Prometheus exporters should fetch the data on demand
Currently, we periodically aggregate the metrics for Prometheus purposes using internal Stork intervals. It means the Prometheus collector always pulls a bit of outdated cached values.
We should avoid using the internal collecting loop and our own intervals. The Prometheus collector should be the one to decide when and how often the data are fetched.
-
The Stork agent should send the statistics-get-all
command and process the response on request to the/metrics
endpoint -
The Stork server should retrieve the adjusted utilizations and machine counters on request to the /metrics
endpoint