Monitor the number of busy Unicorn processes

I see that in gitlab-com/infrastructure#1883, we are experimenting with adding more app servers. Do we know how many unicorn workers are busy for any given node? If not, one technique is to use Raindrops in such a way:

We should log these into Prometheus somehow (e.g. via InfluxDB or statsd if the built-in unicorn scraping isn't ready yet).

Edited by Craig Miskell