search

metrics-rs / metrics

A metrics ecosystem for Rust.
MIT License
1.08k stars 148 forks source link

Specific endpoint for prometheus exporter health checking #435

Closed biancamip closed 7 months ago

biancamip commented 7 months ago

Problem: A service that installs a prometheus metrics exporter will return all metrics data on any request, regardless of the endpoint requested. This presents an issue when doing health checks using a service networking solution such as Consul, which performs requests on short intervals and spreads the response status + body across all server nodes. Since the response body contains all metrics collected and will probably change from one check to the next, the issue is that this behavior will generate a high network traffic, resulting in delays, elevated costs, and so on.

Solution: For health checking, the response could be as simple as a successful status code with a short body content. I've implemented this for the path "/health", while maintaining the previous behavior for all other cases.

If necessary, this solution could be adjusted to allow configuration of the health path on PrometheusBuilder struct initialization, and it could also be disabled in case the current behavior is desired, but I think this minor change would make significant difference for now.

tobz commented 7 months ago

@biancamip Apologies on the delay reviewing this. 👋🏻

I've made two small little changes, but otherwise, I think this is fine to merge as soon as CI passes.

tobz commented 6 months ago

Forgot to update the changelog prior to release, but this has now been released as part of metrics-exporter-prometheus@v0.13.1.

Thanks again for your contribution. ❤️