Currently our codebase tracks two separate metrics, lets say a.b.c.success and a.b.c.failure. We always increment the metric by a value of 1 each time.
Is it necessary to separate these metrics? Or can we "increment" the metric by a value of 0 for errors, and 1 for success, and if we wanted a total count of failures, we could subtract the sum in a time period from total count?
I tried to confirm this could work through the metric explorer by looking at graphs for both sum:a.b.c.success{*}.as_count() and count:a.b.c.success{*}.as_count(). Since I'm always incrementing the value by 1, I would expect these values to be the same but the count value is significantly higher (and always in increments of 10 per time slice?)
