Struct Metrics 

pub struct Metrics { /* private fields */ }

Prometheus metrics handle for the control plane.

Wraps a PrometheusHandle and the process start time used to compute lightshuttle_uptime_seconds at scrape time.

Lifecycle

Call Metrics::install once per process to register the global Prometheus recorder. Pass the resulting value (wrapped in an std::sync::Arc) to crate::ControlState::with_metrics so GET /metrics can render a live snapshot.

For tests and embedders that do not need metrics, build with Metrics::for_test, which does not touch the global recorder.

Tracked metrics

Metric name	Kind	Description
lightshuttle_restart_total	counter	Accepted restart requests
lightshuttle_lifecycle_event_duration_seconds	histogram	Seconds from start to healthy
lightshuttle_resources	gauge (per status label)	Managed resource count
lightshuttle_uptime_seconds	gauge	Process uptime

Implementations

impl Metrics

pub fn install() -> Self

Install the global Prometheus recorder and describe every metric. Call exactly once per process, before any metric is recorded.

Panics

Panics if a global recorder is already installed.

pub fn for_test() -> Self

Build a non-installing handle for tests.

The metrics! macros always target the globally installed recorder, which this constructor never sets. The returned handle therefore renders an empty snapshot regardless of any metric recorded elsewhere. Use Self::install plus super::ControlState::with_metrics to serve live metrics.

pub fn render(&self, status_counts: &[(&str, u64)]) -> String

Render the current metrics snapshot in Prometheus text format.

Before serialising, this method refreshes the two scrape-time gauges:

• lightshuttle_resources{status="<s>"} for each (status, count) pair in status_counts.
• lightshuttle_uptime_seconds derived from the process start time.

The returned string is suitable for serving directly as the body of GET /metrics with content type text/plain; version=0.0.4.