Difference between nginx-ingress and yabeda metrics

[vterdunov]

Hi. I have a problem with metrics value consistency. There is a backend deployed in k8s. Also there is a nginx-ingress controller with metrics enabled. Let's compare RPS through the backend service.

Here is yabeda graph

Here is a ingress-nginx graph

Looks like yabeda's metrics are doubled.

[pasl]

In my setup I have the same conclusion looking at rails_requests_total.

For 1 get query the rails_requests_total counter get +2 each time.