Operators of older, legacy APIs may want to identify which WebActions have not had any requests in recent history. This aides in pruning unused code.
Currently, the http_request_latency_ms count metric can be used to identify WebActions (action label) that are actively used. To my knowledge, no metric exists which allows identifying unused endpoints.
I propose we add a Gauge with an action label which is incremented from zero to one during application startup for each WebAction registered/bound via WebActionModule and WebActionEntry. Each instance of the Misk application (i.e. each container, pod, server, etc.) will report the WebActions it is serving.
This will enable operators to compare the set of WebActions available to the set which is actively used.