From bfda29193a4bf15246a0703a4f8638a1ca93d028 Mon Sep 17 00:00:00 2001
From: Rain Seven <128443127+RAiNY7Study@users.noreply.github.com>
Date: Sun, 16 Aug 2026 21:28:01 +0800
Subject: [PATCH] feat: add operational health and metrics endpoints (#37)

Co-authored-by: Meng Meng <227010654+MengMengCode@users.noreply.github.com>
---
 internal/server/health_endpoints.go      | 52 ++++++++++++++++++
 internal/server/health_endpoints_test.go | 67 ++++++++++++++++++++++++
 internal/server/server.go                |  3 ++
 3 files changed, 122 insertions(+)
 create mode 100644 internal/server/health_endpoints.go
 create mode 100644 internal/server/health_endpoints_test.go

diff --git a/internal/server/health_endpoints.go b/internal/server/health_endpoints.go
new file mode 100644
index 0000000..f357b53
--- /dev/null
+++ b/internal/server/health_endpoints.go
@@ -0,0 +1,52 @@
+package server
+
+import (
+	"context"
+	"fmt"
+	"net/http"
+	"runtime"
+	"time"
+)
+
+func (s *Server) handleLiveness(w http.ResponseWriter, r *http.Request) {
+	if !requireMethod(w, r, http.MethodGet) {
+		return
+	}
+	w.Header().Set("Cache-Control", "no-store")
+	writeJSON(w, http.StatusOK, map[string]any{"status": "ok"})
+}
+
+func (s *Server) handleReadiness(w http.ResponseWriter, r *http.Request) {
+	if !requireMethod(w, r, http.MethodGet) {
+		return
+	}
+	ctx, cancel := context.WithTimeout(r.Context(), 2*time.Second)
+	defer cancel()
+	if err := s.store.Ready(ctx); err != nil {
+		writeJSON(w, http.StatusServiceUnavailable, map[string]any{"status": "not_ready"})
+		return
+	}
+	w.Header().Set("Cache-Control", "no-store")
+	writeJSON(w, http.StatusOK, map[string]any{"status": "ready"})
+}
+
+// handleMetrics exposes only process-level, non-identifying Prometheus data.
+// Device IDs, SIM identities, phone numbers and proxy information never enter
+// this unauthenticated endpoint.
+func (s *Server) handleMetrics(w http.ResponseWriter, r *http.Request) {
+	if !requireMethod(w, r, http.MethodGet) {
+		return
+	}
+	ready := 0
+	ctx, cancel := context.WithTimeout(r.Context(), time.Second)
+	if s.store.Ready(ctx) == nil {
+		ready = 1
+	}
+	cancel()
+	w.Header().Set("Content-Type", "text/plain; version=0.0.4; charset=utf-8")
+	w.Header().Set("Cache-Control", "no-store")
+	fmt.Fprint(w, "# HELP vocat_up Whether the process is running.\n# TYPE vocat_up gauge\nvocat_up 1\n")
+	fmt.Fprintf(w, "# HELP vocat_ready Whether the database is ready.\n# TYPE vocat_ready gauge\nvocat_ready %d\n", ready)
+	fmt.Fprintf(w, "# HELP vocat_uptime_seconds Process uptime.\n# TYPE vocat_uptime_seconds counter\nvocat_uptime_seconds %.0f\n", time.Since(s.startedAt).Seconds())
+	fmt.Fprintf(w, "# HELP vocat_go_goroutines Current Go goroutines.\n# TYPE vocat_go_goroutines gauge\nvocat_go_goroutines %d\n", runtime.NumGoroutine())
+}
diff --git a/internal/server/health_endpoints_test.go b/internal/server/health_endpoints_test.go
new file mode 100644
index 0000000..24fe1bd
--- /dev/null
+++ b/internal/server/health_endpoints_test.go
@@ -0,0 +1,67 @@
+package server
+
+import (
+	"io"
+	"net/http"
+	"strings"
+	"testing"
+)
+
+func TestOperationalHealthEndpointsAreAnonymousAndNonIdentifying(t *testing.T) {
+	app := newTestApplication(t)
+
+	tests := []struct {
+		path        string
+		contentType string
+		contains    string
+	}{
+		{path: "/healthz", contentType: "application/json", contains: `"status":"ok"`},
+		{path: "/readyz", contentType: "application/json", contains: `"status":"ready"`},
+		{path: "/metrics", contentType: "text/plain", contains: "vocat_ready 1"},
+	}
+	for _, test := range tests {
+		t.Run(test.path, func(t *testing.T) {
+			response, err := app.client.Get(app.server.URL + test.path)
+			if err != nil {
+				t.Fatal(err)
+			}
+			defer response.Body.Close()
+			body, err := io.ReadAll(response.Body)
+			if err != nil {
+				t.Fatal(err)
+			}
+			if response.StatusCode != http.StatusOK {
+				t.Fatalf("status = %d, body = %s", response.StatusCode, body)
+			}
+			if !strings.Contains(response.Header.Get("Content-Type"), test.contentType) {
+				t.Fatalf("Content-Type = %q", response.Header.Get("Content-Type"))
+			}
+			if !strings.Contains(string(body), test.contains) {
+				t.Fatalf("body = %q, want %q", body, test.contains)
+			}
+			for _, forbidden := range []string{"imsi", "iccid", "msisdn", "proxy", "device_id"} {
+				if strings.Contains(strings.ToLower(string(body)), forbidden) {
+					t.Fatalf("body exposes forbidden label %q: %s", forbidden, body)
+				}
+			}
+		})
+	}
+}
+
+func TestOperationalHealthEndpointsRejectPOST(t *testing.T) {
+	app := newTestApplication(t)
+	for _, path := range []string{"/healthz", "/readyz", "/metrics"} {
+		request, err := http.NewRequest(http.MethodPost, app.server.URL+path, nil)
+		if err != nil {
+			t.Fatal(err)
+		}
+		response, err := app.client.Do(request)
+		if err != nil {
+			t.Fatal(err)
+		}
+		response.Body.Close()
+		if response.StatusCode != http.StatusMethodNotAllowed {
+			t.Fatalf("%s status = %d, want %d", path, response.StatusCode, http.StatusMethodNotAllowed)
+		}
+	}
+}
diff --git a/internal/server/server.go b/internal/server/server.go
index b4cbcf3..2156630 100644
--- a/internal/server/server.go
+++ b/internal/server/server.go
@@ -144,6 +144,9 @@ func New(options Options) (*Server, error) {
 	server.loadUILanguage(context.Background())
 
 	mux := http.NewServeMux()
+	mux.HandleFunc("/healthz", server.handleLiveness)
+	mux.HandleFunc("/readyz", server.handleReadiness)
+	mux.HandleFunc("/metrics", server.handleMetrics)
 	mux.HandleFunc("/api/health", server.handleHealth)
 	mux.HandleFunc("/api/auth/login", server.handleLogin)
 	mux.HandleFunc("/api/auth/session", server.handleSession)