Function Metrics

Collects execution metrics for the annotated function.

The following metrics are collected:

• Execution time.
• Success (if function returns a core::result::Result-like type.

The metrics are sent to recorders, which can be thought as backends, which can send for metric aggregation services, for example.

Currently, there are the following recorders implemented:

• Log: simply log the metrics.
• Prometheus: collect the metrics so they can be sent to Prometheus.

lib.rs:

Function Metrics

Collects execution metrics for the annotated function.

The following metrics are collected:

• Execution time.
• Success (if function returns a core::result::Result-like type.

Usage

Include the dependency and enable the desired recorders.

For example, the Prometheus recorder can be enabled like the following:

metrics-fn = { version = "0.1", features = [ "record-prometheus" ] }

After that, just annotate the desired functions with #[measure].

use std::thread::sleep;
use std::time::Duration;

use metrics_fn_codegen::measure;

#[measure]
pub fn super_slow_function(a: u64, b: u64) {
	sleep(Duration::from_millis(a * b));
}

Notice that there is a performance impact due to the metric collection on every call to the function. Thus it is advised to use the annotation on functions where the performance impact is negligible, like on function that makes calls to external services or database queries.

Recorders

The metrics are collected and sent to recorders, which can integrate with metrics systems like Prometheus.

In order to keep dependencies to a minimum, for each recorder there is an accompanying feature. These features are composed by record- and metrics system it enables. By default, all recorders are disabled.

The following metrics systems are currently supported:

• Log (record-log): simply log the collected metrics, using log::info!; not really a metric collection system.
• Prometheus (record-prometheus): collect metrics to the default Prometheus metric registry, using the Prometheus crate.

Log Recorder (record-log)

The log recorder is not really useful: it was created to help with development.

It simply logs the metrics using log::info! or log::error!.

Prometheus Recorder (record-prometheus)

The Prometheus recorder collects metrics about the functions annotated with #[measure] to the application_method_timings histogram metric.

This metric has the following labels:

• mod: module name.
• fn: function name.
• res: result (Ok or Err if the annotated function result is named Result; always Ok otherwise).

The following buckets are used: 0.005, 0.01, 0.025, 0.05, 0.1, 0.25, 0.5, 1.0, 2.5, 5.0, 10.0.

In order to expose the collected metrics, a web application framework like Rocket or Actix Web must be manually included and configured.

See metrics-fn-example-prometheus for an example.

Versioning

Semantic Versioning is used. Changes to the collected metrics are also considered in the versioning.

To-Do List

The following are planned features, by priority:

This list is not a commitment: items might be added or removed of the list.

• [MEDIUM] Record type name associated with impl.
• [LOW] Configurable Prometheus metrics names and buckets.