feat(ops): protect metrics, rate-limit auth, agent secret timing, e2e smoke

Bearer для /metrics (EVOBGP_METRICS_TOKEN); rate limit /v1/auth/config; constant-time agent secret; OTel stub; Playwright smoke; HTTP_PROXY note; checklist обновлён.

Co-authored-by: Cursor <[email protected]>

apps/web/e2e/smoke.spec.ts

@@ -0,0 +1,32 @@
+import { test, expect } from '@playwright/test'
+
+/**
+ * Smoke against local API (+ optional UI). Profile: compose microvps-full.
+ *
+ * Env:
+ * - EVOBGP_E2E_API_URL (default http://127.0.0.1:8080)
+ * - EVOBGP_E2E_BASE_URL (default http://127.0.0.1:5173)
+ * - EVOBGP_E2E_TOKEN (default dev)
+ */
+const apiBase = process.env.EVOBGP_E2E_API_URL ?? 'http://127.0.0.1:8080'
+const uiBase = process.env.EVOBGP_E2E_BASE_URL ?? 'http://127.0.0.1:5173'
+const token = process.env.EVOBGP_E2E_TOKEN ?? 'dev'
+
+test.describe('EvoBGP smoke', () => {
+  test('API health and modules list', async ({ request }) => {
+    const health = await request.get(`${apiBase}/v1/health`)
+    expect(health.ok()).toBeTruthy()
+
+    const mods = await request.get(`${apiBase}/v1/modules?limit=10`, {
+      headers: { Authorization: `Bearer ${token}` },
+    })
+    expect([200, 401]).toContain(mods.status())
+  })
+
+  test('UI loads when available', async ({ page }) => {
+    test.skip(!process.env.EVOBGP_E2E_UI, 'set EVOBGP_E2E_UI=1 to enable UI smoke')
+    await page.goto(uiBase)
+    await page.waitForLoadState('domcontentloaded')
+    await expect(page.locator('body')).toBeVisible()
+  })
+})

apps/web/package.json

@@ -10,6 +10,7 @@
     "typecheck": "tsr generate && tsc --noEmit",
     "lint": "eslint .",
     "test": "vitest run",
+    "test:e2e": "playwright test",
     "openapi:gen": "openapi-typescript ../../docs/openapi.yaml -o src/types/api.gen.ts",
     "openapi:check": "openapi-typescript ../../docs/openapi.yaml -o src/types/api.gen.check.ts && diff -q src/types/api.gen.ts src/types/api.gen.check.ts && rm -f src/types/api.gen.check.ts"
   },
@@ -43,6 +44,7 @@
   },
   "devDependencies": {
     "@eslint/js": "^9.0.0",
+    "@playwright/test": "^1.62.1",
     "@tailwindcss/vite": "^4.1.0",
     "@tanstack/router-cli": "^1.130.0",
     "@tanstack/router-plugin": "^1.130.0",

apps/web/playwright.config.ts

@@ -0,0 +1,14 @@
+import { defineConfig, devices } from '@playwright/test'
+
+export default defineConfig({
+  testDir: './e2e',
+  timeout: 60_000,
+  fullyParallel: false,
+  forbidOnly: !!process.env.CI,
+  retries: process.env.CI ? 1 : 0,
+  use: {
+    baseURL: process.env.EVOBGP_E2E_BASE_URL ?? 'http://127.0.0.1:5173',
+    trace: 'on-first-retry',
+  },
+  projects: [{ name: 'chromium', use: { ...devices['Desktop Chrome'] } }],
+})

apps/web/vite.config.ts

@@ -35,5 +35,6 @@ export default defineConfig({
   test: {
     environment: 'happy-dom',
     globals: true,
+    exclude: ['**/node_modules/**', '**/e2e/**', '**/dist/**'],
   },
 })

cmd/evobgp-api/main.go

@@ -52,6 +52,8 @@ func main() {
 	observability.SetBuildInfo(version.Version, version.GitSHA)
 	ctx, stop := signal.NotifyContext(context.Background(), os.Interrupt, syscall.SIGTERM)
 	defer stop()
+	otelShutdown := observability.StartOTelIfEnabled(ctx)
+	defer func() { _ = otelShutdown(context.Background()) }()
 
 	srv.StartBackground(ctx)
 	startBirdMetricsPoller(ctx)

docs/production-checklist.md

@@ -17,8 +17,12 @@
 
 - `EVOBGP_JOB_MAX_CONCURRENT=16`, `EVOBGP_DB_MAX_CONNS=25`, `EVOBGP_COLLECT_CONCURRENCY=16` при росте tenants.
 - `EVOBGP_NODE_DISPATCH_INSECURE_TLS=0` — только валидный TLS к agent.
-- Ограничить `/metrics` сетевой политикой или reverse proxy.
+- Ограничить `/metrics`: `EVOBGP_METRICS_TOKEN` (Bearer) и/или сетевая политика / reverse proxy.
-- Профиль `evobgp-all` или HA API + персистентная `job_audit` (PostgreSQL).
+- `EVOBGP_AUTH_RATE_LIMIT` (default 60/min per IP) на `GET /v1/auth/config`.
+- `HTTP_PROXY` / `HTTPS_PROXY` / `NO_PROXY` для CDN/DoH в censored сетях.
+- Опционально `EVOBGP_OTEL_ENDPOINT` (stub opt-in для будущего OTel export).
+- Профиль `evobgp-all` или HA API + персистентная `job_audit` (PostgreSQL SKIP LOCKED reclaim).
+- E2E smoke: `pnpm --filter @evobgp/web run test:e2e` против API (`EVOBGP_E2E_API_URL`) / UI (`EVOBGP_E2E_UI=1`).
 - Мониторинг drift: `evobgp-deploy`, `last_applied_revision_id` vs published.
 
 ## Не использовать в prod

internal/agentserver/server.go

@@ -2,6 +2,7 @@ package agentserver
 
 import (
 	"context"
+	"crypto/subtle"
 	"encoding/json"
 	"fmt"
 	"log"
@@ -157,7 +158,8 @@ func (s *Server) authorize(r *http.Request) bool {
 	if !strings.HasPrefix(h, prefix) {
 		return false
 	}
-	return strings.TrimSpace(h[len(prefix):]) == secret
+	got := strings.TrimSpace(h[len(prefix):])
+	return subtle.ConstantTimeCompare([]byte(got), []byte(secret)) == 1
 }
 
 func writeJSON(w http.ResponseWriter, status int, v any) {

internal/httpapi/ratelimit.go

@@ -0,0 +1,68 @@
+package httpapi
+
+import (
+	"net/http"
+	"os"
+	"strconv"
+	"strings"
+	"sync"
+	"time"
+)
+
+// authRateLimiter is a simple per-IP token bucket for public auth-ish endpoints.
+type authRateLimiter struct {
+	mu     sync.Mutex
+	hits   map[string][]time.Time
+	limit  int
+	window time.Duration
+}
+
+func newAuthRateLimiter() *authRateLimiter {
+	limit := 60
+	if n, err := strconv.Atoi(strings.TrimSpace(os.Getenv("EVOBGP_AUTH_RATE_LIMIT"))); err == nil && n > 0 {
+		limit = n
+	}
+	return &authRateLimiter{
+		hits:   make(map[string][]time.Time),
+		limit:  limit,
+		window: time.Minute,
+	}
+}
+
+func (l *authRateLimiter) allow(ip string) bool {
+	if l == nil {
+		return true
+	}
+	now := time.Now()
+	l.mu.Lock()
+	defer l.mu.Unlock()
+	cut := now.Add(-l.window)
+	arr := l.hits[ip]
+	kept := arr[:0]
+	for _, t := range arr {
+		if t.After(cut) {
+			kept = append(kept, t)
+		}
+	}
+	if len(kept) >= l.limit {
+		l.hits[ip] = kept
+		return false
+	}
+	kept = append(kept, now)
+	l.hits[ip] = kept
+	return true
+}
+
+func (s *Server) withAuthRateLimit(next http.HandlerFunc) http.HandlerFunc {
+	if s.authLimiter == nil {
+		s.authLimiter = newAuthRateLimiter()
+	}
+	return func(w http.ResponseWriter, r *http.Request) {
+		ip := clientIP(r)
+		if !s.authLimiter.allow(ip) {
+			writeProblem(w, http.StatusTooManyRequests, "Too Many Requests", "auth rate limit exceeded")
+			return
+		}
+		next(w, r)
+	}
+}

internal/httpapi/routes.go

@@ -30,12 +30,12 @@ func (s *Server) Handler() http.Handler {
 	s.registerV1(v1)
 	wrappedV1 := http.StripPrefix("/v1", v1)
 
-	s.mux.Handle("GET /metrics", observability.MetricsHandler())
+	s.mux.Handle("GET /metrics", observability.ProtectMetrics(observability.MetricsHandler()))
 	s.mux.HandleFunc("GET /version", s.handleVersion)
 	s.mux.HandleFunc("GET /v1/health", s.handleHealth)
 	s.mux.HandleFunc("GET /v1/ready", s.handleReady)
 	s.mux.HandleFunc("GET /v1/version", s.handleVersion)
-	s.mux.HandleFunc("GET /v1/auth/config", s.handleAuthConfigPublic)
+	s.mux.HandleFunc("GET /v1/auth/config", s.withAuthRateLimit(s.handleAuthConfigPublic))
 	// Firewall subsystem moved to the standalone EvoFirewall service; see docs/firewall.md.
 	// Registered on the public mux so it wins over the "/v1/" subtree below regardless of auth.
 	s.mux.HandleFunc("/v1/firewall/", s.handleFirewallGone)

internal/httpapi/server.go

@@ -29,6 +29,7 @@ type Server struct {
 	maintConfig             *maintenance.ConfigProvider
 	maintStats              *maintenance.DBStatsProvider
 	jobs                    *jobs.Registry
+	authLimiter             *authRateLimiter
 	bundlePriv              ed25519.PrivateKey
 	keyResolver             *apiKeyResolver
 	firewallResolver        *firewallTokenResolver

internal/httpclient/httpclient.go

@@ -14,6 +14,7 @@ import (
 const DefaultTimeout = 45 * time.Second
 
 // New returns an HTTP client with timeout and tuned idle connection pooling.
+// Standard proxy env (HTTP_PROXY / HTTPS_PROXY / NO_PROXY) is honored via the cloned DefaultTransport.
 func New(timeout time.Duration) *http.Client {
 	if timeout <= 0 {
 		timeout = DefaultTimeout

internal/observability/metrics_auth.go

@@ -0,0 +1,32 @@
+package observability
+
+import (
+	"crypto/subtle"
+	"net/http"
+	"os"
+	"strings"
+)
+
+// ProtectMetrics wraps the Prometheus handler. When EVOBGP_METRICS_TOKEN is set,
+// scrapes must send Authorization: Bearer <token> (constant-time compare).
+// Empty token keeps /metrics open (dev / private network).
+func ProtectMetrics(next http.Handler) http.Handler {
+	token := strings.TrimSpace(os.Getenv("EVOBGP_METRICS_TOKEN"))
+	if token == "" {
+		return next
+	}
+	want := []byte(token)
+	return http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		h := r.Header.Get("Authorization")
+		const p = "Bearer "
+		got := ""
+		if strings.HasPrefix(h, p) {
+			got = strings.TrimSpace(h[len(p):])
+		}
+		if subtle.ConstantTimeCompare([]byte(got), want) != 1 {
+			http.Error(w, "Unauthorized", http.StatusUnauthorized)
+			return
+		}
+		next.ServeHTTP(w, r)
+	})
+}

internal/observability/otel.go

@@ -0,0 +1,21 @@
+package observability
+
+import (
+	"context"
+	"os"
+	"strings"
+
+	"evobgp/internal/logging"
+)
+
+// StartOTelIfEnabled is an opt-in stub for OpenTelemetry (EVOBGP_OTEL_ENDPOINT).
+// Full exporter wiring lands when the ops stack provides a collector; this logs intent only.
+func StartOTelIfEnabled(ctx context.Context) (shutdown func(context.Context) error) {
+	ep := strings.TrimSpace(os.Getenv("EVOBGP_OTEL_ENDPOINT"))
+	if ep == "" {
+		return func(context.Context) error { return nil }
+	}
+	logging.Default().Info("otel opt-in enabled (stub; export not wired yet)", "endpoint", ep)
+	_ = ctx
+	return func(context.Context) error { return nil }
+}

pnpm-lock.yaml

@@ -117,6 +117,9 @@ importers:
       '@eslint/js':
         specifier: ^9.0.0
         version: 9.39.4
+      '@playwright/test':
+        specifier: ^1.62.1
+        version: 1.62.1
       '@tailwindcss/vite':
         specifier: ^4.1.0
         version: 4.3.2([email protected](@types/[email protected])([email protected])([email protected]))
@@ -770,6 +773,11 @@ packages:
   '@octokit/[email protected]':
     resolution: {integrity: sha512-sKq+9r1Mm4efXW1FCk7hFSeJo4QKreL/tTbR0rz/qx/r1Oa2VV83LTA/H/MuCOX7uCIJmQVRKBcbmWoySjAnSg==}
 
+  '@playwright/[email protected]':
+    resolution: {integrity: sha512-DTcUc8qii+cpHvtOwggMtBRMjKZHXYWdw8syRYu2vtzuq4Wxphqq4NfCs5Zt44L6mA8rfDfj+PHnxFc/FeK6mQ==}
+    engines: {node: '>=20'}
+    hasBin: true
+
   '@pnpm/[email protected]':
     resolution: {integrity: sha512-htyl8TWnKL7K/ESFa1oW2UB5lVDxuF5DpM7tBi6Hu2LNL3mWkIzNLG6N4zoCUP1lCKNxWy/3iu8mS8MvToGd6w==}
     engines: {node: '>=12.22.0'}
@@ -2239,6 +2247,11 @@ packages:
     resolution: {integrity: sha512-w8ZNZr2mKIc7qeNaQ9AVPT1+iFaI+Avd4xudVOvdDJ8VytREi1Ft5Ih7hd9jjehod8vAM5GMsfQ/TpPf4EyoEA==}
     engines: {node: '>=14.14'}
 
+  [email protected]:
+    resolution: {integrity: sha512-xiqMQR4xAeHTuB9uWm+fFRcIOgKBMiOBP+eXiyT7jsgVCq1bkVygt00oASowB7EdtpOHaaPgKt812P9ab+DDKA==}
+    engines: {node: ^8.16.0 || ^10.6.0 || >=11.0.0}
+    os: [darwin]
+
   [email protected]:
     resolution: {integrity: sha512-5xoDfX+fL7faATnagmWPpbFtwh/R77WmMMqqHGS65C3vvB0YHrgF+B1YmZ3441tMj5n63k0212XNoJwzlhffQw==}
     engines: {node: ^8.16.0 || ^10.6.0 || >=11.0.0}
@@ -3124,6 +3137,16 @@ packages:
     resolution: {integrity: sha512-C+VUP+8jis7EsQZIhDYmS5qlNtjv2yP4SNtjXK9AP1ZcTRlnSfuumaTnRfYZnYgUUYVIKqL0fRvmUGDV2fmp6g==}
     engines: {node: '>=4'}
 
+  [email protected]:
+    resolution: {integrity: sha512-wPYSwEBJY9GHraISXqyqtx0na0LpO3XEX7jNDhntbex7tzUS7kLnZsOlFruFJB4Hi/rhDMjXGqHewDZ68nYZVw==}
+    engines: {node: '>=20'}
+    hasBin: true
+
+  [email protected]:
+    resolution: {integrity: sha512-0M+L3LAD8/nm554LOla9Ayx0j0tmFZ0FBcoQ7F1VuVHpM/XpiC8RcDzBQB8W5+hA8L22THxELzeF+2WcUzvcLg==}
+    engines: {node: '>=20'}
+    hasBin: true
+
   [email protected]:
     resolution: {integrity: sha512-Nc3IT5yHzflTfbjgqWcCPpo7DaKy4FnpB0l/zCAW0Tc7jxAiuqSxHasntB3D7887LSrA93kDJ9IXovxJYxyLCA==}
     engines: {node: '>=4'}
@@ -4492,6 +4515,10 @@ snapshots:
     dependencies:
       '@octokit/openapi-types': 27.0.0
 
+  '@playwright/[email protected]':
+    dependencies:
+      playwright: 1.62.1
+
   '@pnpm/[email protected]': {}
 
   '@pnpm/[email protected]':
@@ -5996,6 +6023,9 @@ snapshots:
       jsonfile: 6.2.1
       universalify: 2.0.1
 
+  [email protected]:
+    optional: true
+
   [email protected]:
     optional: true
 
@@ -6688,6 +6718,14 @@ snapshots:
       find-up: 2.1.0
       load-json-file: 4.0.0
 
+  [email protected]: {}
+
+  [email protected]:
+    dependencies:
+      playwright-core: 1.62.1
+    optionalDependencies:
+      fsevents: 2.3.2
+
   [email protected]: {}
 
   [email protected]: