Add metrics and traces #111

New Issue

Closed

opened 2021-09-25 15:51:19 +00:00 by maximilien · 1 comment

maximilien commented

2021-09-25 15:51:19 +00:00

Owner

The current trending framework in this field is OpenTelemetry, an open data format that can be leveraged by multiple vendors.

On the metric side, the openmetrics format is very similar to the prometheus data format.

My proposition is to implement https://github.com/open-telemetry/opentelemetry-rust in garage, and:

expose a openmetric endpoint with runtime information
have a default tracing framework in place with at least HTTP requests
design an easy-ot-use harness enabling devs to leverage those on their laptop and documenting a production usage

The idea here is to enable Garage cluster operators and developpers alike to benefit from observability feature, namely metrics and tracing, allowing them to better understand what is happening in a production cluster. The current trending framework in this field is OpenTelemetry, an open data format that can be leveraged by multiple vendors. On the metric side, the openmetrics format is very similar to the prometheus data format. My proposition is to implement https://github.com/open-telemetry/opentelemetry-rust in garage, and: 1. expose a openmetric endpoint with runtime information 2. have a default tracing framework in place with at least HTTP requests 3. design an easy-ot-use harness enabling devs to leverage those on their laptop and documenting a production usage