[c-nsp] SNMP Counter Oddness with SUP720-3BXL, 12.2(33)SXH

[Adam Rothschild]

Howdy,

I've been noticing some SNMP-related weirdness of late, whereby a box
will report erroneously high interface counter data every five (5)
minutes on the nose.

This trend seems to occur with Gigabit and Ten Gigabit physical
interfaces on {inbound,outbound} {octet,packet} counters (polled via
SNMPv2, 64-bit).  Interestingly, SVIs ('interface VlanXXX') do not
appear impacted.

The machine in question is a Cat6506 with SUP720-3BXL and an
assortment of 67xx (CFC) blades, running 12.2(33)SXH.  RP and SP CPU
loading is sane, and nothing's particularly out of the norm.  (Well,
the purportedly-resolved "default-originate ... doesn't" bug appears
still present, but that's another can of worms for another day. :-)

I've been told, in broad strokes, that these symptoms match a known
bug, something to do with communication between the LC and RP/SP under
load; I've been unable to identify specifics by rummaging through SR
and SX train release notes.  Any advice, including work-arounds and
suggested code changes, would be greatly appreciated.

Thanks in advance,
-a