Ops platform / Observe
Metrics
The numbers your services expose, charted per service, and what each container uses now.
Metrics charts one metric at a time for the environment, one line per service, over a range you pick. A counter is drawn as a rate per second, a gauge as its average.
Under the chart, every container of the environment shows the CPU and memory it uses right now. That needs nothing from your service.
Exposing your own
Ops reads metrics in the Prometheus text format from a service's /metrics path every 15
seconds. With the Rust ops-sdk, the observability feature has a middleware that counts
requests and a handler that serves /metrics.
Reading a new service's /metrics is switched on by whoever administers your ops; it is
not a setting in the console yet.
How long
Full detail is kept for three days, and five minute averages for thirteen months. A range longer than a day is drawn from the averages.