uptime

uptime

English | 中文

Tiny uptime history middleware for Go net/http.

• Records heartbeat samples in the background
• Shows daily uptime bars for the last N days
• Uses SQLite for single-machine deployments, or PostgreSQL for shared multi-instance deployments
• Works without Prometheus, Grafana, or an external monitor
• Complements gofurry/monitor: monitor shows current runtime state, uptime shows historical availability

Install

go get github.com/gofurry/uptime

Quick Start

package main

import (
	"log"
	"net/http"

	"github.com/gofurry/uptime"
	"github.com/gofurry/uptime/store/sqlite"
)

func main() {
	up, err := uptime.New(uptime.Config{
		ServiceID:   "demo-api",
		ServiceName: "Demo API",
		Store: sqlite.New(sqlite.Config{
			Path: "./uptime.db",
		}),
	})
	if err != nil {
		log.Fatal(err)
	}
	defer up.Close()

	mux := http.NewServeMux()
	mux.Handle("/uptime", up.Handler())
	mux.Handle("/uptime/", up.Handler())
	mux.Handle("/", up.Middleware(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
		_, _ = w.Write([]byte("hello"))
	})))

	log.Fatal(http.ListenAndServe(":8080", mux))
}

Open:

• http://localhost:8080/uptime
• http://localhost:8080/uptime/api/status

Fiber

uptime is built on net/http. Fiber is based on fasthttp, so the safest integration is to create one uptime instance during startup and expose the uptime handler through Fiber's official adaptor.

Important: do not call uptime.New inside a Fiber handler. Each Uptime instance opens the store and starts a background heartbeat goroutine.

package main

import (
	"log"

	"github.com/gofiber/fiber/v2"
	"github.com/gofiber/fiber/v2/middleware/adaptor"
	"github.com/gofurry/uptime"
	"github.com/gofurry/uptime/store/sqlite"
)

func main() {
	up, err := uptime.New(uptime.Config{
		ServiceID:   "demo-api",
		ServiceName: "Demo API",
		Store: sqlite.New(sqlite.Config{
			Path: "./uptime.db",
		}),
	})
	if err != nil {
		log.Fatal(err)
	}
	defer up.Close()

	app := fiber.New()
	uptimeHandler := adaptor.HTTPHandler(up.Handler())
	app.All("/uptime", uptimeHandler)
	app.All("/uptime/*", uptimeHandler)

	app.Get("/", func(c *fiber.Ctx) error {
		return c.SendString("hello")
	})

	log.Fatal(app.Listen(":8080"))
}

Open http://localhost:8080/uptime.

The /uptime/* route is needed for /uptime/api/status, which is used by the dashboard refresh and by custom clients. uptime records service availability from its own heartbeat ticker, so you do not need to wrap every Fiber business route.

Demo Data

Generate a local uptime.db with multiple services, multiple instances, and 90 days of history:

go run ./cmd/uptime-demo-data -path ./uptime.db -reset=true
go run ./examples/basic

The example service writes its own heartbeat while the dashboard reads every service stored in the same database file.

Dashboard Only

uptime does not depend on business requests. Heartbeats are written by a background ticker, so this is valid:

mux.Handle("/uptime", up.Handler())
mux.Handle("/uptime/", up.Handler())

Middleware is a pass-through adapter. It is provided for normal net/http integration style and future request-aware features.

PostgreSQL

Use store/postgres when multiple service instances need to share one central uptime database:

package main

import (
	"log"
	"net/http"

	"github.com/gofurry/uptime"
	"github.com/gofurry/uptime/store/postgres"
)

func main() {
	up, err := uptime.New(uptime.Config{
		ServiceID:   "demo-api",
		ServiceName: "Demo API",
		Store: postgres.New(postgres.Config{
			Host:        "127.0.0.1",
			Port:        5432,
			Database:    "postgres",
			Username:    "postgres",
			Password:    "password",
			SSLMode:     "disable",
			Schema:      "public",
			TablePrefix: "uptime_",
		}),
	})
	if err != nil {
		log.Fatal(err)
	}
	defer up.Close()

	mux := http.NewServeMux()
	mux.Handle("/uptime", up.Handler())
	mux.Handle("/uptime/", up.Handler())
	log.Fatal(http.ListenAndServe(":8080", mux))
}

You can also pass postgres.Config{DSN: "postgres://..."}. The PostgreSQL store creates its schema, tables, and indexes automatically. The default table names are uptime_services, uptime_instances, uptime_samples, uptime_daily, and uptime_alert_state; use TablePrefix or Tables for custom names.

Alert Hook

Alerts are optional and disabled by default. Configure Alert.Hook to receive deduplicated service status transitions:

up, err := uptime.New(uptime.Config{
	ServiceID: "dashboard",
	Store:    store,
	Alert: uptime.AlertConfig{
		Hook: func(ctx context.Context, event uptime.AlertEvent) error {
			log.Printf("%s changed from %s to %s", event.ServiceID, event.PreviousStatus, event.CurrentStatus)
			return nil
		},
	},
})

Built-in SQLite and PostgreSQL stores persist alert state, so when several instances share one store only one instance claims a given status transition. The first observed state seeds the alert state and does not notify by default; set NotifyOnFirstDown if an already-down service should notify on first observation.

The hook is for delivery only. Send Slack, email, webhooks, or custom messages from user code.

External Probe

Core uptime records in-process heartbeats. External HTTP checks live in the optional probe package:

p, err := probe.New(probe.Config{
	ServiceID:      "homepage-probe",
	ServiceName:    "Homepage",
	URL:            "https://example.com/health",
	ExpectedStatus: []int{http.StatusOK},
	Interval:       30 * time.Second,
	Timeout:        5 * time.Second,
	Store:          store,
})
if err != nil {
	log.Fatal(err)
}
defer p.Close()

A successful probe writes a heartbeat for its synthetic service. A failed probe writes nothing, so missing slots naturally appear as downtime in the existing dashboard.

Snapshots and Custom UI

The built-in dashboard and JSON API use CachedSnapshot to avoid querying the store on every request. You can use the same API to build your own page or copy the status into Redis, Memcached, or another application cache:

snapshot, err := up.CachedSnapshot(r.Context())
if err != nil {
	http.Error(w, "uptime unavailable", http.StatusInternalServerError)
	return
}

_ = json.NewEncoder(w).Encode(snapshot)

Use Snapshot(ctx) when you explicitly need a fresh store read:

fresh, err := up.Snapshot(ctx)

Snapshot and CachedSnapshot return the same structure as /uptime/api/status.

Configuration

ServiceID and Store are required. The core package does not import SQLite automatically.

Defaults:

ServiceID should be a stable business identity such as api, worker, or gofurry-api. Do not generate a new service ID on each start, or history will be split across services.

How It Works

The process writes one up heartbeat every sample interval. Missing heartbeat slots are treated as downtime.

For a 3 second interval:

expected slots per normal day = 24 * 60 * 60 / 3 = 28800
uptime rate = distinct up slots / expected slots

For multiple instances of the same service, a slot is up when any instance writes a heartbeat for that slot.

When multiple services share the same store, the dashboard/API calculate current status, today's expected slots, missing-day expected slots, and estimated downtime with each service's stored sample interval.

Raw samples are kept for today and yesterday. Older samples are rolled up into daily snapshots and then cleaned up.

SQLite Notes

The SQLite store uses the pure-Go modernc.org/sqlite driver and configures:

PRAGMA journal_mode = WAL;
PRAGMA synchronous = NORMAL;
PRAGMA busy_timeout = 5000;

SQLite is intended for one machine. Multiple local processes may share the same database file, but network filesystems such as NFS are not recommended.

PostgreSQL Notes

The PostgreSQL store uses github.com/jackc/pgx/v5/stdlib through database/sql. It is intended for shared deployments where multiple processes or machines write to the same uptime store.

Configurable PostgreSQL fields:

Username is required when DSN is empty. Tables can override each table name individually.

Security

The dashboard is public by default. Put authentication, IP allowlists, or reverse proxy rules outside this package when the endpoint is exposed beyond trusted networks.

The middleware does not read request bodies, capture response bodies, log sensitive headers, or store request contexts.

Dashboard

The built-in page has no external assets. The frontend is maintained as embedded page.html, style.css, and app.js files under internal/ui, matching the structure used by gofurry/monitor.

It supports light and dark theme modes, plus English and simplified Chinese labels. The last selected theme and language are stored in browser local storage.

Daily bars use a custom hover card instead of the browser's native tooltip. The card is anchored to the active bar and centered below it.

Concurrency

Uptime instances and the SQLite store are safe for concurrent use after construction. Snapshot cache reads are protected by a mutex and return cloned payloads, so caller-side mutation does not affect the internal cache. Runtime heartbeat failures are recorded in memory and shown as degraded storage status; they do not affect business handlers.

Storage Extensibility

SQLite and PostgreSQL stores are provided. Additional databases can be added through the existing Store interface.

Related Documents

• 中文文档
• Contributing / 贡献指南
• Security Policy / 安全政策