feat: add bounded process metrics and private monitoring endpoint

internal/observability/runtime.go

@@ -0,0 +1,64 @@
+package observability
+
+import (
+	"context"
+	"database/sql"
+	"errors"
+	"runtime"
+	"time"
+)
+
+// initializeProcess is called once by the registry constructor.
+func (r *Registry) initializeProcess() {
+	if r == nil {
+		return
+	}
+	r.SetGauge("orbit_process_start_time_seconds", float64(r.started.Unix()), nil)
+	r.SetGauge("orbit_build_info", 1, Labels{"go_version": runtime.Version()})
+	r.RecordProcess()
+	for _, component := range components {
+		labels := Labels{"component": component}
+		r.SetGauge("orbit_runtime_batch_last_success_timestamp_seconds", 0, labels)
+		for _, outcome := range []string{"success", "error", "cancelled"} {
+			r.AddCounter("orbit_runtime_batch_total", 0, Labels{"component": component, "outcome": outcome})
+		}
+	}
+	for _, route := range []string{"worker_connect", "event_connect"} {
+		r.SetGauge("orbit_websocket_connections", 0, Labels{"route": route})
+		r.AddCounter("orbit_websocket_connections_total", 0, Labels{"route": route})
+	}
+}
+func (r *Registry) RecordProcess() {
+	if r == nil {
+		return
+	}
+	r.SetGauge("orbit_process_uptime_seconds", time.Since(r.started).Seconds(), nil)
+	r.SetGauge("orbit_go_goroutines", float64(runtime.NumGoroutine()), nil)
+}
+func (r *Registry) RecordBatch(component string, started time.Time, err error) {
+	if r == nil {
+		return
+	}
+	outcome := "success"
+	if err != nil {
+		outcome = "error"
+		if errors.Is(err, context.Canceled) {
+			outcome = "cancelled"
+		}
+	}
+	r.AddCounter("orbit_runtime_batch_total", 1, Labels{"component": component, "outcome": outcome})
+	r.ObserveHistogram("orbit_runtime_batch_duration_seconds", time.Since(started).Seconds(), Labels{"component": component})
+	if err == nil {
+		r.SetGauge("orbit_runtime_batch_last_success_timestamp_seconds", float64(time.Now().Unix()), Labels{"component": component})
+	}
+}
+
+// RecordDatabase reads the local connection-pool snapshot; it issues no SQL.
+func (r *Registry) RecordDatabase(s sql.DBStats) {
+	r.SetGauge("orbit_db_connections_open", float64(s.OpenConnections), nil)
+	r.SetGauge("orbit_db_connections_in_use", float64(s.InUse), nil)
+	r.SetGauge("orbit_db_connections_idle", float64(s.Idle), nil)
+	r.SetGauge("orbit_db_connections_max_open", float64(s.MaxOpenConnections), nil)
+	r.SetCounter("orbit_db_wait_count_total", float64(s.WaitCount), nil)
+	r.SetCounter("orbit_db_wait_duration_seconds_total", s.WaitDuration.Seconds(), nil)
+}