fix: 初验针对修改

internal/health/readiness.go

@@ -0,0 +1,115 @@
+package health
+
+import (
+	"context"
+	"fmt"
+	"net/http"
+	"runtime/debug"
+	"strings"
+	"time"
+
+	"git.apinb.com/ops/logs/internal/config"
+	"git.apinb.com/ops/logs/internal/impl"
+	"git.apinb.com/ops/logs/internal/ingest"
+	"github.com/gin-gonic/gin"
+	"gorm.io/gorm"
+)
+
+type CheckResult struct {
+	Name     string `json:"name"`
+	Required bool   `json:"required"`
+	Ready    bool   `json:"ready"`
+	Message  string `json:"message,omitempty"`
+}
+
+type Readiness struct {
+	Ready     bool                  `json:"ready"`
+	CheckedAt time.Time             `json:"checked_at"`
+	Checks    []CheckResult         `json:"checks,omitempty"`
+	Workers   []ingest.WorkerStatus `json:"workers,omitempty"`
+	Version   string                `json:"version"`
+}
+
+func Evaluate(ctx context.Context) Readiness {
+	result := Readiness{Ready: true, CheckedAt: time.Now().UTC(), Workers: ingest.WorkerStatuses(), Version: runtimeVersion()}
+	result.add("config", true, config.Validate())
+	result.add("database", true, pingDatabase(ctx, impl.DBService))
+	workerMap := make(map[string]ingest.WorkerStatus, len(result.Workers))
+	for _, worker := range result.Workers {
+		workerMap[worker.Name] = worker
+	}
+	for _, name := range requiredWorkers() {
+		worker, exists := workerMap[name]
+		if !exists || !worker.Running {
+			message := "worker 未启动"
+			if exists && worker.LastError != "" {
+				message = worker.LastError
+			}
+			result.add(name, true, fmt.Errorf("%s", message))
+			continue
+		}
+		result.add(name, true, nil)
+	}
+	return result
+}
+
+func runtimeVersion() string {
+	info, ok := debug.ReadBuildInfo()
+	if !ok || strings.TrimSpace(info.Main.Version) == "" {
+		return "unknown"
+	}
+	return info.Main.Version
+}
+
+func Ready(c *gin.Context) {
+	result := Evaluate(c.Request.Context())
+	status := http.StatusOK
+	if !result.Ready {
+		status = http.StatusServiceUnavailable
+	}
+	c.JSON(status, gin.H{"ready": result.Ready})
+}
+
+func Status(c *gin.Context) {
+	result := Evaluate(c.Request.Context())
+	status := http.StatusOK
+	if !result.Ready {
+		status = http.StatusServiceUnavailable
+	}
+	c.JSON(status, result)
+}
+
+func requiredWorkers() []string {
+	workers := []string{"rule_refresher", "alert_dispatcher"}
+	if strings.TrimSpace(config.Spec.Ingest.SyslogListenAddr) != "" {
+		workers = append(workers, "syslog_udp")
+	}
+	if strings.TrimSpace(config.Spec.Ingest.TrapListenAddr) != "" {
+		workers = append(workers, "trap_udp")
+	}
+	return workers
+}
+
+func (r *Readiness) add(name string, required bool, err error) {
+	check := CheckResult{Name: name, Required: required, Ready: err == nil}
+	if err != nil {
+		check.Message = strings.TrimSpace(err.Error())
+		if required {
+			r.Ready = false
+		}
+	}
+	r.Checks = append(r.Checks, check)
+}
+
+func pingDatabase(ctx context.Context, db *gorm.DB) error {
+	if db == nil {
+		return fmt.Errorf("数据库未初始化")
+	}
+	sqlDB, err := db.DB()
+	if err != nil {
+		return err
+	}
+	checkCtx, cancel := context.WithTimeout(ctx, 2*time.Second)
+	defer cancel()
+	return sqlDB.PingContext(checkCtx)
+}