Right now the Kube deployer exports all metrics every second. The metrics plugin can decide how to manipulate these metrics.
However, there are 2 knobs that I think it makes sense for the user to have control on:
how often to export metrics from the deployer side
whether the user wants auto-generated metrics
Reasons why these 2 knobs are important to be controlled by the user:
simpler for the user to write a metrics plugin; if we export every second, and they want to export to Prometheus every 30 seconds, they have to write additional logic to buffer and aggregate the metrics on 30 seconds windows
auto-generated metrics are very expensive; e.g., our histgram metrics have large number of buckets that lead to very expensive metrics; one option is to tune the number of buckets for the auto-generated metrics; however, this might be tricky; given that the auto-generated metrics might be only "sometimes" useful to the user, it's better to not export them for now, unless the user explicitly wants to do that
Right now the Kube deployer exports all metrics every second. The metrics plugin can decide how to manipulate these metrics.
However, there are 2 knobs that I think it makes sense for the user to have control on:
Reasons why these 2 knobs are important to be controlled by the user: