feat: 建立可观测性、容量与恢复验证基线

server-go/internal/service/ai_client.go

@@ -77,6 +77,7 @@ func NewAIClient(baseURL string) *AIClient {
 
 // Detect 上传图片到 ai-service /detect，返回检测结果（无状态接口，天然幂等）
 func (c *AIClient) Detect(ctx context.Context, imageBytes []byte, filename string) (*AIDetectResponse, error) {
+	start := time.Now()
 	var body bytes.Buffer
 	writer := multipart.NewWriter(&body)
 	part, err := writer.CreateFormFile("file", filename)
@@ -97,7 +98,9 @@ func (c *AIClient) Detect(ctx context.Context, imageBytes []byte, filename strin
 	req.Header.Set("Content-Type", writer.FormDataContentType())
 
 	resp, err := c.client.Do(req)
+	latency := time.Since(start)
 	if err != nil {
+		RecordDependency(DependencyAI, err, latency)
 		return nil, err
 	}
 	defer resp.Body.Close()
@@ -107,12 +110,16 @@ func (c *AIClient) Detect(ctx context.Context, imageBytes []byte, filename strin
 		return nil, err
 	}
 	if resp.StatusCode != http.StatusOK {
-		return nil, fmt.Errorf("ai-service /detect 返回 %d: %s", resp.StatusCode, string(data))
+		err := fmt.Errorf("ai-service /detect 返回 %d: %s", resp.StatusCode, string(data))
+		RecordDependency(DependencyAI, err, latency)
+		return nil, err
 	}
 
 	var out AIDetectResponse
 	if err := json.Unmarshal(data, &out); err != nil {
+		RecordDependency(DependencyAI, err, latency)
 		return nil, err
 	}
+	RecordDependency(DependencyAI, nil, latency)
 	return &out, nil
 }

server-go/internal/service/dependency_metrics.go

@@ -0,0 +1,86 @@
+package service
+
+import (
+	"sort"
+	"sync"
+	"time"
+)
+
+// Dependency 依赖名称。
+type Dependency string
+
+const (
+	DependencyPostgres Dependency = "postgresql"
+	DependencyRedis    Dependency = "redis"
+	DependencyMQTT     Dependency = "mqtt"
+	DependencyIoTDB    Dependency = "iotdb"
+	DependencyS3       Dependency = "s3"
+	DependencyAI       Dependency = "ai"
+	DependencyWVP      Dependency = "wvp"
+	DependencyWechat   Dependency = "wechat"
+	DependencyWeather  Dependency = "weather"
+)
+
+// DependencyMetric 依赖调用指标。
+type DependencyMetric struct {
+	Requests      int64      `json:"requests"`
+	Failures      int64      `json:"failures"`
+	LastLatencyMs float64    `json:"lastLatencyMs"`
+	P50LatencyMs  float64    `json:"p50LatencyMs"`
+	P95LatencyMs  float64    `json:"p95LatencyMs"`
+	LastSuccessAt *time.Time `json:"lastSuccessAt,omitempty"`
+	LastFailureAt *time.Time `json:"lastFailureAt,omitempty"`
+	LastError     string     `json:"lastError,omitempty"`
+}
+
+var dependencyMetrics sync.Map
+
+// RecordDependency 记录一次依赖调用。
+func RecordDependency(dep Dependency, err error, latency time.Duration) {
+	now := time.Now()
+	value, _ := dependencyMetrics.LoadOrStore(dep, &DependencyMetric{})
+	metric := value.(*DependencyMetric)
+	metric.Requests++
+	metric.LastLatencyMs = float64(latency.Microseconds()) / 1000
+	if err != nil {
+		metric.Failures++
+		metric.LastFailureAt = &now
+		metric.LastError = err.Error()
+	} else {
+		metric.LastSuccessAt = &now
+		metric.LastError = ""
+	}
+}
+
+// DependencySnapshot 返回当前依赖指标快照。
+func DependencySnapshot() map[Dependency]DependencyMetric {
+	result := map[Dependency]DependencyMetric{}
+	dependencyMetrics.Range(func(key, value interface{}) bool {
+		dep := key.(Dependency)
+		metric := *value.(*DependencyMetric)
+		result[dep] = metric
+		return true
+	})
+	return result
+}
+
+// RecordDependencyHistogram 预留直方图能力；当前仅更新最近一次指标。
+func RecordDependencyHistogram(dep Dependency, err error, latency time.Duration, samples []float64) {
+	RecordDependency(dep, err, latency)
+	if len(samples) == 0 {
+		return
+	}
+	value, _ := dependencyMetrics.LoadOrStore(dep, &DependencyMetric{})
+	metric := value.(*DependencyMetric)
+	sort.Float64s(samples)
+	metric.P50LatencyMs = percentile(samples, 0.5)
+	metric.P95LatencyMs = percentile(samples, 0.95)
+}
+
+func percentile(sorted []float64, p float64) float64 {
+	if len(sorted) == 0 {
+		return 0
+	}
+	idx := int(float64(len(sorted)-1) * p)
+	return sorted[idx]
+}