chore: harden production readiness gates and runbooks

backend/providers/http/engine.go

@@ -2,6 +2,7 @@ package http
 
 import (
 	"context"
+	"database/sql"
 	"errors"
 	"fmt"
 	"net"
@@ -9,9 +10,13 @@ import (
 	"strings"
 	"time"
 
+	"quyun/v2/app/errorx"
+	"quyun/v2/providers/storage"
+
 	logrus "github.com/sirupsen/logrus"
 	"go.ipao.vip/atom/container"
 	"go.ipao.vip/atom/opt"
+	"go.uber.org/dig"
 
 	"github.com/gofiber/fiber/v3"
 	"github.com/gofiber/fiber/v3/middleware/compress"
@@ -22,8 +27,6 @@ import (
 	"github.com/gofiber/fiber/v3/middleware/recover"
 	"github.com/gofiber/fiber/v3/middleware/requestid"
 	"github.com/samber/lo"
-
-	"quyun/v2/app/errorx"
 )
 
 func DefaultProvider() container.ProviderContainer {
@@ -36,8 +39,10 @@ func DefaultProvider() container.ProviderContainer {
 }
 
 type Service struct {
-	conf   *Config
-	Engine *fiber.App
+	conf        *Config
+	Engine      *fiber.App
+	healthCheck func(context.Context) error
+	readyCheck  func(context.Context) error
 }
 
 var errTLSCertKeyRequired = errors.New("tls cert and key must be set")
@@ -98,7 +103,11 @@ func Provide(opts ...opt.Option) error {
 		return err
 	}
 
-	return container.Container.Provide(func() (*Service, error) {
+	return container.Container.Provide(func(params struct {
+		dig.In
+		DB      *sql.DB          `optional:"true"`
+		Storage *storage.Storage `optional:"true"`
+	}) (*Service, error) {
 		engine := fiber.New(fiber.Config{
 			StrictRouting:      true,
 			CaseSensitive:      true,
@@ -198,8 +207,14 @@ func Provide(opts ...opt.Option) error {
 			}))
 		}
 
-		engine.Get("/healthz", func(c fiber.Ctx) error { return c.SendStatus(fiber.StatusNoContent) })
-		engine.Get("/readyz", func(c fiber.Ctx) error { return c.SendStatus(fiber.StatusNoContent) })
+		service := &Service{
+			Engine: engine,
+			conf:   &config,
+		}
+		service.healthCheck = service.buildHealthCheck()
+		service.readyCheck = service.buildReadyCheck(params.DB, params.Storage)
+		engine.Get("/healthz", service.handleHealthz)
+		engine.Get("/readyz", service.handleReadyz)
 
 		engine.Hooks().OnPostShutdown(func(err error) error {
 			if err != nil {
@@ -210,14 +225,72 @@ func Provide(opts ...opt.Option) error {
 			return nil
 		})
 
-		return &Service{
-			Engine: engine,
-			conf:   &config,
-		}, nil
+		return service, nil
 	}, o.DiOptions()...)
 }
 
 // buildCORSConfig converts provider Cors config into fiber cors.Config
+func (svc *Service) buildHealthCheck() func(context.Context) error {
+	return func(_ context.Context) error {
+		return nil
+	}
+}
+
+func (svc *Service) buildReadyCheck(db *sql.DB, store *storage.Storage) func(context.Context) error {
+	var dbPing func(context.Context) error
+	if db != nil {
+		dbPing = func(ctx context.Context) error {
+			pingCtx, cancel := context.WithTimeout(ctx, 1500*time.Millisecond)
+			defer cancel()
+
+			return db.PingContext(pingCtx)
+		}
+	}
+
+	return newReadyCheck(dbPing, store)
+}
+
+func newReadyCheck(dbPing func(context.Context) error, store *storage.Storage) func(context.Context) error {
+	return func(ctx context.Context) error {
+		if dbPing != nil {
+			if err := dbPing(ctx); err != nil {
+				return errorx.ErrServiceUnavailable.WithCause(err).WithMsg("database not ready")
+			}
+		}
+		if store != nil && store.Config != nil && strings.EqualFold(strings.TrimSpace(store.Config.Type), "s3") && store.Config.CheckOnBoot {
+			if strings.TrimSpace(store.Config.Endpoint) == "" || strings.TrimSpace(store.Config.Bucket) == "" {
+				return errorx.ErrServiceUnavailable.WithMsg("storage not ready")
+			}
+		}
+
+		return nil
+	}
+}
+
+func (svc *Service) handleHealthz(c fiber.Ctx) error {
+	ctx, cancel := context.WithTimeout(context.Background(), 2*time.Second)
+	defer cancel()
+	if svc.healthCheck != nil {
+		if err := svc.healthCheck(ctx); err != nil {
+			return errorx.SendError(c, err)
+		}
+	}
+
+	return c.SendStatus(fiber.StatusNoContent)
+}
+
+func (svc *Service) handleReadyz(c fiber.Ctx) error {
+	ctx, cancel := context.WithTimeout(context.Background(), 2*time.Second)
+	defer cancel()
+	if svc.readyCheck != nil {
+		if err := svc.readyCheck(ctx); err != nil {
+			return errorx.SendError(c, err)
+		}
+	}
+
+	return c.SendStatus(fiber.StatusNoContent)
+}
+
 func buildCORSConfig(c *Cors) *cors.Config {
 	if c == nil {
 		return nil