Merge pull request #2570 from wucm667/fix/ops-sla-exclude-ip-denied

fix(ops): 用户 IP 限制导致的 ACCESS_DENIED 不计入 SLA 错误

backend/internal/handler/ops_error_logger.go

@@ -1220,15 +1220,19 @@ func classifyOpsIsRetryable(errType string, statusCode int) bool {
 func classifyOpsErrorLog(c *gin.Context, errType, message, code string, status int) (phase string, isBusinessLimited bool, errorOwner string, errorSource string) {
 	phase = classifyOpsPhase(errType, message, code)
 	routingCapacityLimited := isOpsRoutingCapacityLimited(c)
+	clientBusinessLimited := service.HasOpsClientBusinessLimited(c)
 	upstreamError := hasOpsUpstreamErrorContext(c)
 	if upstreamError && !routingCapacityLimited {
 		phase = "upstream"
 	}
+	if clientBusinessLimited && !upstreamError && !routingCapacityLimited {
+		phase = "auth"
+	}
 	if routingCapacityLimited {
 		phase = "routing"
 	}
 	localClientAuthError := !upstreamError && phase == "auth" && isOpsClientAuthError(code, strings.ToLower(message))
-	isBusinessLimited = routingCapacityLimited || classifyOpsIsBusinessLimited(errType, phase, code, status, message, localClientAuthError)
+	isBusinessLimited = routingCapacityLimited || clientBusinessLimited || classifyOpsIsBusinessLimited(errType, phase, code, status, message, localClientAuthError)
 	errorOwner = classifyOpsErrorOwner(phase, message)
 	errorSource = classifyOpsErrorSource(phase, message)
 	return phase, isBusinessLimited, errorOwner, errorSource

backend/internal/handler/ops_error_logger_test.go

@@ -370,6 +370,37 @@ func TestClassifyOpsAuthClientErrorsExcludedFromSLA(t *testing.T) {
 	}
 }
 
+func TestClassifyOpsIPRestrictionAccessDeniedExcludedFromSLA(t *testing.T) {
+	gin.SetMode(gin.TestMode)
+	rec := httptest.NewRecorder()
+	c, _ := gin.CreateTestContext(rec)
+	service.MarkOpsClientBusinessLimited(c, service.OpsClientBusinessLimitedReasonIPRestriction)
+
+	errType := normalizeOpsErrorType("api_error", "ACCESS_DENIED")
+	phase, isBusinessLimited, errorOwner, errorSource := classifyOpsErrorLog(c, errType, "Access denied", "ACCESS_DENIED", http.StatusForbidden)
+
+	require.Equal(t, "api_error", errType)
+	require.Equal(t, "auth", phase)
+	require.True(t, isBusinessLimited)
+	require.Equal(t, "client", errorOwner)
+	require.Equal(t, "client_request", errorSource)
+}
+
+func TestClassifyOpsOtherErrorsStillCountForSLA(t *testing.T) {
+	gin.SetMode(gin.TestMode)
+	rec := httptest.NewRecorder()
+	c, _ := gin.CreateTestContext(rec)
+
+	errType := normalizeOpsErrorType("api_error", "INTERNAL_ERROR")
+	phase, isBusinessLimited, errorOwner, errorSource := classifyOpsErrorLog(c, errType, "Failed to validate API key", "INTERNAL_ERROR", http.StatusInternalServerError)
+
+	require.Equal(t, "api_error", errType)
+	require.Equal(t, "internal", phase)
+	require.False(t, isBusinessLimited)
+	require.Equal(t, "platform", errorOwner)
+	require.Equal(t, "gateway", errorSource)
+}
+
 func TestClassifyOpsUnsupportedModelExcludedFromSLA(t *testing.T) {
 	tests := []string{
 		"No available accounts: no available accounts supporting model: made-up-model",

backend/internal/server/middleware/api_key_auth.go

@@ -92,6 +92,7 @@ func apiKeyAuthWithSubscription(apiKeyService *service.APIKeyService, subscripti
 			clientIP := ip.GetTrustedClientIP(c)
 			allowed, _ := ip.CheckIPRestrictionWithCompiledRules(clientIP, apiKey.CompiledIPWhitelist, apiKey.CompiledIPBlacklist)
 			if !allowed {
+				service.MarkOpsClientBusinessLimited(c, service.OpsClientBusinessLimitedReasonIPRestriction)
 				AbortWithError(c, 403, "ACCESS_DENIED", "Access denied")
 				return
 			}

backend/internal/server/middleware/api_key_auth_test.go

@@ -333,6 +333,15 @@ func TestAPIKeyAuthIPRestrictionDoesNotTrustSpoofedForwardHeaders(t *testing.T)
 	apiKeyService := service.NewAPIKeyService(apiKeyRepo, nil, nil, nil, nil, nil, cfg)
 	router := gin.New()
 	require.NoError(t, router.SetTrustedProxies(nil))
+	var markedBusinessLimited bool
+	var businessLimitedReason string
+	router.Use(func(c *gin.Context) {
+		c.Next()
+		markedBusinessLimited = service.HasOpsClientBusinessLimited(c)
+		if v, ok := c.Get(service.OpsClientBusinessLimitedReasonKey); ok {
+			businessLimitedReason, _ = v.(string)
+		}
+	})
 	router.Use(gin.HandlerFunc(NewAPIKeyAuthMiddleware(apiKeyService, nil, cfg)))
 	router.GET("/t", func(c *gin.Context) {
 		c.JSON(http.StatusOK, gin.H{"ok": true})
@@ -349,6 +358,8 @@ func TestAPIKeyAuthIPRestrictionDoesNotTrustSpoofedForwardHeaders(t *testing.T)
 
 	require.Equal(t, http.StatusForbidden, w.Code)
 	require.Contains(t, w.Body.String(), "ACCESS_DENIED")
+	require.True(t, markedBusinessLimited)
+	require.Equal(t, service.OpsClientBusinessLimitedReasonIPRestriction, businessLimitedReason)
 }
 
 func TestAPIKeyAuthTouchesLastUsedOnSuccess(t *testing.T) {

backend/internal/service/ops_upstream_context.go

@@ -36,6 +36,12 @@ const (
 	// OpsSkipPassthroughKey 由 applyErrorPassthroughRule 在命中 skip_monitoring=true 的规则时设置。
 	// ops_error_logger 中间件检查此 key，为 true 时跳过错误记录。
 	OpsSkipPassthroughKey = "ops_skip_passthrough"
+
+	// Client-side configuration denials should remain visible in ops_error_logs,
+	// but should be excluded from SLA/error-rate calculations.
+	OpsClientBusinessLimitedKey                 = "ops_client_business_limited"
+	OpsClientBusinessLimitedReasonKey           = "ops_client_business_limited_reason"
+	OpsClientBusinessLimitedReasonIPRestriction = "api_key_ip_restriction"
 )
 
 func setOpsUpstreamRequestBody(c *gin.Context, body []byte) {
@@ -53,6 +59,28 @@ func SetOpsLatencyMs(c *gin.Context, key string, value int64) {
 	c.Set(key, value)
 }
 
+func MarkOpsClientBusinessLimited(c *gin.Context, reason string) {
+	if c == nil {
+		return
+	}
+	c.Set(OpsClientBusinessLimitedKey, true)
+	if reason = strings.TrimSpace(reason); reason != "" {
+		c.Set(OpsClientBusinessLimitedReasonKey, reason)
+	}
+}
+
+func HasOpsClientBusinessLimited(c *gin.Context) bool {
+	if c == nil {
+		return false
+	}
+	v, ok := c.Get(OpsClientBusinessLimitedKey)
+	if !ok {
+		return false
+	}
+	marked, _ := v.(bool)
+	return marked
+}
+
 // SetOpsUpstreamError is the exported wrapper for setOpsUpstreamError, used by
 // handler-layer code (e.g. failover-exhausted paths) that needs to record the
 // original upstream status code before mapping it to a client-facing code.