Recommendations for integrating with Prometheus?

IIUC, if you want to get metrics from rfc39 into prometheus, you configure rfc39 to run with a generous metrics_delay, and configure prometheus to scrape the endpoint that exists while rfc39 is running.

IMO, this is a bit funky:

• It's makes it hard to monitor how long the rfc39 job ran for (because it just sleeps for a while at the end)
• When the job is not running, then prometheus scrapes fail, which makes it hard/annoying to monitor for failed scrapes. See https://github.com/NixOS/infra/issues/551 for details.

I've done some research, and this doesn't seem to be how Prometheus recommends scraping batch jobs. Prometheus recommends either using Prometheus Pushgateway or Node Exporter's textfile collector.

Is it possible to use one of those approaces with rfc39? I see there's a --dump-metrics option, is that meant to be used with the textfile collector?

Start with src/cli.rs and src/main.rs, especially --dump-metrics, metrics_delay, and the current endpoint behavior. Compare the Prometheus Pushgateway and node_exporter textfile collector approaches mentioned in the issue, then clarify which integration is wanted; done would require an agreed design and documented acceptance criteria.