supposed to be rate-limiting ARP
requests (on a per-VLAN basis) somehow?
-- 
Chris Adams <cmadams at hiwaay.net>
Systems and Network Administrator - HiWAAY Internet Services
I don't speak for anybody but myself - that's enough trouble.