scrape

One fetch by Prometheus of a service's metrics page over HTTP, whose values are then stored with a timestamp.

In more detail

The service only publishes the page; it does not know Prometheus exists and does not buffer, retry or push anything. Because Prometheus does the fetching, it knows whether each fetch worked and records that in the up metric. Rendering the page must be cheap: no database calls while building it.

Example

curl -s http://localhost:8001/metrics | grep http_requests_total does by hand exactly what a scrape does.

Related terms

Taught in

References

Certification courses