I'm not sure why you roll up interface errors and discards to hourly and daily sums in the interface table but by doing so you make it impossible to differentiate between a line being bounced and generating 100's or 1000's of errors in an instant and a line that is generating errors every polling cycle. In order to work around this and create an alert based on the latter I have to create a UnDP based on the exact same OID you are already polling in order to get the responses into a table where I can set an alert based on something like 1000 or more inbound errors for 3 consecutive polling cycles. If there is some way to do this other than via a device poller let me know, otherwise please consider this feature request.
