Struct PrometheusMetrics 

pub struct PrometheusMetrics { /* private fields */ }

In-memory Prometheus-format HTTP metrics collector.

This collector stores counters and bounded duration histograms in process memory and renders Prometheus text exposition. It is useful for small services, examples, and tests; it is not a durable metrics store and values reset on process restart. The default exclusions are /health/live, /health/ready, and /metrics.

Label cardinality

By default the collector records every distinct route label it observes, so the caller is responsible for keeping cardinality bounded — prefer route patterns (e.g. "/users/:id") over concrete paths. To harden against accidental high-cardinality labels (which would grow memory without bound in a long-running process), apply PrometheusMetrics::with_max_series: once the configured number of distinct route labels has been admitted, every further distinct label collapses into a single "<overflow>" route.

use axum::{Router, routing::get};
use nidus_http::middleware::{PrometheusMetrics, route_metrics_layer};

let metrics = PrometheusMetrics::new();
let app = Router::new()
    .route("/users/:id", get(show_user))
    .route_layer(route_metrics_layer("/users/:id", metrics.clone()))
    .merge(metrics.routes());

Implementations

impl PrometheusMetrics

pub fn new() -> Self

Creates a Prometheus metrics collector with default internal route exclusions.

The collector is unbounded by default (every distinct route label is recorded); use Self::with_max_series to cap label cardinality.

pub fn exclude_route(self, route: impl Into<String>) -> Self

Adds a route pattern to exclude from recording.

Match the exact route label emitted by the metrics layer, such as a static route supplied to route_metrics_layer or an Axum matched path.

pub fn with_max_series(self, max_series: usize) -> Self

Bounds the number of distinct route labels retained in memory.

The first max_series distinct route labels are recorded normally; any further distinct label collapses into a single shared "<overflow>" route. This prevents unbounded memory growth when a layer accidentally emits high-cardinality labels (for example concrete request paths) while still keeping the already-admitted routes intact. Without this cap the collector records every distinct label it observes.

pub fn layer(&self) -> MetricsLayer<Self>

Creates a metrics layer backed by this collector.

The layer records request totals, errors, in-flight counts, and bounded duration histograms. It does not expose a scrape endpoint; use Self::routes or Self::routes_at for that.

pub fn routes(&self) -> Router

Creates a /metrics route for this collector.

pub fn routes_at(&self, path: &'static str) -> Router

Creates a metrics route at a custom path.

pub fn render(&self) -> String

Renders metrics in Prometheus text exposition format.

The output includes nidus_http_requests_total, nidus_http_request_duration_seconds_count, nidus_http_request_duration_seconds_sum, nidus_http_in_flight_requests, and nidus_http_errors_total.

Trait Implementations

impl Clone for PrometheusMetrics

fn clone(&self) -> PrometheusMetrics

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for PrometheusMetrics

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl Default for PrometheusMetrics

fn default() -> Self

Returns the “default value” for a type.

impl HttpMetricsHook for PrometheusMetrics

fn on_request(&self, method: &Method, route: Option<&str>)

Records that a request entered the service.

fn on_response( &self, method: &Method, route: Option<&str>, status: StatusCode, latency: Duration, )

Records that a response left the service.

fn on_error(&self, method: &Method, route: Option<&str>, latency: Duration)

Records that the inner service returned an error before producing a response.