At the moment we are running an inference server in azure machine learning. We would like to be able to expose existing metrics to prometheus as well as create our own custom metrics, all described below
Expose existing metrics:
I would like a breakdown of number of requests by status/average latency etc... I assume azure keeps track of this somewhere but I can't see it anywhere, even if I could I want to expose it so prometheus can scrape it.
Expose custom metrics:
Our inference code has a number of slow steps, we want to be able to expose metrics about each section of this code so that we can see which parts of our application are the bottleneck.
