I'm doing a large-scale snmpwalk against an ASR9k (with IOS-XR v4.2.0) running as a provider edge router (full bgp table) and pulling the full contents of the BGP route table. On other routers, this completes within my timeout window, but not on the ASR9k.
Figuring that this has to do with CoPP rate-limits, I've adjusted the rate-limits to ridiculously high values. E.g.
lpts pifib hardware police location 0/0/CPU0
flow snmp rate 10000000
But still, the walk doesn't complete in an acceptable amount of time. Manual snmpwalks display a rate slower than even 7600s, with occassional stutters. CPU on the box doesn't even register that anything extraordinary is going on (@ 2 - 3%), and "show lpts pifib hardware police location" shows that there are 0 drops against SNMP.
I figure it's likely a knob I haven't turned yet - either some traffic shaping mechanism or some combination of process scheduling/priority with SNMP.
Any ideas out there?
Thanks,
-- Gil
