opinionated_telemetry: Easy-to-use backend for metrics and tracing

This is a single crate that provides a simple, opinionated backend for metrics and tracing. The goal is to easily enable an existing Rust app to emit metrics and traces, without having to spend a lot of effort on configuration.

Documentation

This crate's opinions

All opionated software should try to list what opinions it holds, so users know whether it's a good fit for them. Here are the opinions of this crate:

• Tracing and metrics are incredibly useful, and should be ubiquitous.
  • This includes CLI tools, not just servers!
• Prometheus and Grafana are a solid combination for metrics and dashboards.
  • CLI tools should normally use prom-aggregation-gateway instead of Prometheus pushgateway.
• OpenTelemetry and W3C Trace Context are a popular choice for tracing.
  • But tracing backends are less standardized than Prometheus.
• Rust's tracing and metrics façades are good enough to handle the basics, and reasonably standard.
• "Labels" on metrics should be carefully chosen and "low-arity" (having few possible values). Therefore, inheriting labels from parent scopes is almost always the wrong thing to do.

Supported backends

Tracing:

• Vendors
  • Google Cloud Trace
• Jaeger (not yet supported, but we'd love a PR)
• Debug (printed to stderr)

Metrics:

• Prometheus (scraping)
• Prometheus (push gateway)
• Debug (logged via tracing)