Metric values no longer increasing monotonically

[breed808]

With the metric clearing introduced in #177 , metric values are now increasing or decreasing over time. I.E. not behaving like Counters.

Example:

I think there's two options to consider here:

1. Reverting the change in Fix memory leak: clear stale metrics for each zone/account before fetching #177 and continuing to Add() observed Cloudflare values to the Counter metrics, or
2. Switch affected metrics from CounterVecs to GaugeVecs and Set() the metric based on the observed value from Cloudflare.

I'm partial to the second option, as the exporter has been observing requests over a time period, rather than request totals. As a result I think Gauges better reflect the data being exposed here.

That said I'm just a single user of the exporter, so it'd be worth discussing here.

EDIT: I should mention I'm happy to raise a PR if the second option is desirable.

[jonnangle]

I'm partial to the second option, as the exporter has been observing requests over a time period, rather than request totals. As a result I think Gauges better reflect the data being exposed here.

I agree - it also helps to avoid the confusion that can arise when people take the rate() of a counter and notice differences between what they see in Prometheus and what's shown on the dashboard in Cloudflare Analytics.

[haad]

Please do so I will merge it right away.