metricµs

Low-latency metrics framework.

Crates

• metricus: core types and the Metrics backend trait (Counter, Histogram).
• metricus_agent: metrics backend that uses background aggregator + exporters (UDP, file, unix sockets).
• metricus_allocator: optional counting global allocator.
• metricus_macros: #[counter] and #[span] helpers.

Quick start

Initialize the default agent backend, then create counters and histograms with tags.

use metricus_agent::MetricsAgent;
use metricus::{empty_tags, Counter, CounterOps, Histogram, HistogramOps};

fn main() -> metricus_agent::Result<()> {
    MetricsAgent::init()?; // start background aggregator with default config

    // create counter with no tags
    let requests = Counter::new("requests_total", empty_tags());
    requests.increment();

    // create histogram with tags
    let latency = Histogram::new("request_latency_ns", &[("service", "api"), ("route", "/v1/orders")]);
    latency.record(1_250);

    Ok(())
}

Macros let you attach counters or spans directly to functions. This will automatically add fn_name tag with instrument method name.

use metricus_macros::{counter, span};

#[counter(measurement = "requests", tags(service = "api"))]
fn handle_request() {
    // work
}

#[span(measurement = "latency", tags(service = "api"))]
fn handle_request_with_span() {
    // work
}

When you want to time a single block, create a span directly from a histogram.

use metricus::{Histogram, HistogramOps};

fn handle_request(histogram: &Histogram) {
    {
        let _span = histogram.span();
        // do the timed work, span records on drop
    }
    // remaining work
}

Custom backends

The project ships with metricus_agent backend that uses background aggregator and various exporters. If you wish to use your own custom backed you need to implement metricus::Metrics and register it via metricus::set_metrics.