Merge pull request #7160 from StarryKira/codex/log-retention-upstream

feat: 支持运维与请求日志的滚动保留设置

backend/cmd/server/wire_gen.go

@@ -185,7 +185,7 @@ func initializeApplication(buildInfo handler.BuildInfo) (*Application, error) {
 	dashboardAggregationRepository := repository.NewDashboardAggregationRepository(db)
 	dashboardStatsCache := repository.NewDashboardCache(redisClient, configConfig)
 	dashboardService := service.NewDashboardService(usageLogRepository, dashboardAggregationRepository, dashboardStatsCache, configConfig)
-	dashboardAggregationService := service.ProvideDashboardAggregationService(dashboardAggregationRepository, timingWheelService, leaderLockCache, db, configConfig)
+	dashboardAggregationService := service.ProvideDashboardAggregationService(dashboardAggregationRepository, timingWheelService, leaderLockCache, db, configConfig, settingRepository)
 	dashboardHandler := admin.NewDashboardHandler(dashboardService, dashboardAggregationService)
 	adminGroupRepository := repository.NewAdminGroupRepository(client, db)
 	adminAccountRepository := repository.NewAdminAccountRepository(client, db, schedulerCache)

backend/internal/repository/dashboard_aggregation_group_usage_test.go

@@ -249,6 +249,18 @@ func TestDashboardAggregationRepositoryCleanupUsageLogsPartitionedSortsAndInvali
 			WillReturnResult(sqlmock.NewResult(0, 0))
 		mock.ExpectCommit()
 	}
+	// The boundary partition is pruned by exact timestamp. tableoid is required
+	// because ctid alone can also identify a retained row in another partition.
+	mock.ExpectBegin()
+	mock.ExpectQuery(`SELECT id FROM usage_group_rollup_state.*FOR UPDATE`).
+		WillReturnRows(sqlmock.NewRows([]string{"id"}).AddRow(1))
+	mock.ExpectQuery(`(?s)SELECT tableoid, ctid.*WHERE created_at < \$1.*WHERE \(tableoid, ctid\) IN.*RETURNING created_at`).
+		WithArgs(cutoff, usageLogsCleanupBatchSize).
+		WillReturnRows(sqlmock.NewRows([]string{"created_at"}).AddRow(cutoff.Add(-time.Hour)))
+	mock.ExpectExec(`UPDATE usage_group_rollup_state`).
+		WithArgs(cutoff.Add(-time.Hour), "Asia/Shanghai").
+		WillReturnResult(sqlmock.NewResult(0, 1))
+	mock.ExpectCommit()
 	mock.ExpectBegin()
 	mock.ExpectQuery(`SELECT closed_before::text, retained_from.*FOR UPDATE`).
 		WillReturnRows(sqlmock.NewRows([]string{"closed_before", "retained_from", "timezone_name"}).
@@ -271,7 +283,7 @@ func TestDashboardAggregationRepositoryCleanupUsageLogsNonPartitionedFailureRoll
 	mock.ExpectBegin()
 	mock.ExpectQuery(`SELECT id FROM usage_group_rollup_state.*FOR UPDATE`).
 		WillReturnRows(sqlmock.NewRows([]string{"id"}).AddRow(1))
-	mock.ExpectQuery(`(?s)SELECT ctid.*ORDER BY created_at ASC, id ASC.*DELETE FROM usage_logs.*RETURNING created_at`).
+	mock.ExpectQuery(`(?s)SELECT tableoid, ctid.*ORDER BY created_at ASC, id ASC.*DELETE FROM usage_logs.*RETURNING created_at`).
 		WithArgs(cutoff, usageLogsCleanupBatchSize).
 		WillReturnRows(sqlmock.NewRows([]string{"created_at"}).AddRow(deletedAt))
 	mock.ExpectExec(`UPDATE usage_group_rollup_state`).

backend/internal/repository/dashboard_aggregation_repo.go

@@ -239,7 +239,10 @@ func (r *dashboardAggregationRepository) CleanupUsageLogs(ctx context.Context, c
 		if err := r.dropUsageLogsPartitions(ctx, cutoff); err != nil {
 			return err
 		}
-	} else if err := r.cleanupUsageLogsBatches(ctx, cutoff); err != nil {
+	}
+	// Remove the expired part of the boundary partition as well, so a retention
+	// window measured in days does not silently round up to a whole month.
+	if err := r.cleanupUsageLogsBatches(ctx, cutoff); err != nil {
 		return err
 	}
 	return r.SyncGroupUsageRollups(ctx, service.GroupUsageTodayStart(r.now()))
@@ -261,14 +264,14 @@ func (r *dashboardAggregationRepository) cleanupUsageLogsBatches(ctx context.Con
 
 		res, err := r.sql.ExecContext(ctx, `
 			WITH victims AS (
-				SELECT ctid
+				SELECT tableoid, ctid
 				FROM usage_logs
 				WHERE created_at < $1
 				ORDER BY created_at ASC, id ASC
 				LIMIT $2
 			)
 			DELETE FROM usage_logs
-			WHERE ctid IN (SELECT ctid FROM victims)
+			WHERE (tableoid, ctid) IN (SELECT tableoid, ctid FROM victims)
 		`, cutoff.UTC(), usageLogsCleanupBatchSize)
 		if err != nil {
 			return err
@@ -298,14 +301,14 @@ func cleanupUsageLogsBatchWithRollupInvalidation(ctx context.Context, db *sql.DB
 	}
 	rows, err := tx.QueryContext(ctx, `
 		WITH victims AS (
-			SELECT ctid
+			SELECT tableoid, ctid
 			FROM usage_logs
 			WHERE created_at < $1
 			ORDER BY created_at ASC, id ASC
 			LIMIT $2
 		)
 		DELETE FROM usage_logs
-		WHERE ctid IN (SELECT ctid FROM victims)
+		WHERE (tableoid, ctid) IN (SELECT tableoid, ctid FROM victims)
 		RETURNING created_at
 	`, cutoff.UTC(), usageLogsCleanupBatchSize)
 	if err != nil {

backend/internal/repository/request_log_retention_integration_test.go

@@ -0,0 +1,46 @@
+//go:build integration
+
+package repository
+
+import (
+	"context"
+	"testing"
+	"time"
+
+	"github.com/stretchr/testify/require"
+)
+
+func TestRequestLogRetention_PartitionBoundaryKeepsRecentRows(t *testing.T) {
+	ctx := context.Background()
+	tx := testTx(t)
+	// Temporary tables shadow the real schema and disappear on rollback.
+	_, err := tx.ExecContext(ctx, `
+		CREATE TEMP TABLE usage_logs (id bigint, created_at timestamptz) PARTITION BY RANGE (created_at);
+		CREATE TEMP TABLE retention_july PARTITION OF usage_logs FOR VALUES FROM ('2026-07-01') TO ('2026-08-01');
+		CREATE TEMP TABLE retention_august PARTITION OF usage_logs FOR VALUES FROM ('2026-08-01') TO ('2026-09-01');
+		INSERT INTO usage_logs VALUES
+			(1, '2026-07-17 23:59:59+00'),
+			(2, '2026-07-18 00:00:00+00'),
+			(3, '2026-08-01 00:00:00+00');
+	`)
+	require.NoError(t, err)
+	var collidingCTIDs bool
+	require.NoError(t, tx.QueryRowContext(ctx, `SELECT a.ctid = b.ctid FROM usage_logs a, usage_logs b WHERE a.id = 1 AND b.id = 3`).Scan(&collidingCTIDs))
+	require.True(t, collidingCTIDs, "fixture must exercise identical row locations in different partitions")
+	repo := newDashboardAggregationRepositoryWithSQL(tx)
+	cutoff := time.Date(2026, 7, 18, 0, 0, 0, 0, time.UTC)
+	for range 2 {
+		require.NoError(t, repo.cleanupUsageLogsBatches(ctx, cutoff))
+		rows, err := tx.QueryContext(ctx, `SELECT id FROM usage_logs ORDER BY id`)
+		require.NoError(t, err)
+		var ids []int64
+		for rows.Next() {
+			var id int64
+			require.NoError(t, rows.Scan(&id))
+			ids = append(ids, id)
+		}
+		require.NoError(t, rows.Err())
+		require.NoError(t, rows.Close())
+		require.Equal(t, []int64{2, 3}, ids)
+	}
+}

backend/internal/service/dashboard_aggregation_service.go

@@ -3,7 +3,9 @@ package service
 import (
 	"context"
 	"database/sql"
+	"encoding/json"
 	"errors"
+	"fmt"
 	"log/slog"
 	"sync/atomic"
 	"time"
@@ -55,6 +57,7 @@ type DashboardAggregationService struct {
 	repo                 DashboardAggregationRepository
 	timingWheel          *TimingWheelService
 	cfg                  config.DashboardAggregationConfig
+	settingRepo          SettingRepository
 	running              int32
 	lastRetentionCleanup atomic.Value // time.Time
 
@@ -95,6 +98,8 @@ func (s *DashboardAggregationService) Start() {
 	}
 	if !s.cfg.Enabled {
 		logger.LegacyPrintf("service.dashboard_aggregation", "[DashboardAggregation] 聚合作业已禁用")
+		// Explicit runtime retention remains available without dashboard aggregation.
+		s.timingWheel.ScheduleRecurring("dashboard:retention", time.Minute, s.runScheduledRetention)
 		return
 	}
 	go s.runStartupGroupUsageSync()
@@ -360,20 +365,38 @@ func (s *DashboardAggregationService) maybeCleanupRetention(ctx context.Context,
 		}
 	}
 
+	usageDays, err := s.requestRetentionDays(ctx)
+	if err != nil {
+		// Never fall back to a shorter window after a settings read failure.
+		logger.LegacyPrintf("service.dashboard_aggregation", "[DashboardAggregation] 读取日志保留设置失败，跳过清理: %v", err)
+		return
+	}
 	hourlyCutoff := now.AddDate(0, 0, -s.cfg.Retention.HourlyDays)
 	dailyCutoff := now.AddDate(0, 0, -s.cfg.Retention.DailyDays)
-	usageCutoff := now.AddDate(0, 0, -s.cfg.Retention.UsageLogsDays)
-	dedupCutoff := now.AddDate(0, 0, -s.cfg.Retention.UsageBillingDedupDays)
+	usageCutoff := now.AddDate(0, 0, -usageDays)
+	dedupDays := s.cfg.Retention.UsageBillingDedupDays
+	if dedupDays <= 0 {
+		dedupDays = 365
+	}
+	dedupDays = max(dedupDays, usageDays)
+	dedupCutoff := now.AddDate(0, 0, -dedupDays)
 
-	aggErr := s.repo.CleanupAggregates(ctx, hourlyCutoff, dailyCutoff)
+	var aggErr, usageErr, dedupErr error
+	if s.cfg.Enabled {
+		aggErr = s.repo.CleanupAggregates(ctx, hourlyCutoff, dailyCutoff)
+	}
 	if aggErr != nil {
 		logger.LegacyPrintf("service.dashboard_aggregation", "[DashboardAggregation] 聚合保留清理失败: %v", aggErr)
 	}
-	usageErr := s.repo.CleanupUsageLogs(ctx, usageCutoff)
+	if usageDays > 0 {
+		usageErr = s.repo.CleanupUsageLogs(ctx, usageCutoff)
+	}
 	if usageErr != nil {
 		logger.LegacyPrintf("service.dashboard_aggregation", "[DashboardAggregation] usage_logs 保留清理失败: %v", usageErr)
 	}
-	dedupErr := s.repo.CleanupUsageBillingDedup(ctx, dedupCutoff)
+	if usageDays > 0 {
+		dedupErr = s.repo.CleanupUsageBillingDedup(ctx, dedupCutoff)
+	}
 	if dedupErr != nil {
 		logger.LegacyPrintf("service.dashboard_aggregation", "[DashboardAggregation] usage_billing_dedup 保留清理失败: %v", dedupErr)
 	}
@@ -382,6 +405,51 @@ func (s *DashboardAggregationService) maybeCleanupRetention(ctx context.Context,
 	}
 }
 
+func (s *DashboardAggregationService) requestRetentionDays(ctx context.Context) (int, error) {
+	days := s.cfg.Retention.UsageLogsDays
+	if !s.cfg.Enabled {
+		days = 0
+	}
+	if s.settingRepo == nil {
+		return days, nil
+	}
+	raw, err := s.settingRepo.GetValue(ctx, SettingKeyOpsRuntimeLogConfig)
+	if errors.Is(err, ErrSettingNotFound) {
+		return days, nil
+	}
+	if err != nil {
+		return 0, err
+	}
+	var cfg struct {
+		RequestRetentionDays *int `json:"request_retention_days"`
+	}
+	if err := json.Unmarshal([]byte(raw), &cfg); err != nil {
+		return 0, err
+	}
+	if cfg.RequestRetentionDays != nil {
+		days = *cfg.RequestRetentionDays
+		if days < 0 || days > 3650 {
+			return 0, fmt.Errorf("invalid request_retention_days: %d", days)
+		}
+	}
+	return days, nil
+}
+
+func (s *DashboardAggregationService) runScheduledRetention() {
+	if !atomic.CompareAndSwapInt32(&s.running, 0, 1) {
+		return
+	}
+	defer atomic.StoreInt32(&s.running, 0)
+	ctx, cancel := context.WithTimeout(context.Background(), defaultDashboardAggregationTimeout)
+	defer cancel()
+	release, ok := tryAcquireSingletonLeaderLock(ctx, s.lockCache, s.db, dashboardAggregationLeaderLockKey, s.instanceID, dashboardAggregationLeaderLockTTL)
+	if !ok {
+		return
+	}
+	defer release()
+	s.maybeCleanupRetention(ctx, time.Now().UTC())
+}
+
 func truncateToDayUTC(t time.Time) time.Time {
 	t = t.UTC()
 	return time.Date(t.Year(), t.Month(), t.Day(), 0, 0, 0, 0, time.UTC)

backend/internal/service/dashboard_aggregation_service_test.go

@@ -11,21 +11,24 @@ import (
 )
 
 type dashboardAggregationRepoTestStub struct {
-	aggregateCalls       int
-	recomputeCalls       int
-	cleanupUsageCalls    int
-	cleanupDedupCalls    int
-	ensurePartitionCalls int
-	lastStart            time.Time
-	lastEnd              time.Time
-	watermark            time.Time
-	aggregateErr         error
-	cleanupAggregatesErr error
-	cleanupUsageErr      error
-	cleanupDedupErr      error
-	ensurePartitionErr   error
-	aggregateCtx         context.Context
-	events               *[]string
+	aggregateCalls        int
+	recomputeCalls        int
+	cleanupUsageCalls     int
+	cleanupUsageCutoff    time.Time
+	cleanupDedupCutoff    time.Time
+	cleanupAggregateCalls int
+	cleanupDedupCalls     int
+	ensurePartitionCalls  int
+	lastStart             time.Time
+	lastEnd               time.Time
+	watermark             time.Time
+	aggregateErr          error
+	cleanupAggregatesErr  error
+	cleanupUsageErr       error
+	cleanupDedupErr       error
+	ensurePartitionErr    error
+	aggregateCtx          context.Context
+	events                *[]string
 }
 
 type dashboardAggregationRollupRepoTestStub struct {
@@ -71,16 +74,19 @@ func (s *dashboardAggregationRepoTestStub) UpdateAggregationWatermark(ctx contex
 }
 
 func (s *dashboardAggregationRepoTestStub) CleanupAggregates(ctx context.Context, hourlyCutoff, dailyCutoff time.Time) error {
+	s.cleanupAggregateCalls++
 	return s.cleanupAggregatesErr
 }
 
 func (s *dashboardAggregationRepoTestStub) CleanupUsageLogs(ctx context.Context, cutoff time.Time) error {
 	s.cleanupUsageCalls++
+	s.cleanupUsageCutoff = cutoff
 	return s.cleanupUsageErr
 }
 
 func (s *dashboardAggregationRepoTestStub) CleanupUsageBillingDedup(ctx context.Context, cutoff time.Time) error {
 	s.cleanupDedupCalls++
+	s.cleanupDedupCutoff = cutoff
 	return s.cleanupDedupErr
 }
 
@@ -214,6 +220,7 @@ func TestDashboardAggregationService_CleanupRetentionFailure_DoesNotRecord(t *te
 	svc := &DashboardAggregationService{
 		repo: repo,
 		cfg: config.DashboardAggregationConfig{
+			Enabled: true,
 			Retention: config.DashboardAggregationRetentionConfig{
 				UsageLogsDays: 1,
 				HourlyDays:    1,
@@ -234,6 +241,7 @@ func TestDashboardAggregationService_CleanupDedupFailure_DoesNotRecord(t *testin
 	svc := &DashboardAggregationService{
 		repo: repo,
 		cfg: config.DashboardAggregationConfig{
+			Enabled: true,
 			Retention: config.DashboardAggregationRetentionConfig{
 				UsageLogsDays: 1,
 				HourlyDays:    1,

backend/internal/service/ops_log_runtime.go

@@ -14,15 +14,23 @@ import (
 )
 
 func defaultOpsRuntimeLogConfig(cfg *config.Config) *OpsRuntimeLogConfig {
+	requestRetentionDays := 90
+	if cfg != nil {
+		requestRetentionDays = cfg.DashboardAgg.Retention.UsageLogsDays
+		if !cfg.DashboardAgg.Enabled {
+			requestRetentionDays = 0
+		}
+	}
 	out := &OpsRuntimeLogConfig{
-		Level:             "info",
-		PersistAccessLogs: false,
-		EnableSampling:    false,
-		SamplingInitial:   100,
-		SamplingNext:      100,
-		Caller:            true,
-		StacktraceLevel:   "error",
-		RetentionDays:     30,
+		Level:                "info",
+		PersistAccessLogs:    false,
+		EnableSampling:       false,
+		SamplingInitial:      100,
+		SamplingNext:         100,
+		Caller:               true,
+		StacktraceLevel:      "error",
+		RetentionDays:        30,
+		RequestRetentionDays: &requestRetentionDays,
 	}
 	if cfg == nil {
 		return out
@@ -60,6 +68,10 @@ func normalizeOpsRuntimeLogConfig(cfg *OpsRuntimeLogConfig, defaults *OpsRuntime
 	if cfg.RetentionDays <= 0 {
 		cfg.RetentionDays = defaults.RetentionDays
 	}
+	if cfg.RequestRetentionDays == nil && defaults.RequestRetentionDays != nil {
+		days := *defaults.RequestRetentionDays
+		cfg.RequestRetentionDays = &days
+	}
 }
 
 func validateOpsRuntimeLogConfig(cfg *OpsRuntimeLogConfig) error {
@@ -85,6 +97,9 @@ func validateOpsRuntimeLogConfig(cfg *OpsRuntimeLogConfig) error {
 	if cfg.RetentionDays < 1 || cfg.RetentionDays > 3650 {
 		return errors.New("retention_days must be between 1 and 3650")
 	}
+	if cfg.RequestRetentionDays != nil && (*cfg.RequestRetentionDays < 0 || *cfg.RequestRetentionDays > 3650) {
+		return errors.New("request_retention_days must be between 0 and 3650")
+	}
 	return nil
 }
 
@@ -139,6 +154,9 @@ func (s *OpsService) UpdateRuntimeLogConfig(ctx context.Context, req *OpsRuntime
 		return nil, err
 	}
 	next := *req
+	if next.RequestRetentionDays == nil {
+		next.RequestRetentionDays = oldCfg.RequestRetentionDays
+	}
 	normalizeOpsRuntimeLogConfig(&next, defaultOpsRuntimeLogConfig(s.cfg))
 	if err := validateOpsRuntimeLogConfig(&next); err != nil {
 		s.auditRuntimeLogConfigFailure(operatorID, oldCfg, &next, "validation_failed: "+err.Error())

backend/internal/service/ops_settings_models.go

@@ -77,11 +77,14 @@ type OpsRuntimeLogConfig struct {
 	Caller            bool   `json:"caller"`
 	StacktraceLevel   string `json:"stacktrace_level"`
 	// RetentionDays is the authoritative retention for ops_system_logs.
-	RetentionDays   int            `json:"retention_days"`
-	Source          string         `json:"source,omitempty"`
-	UpdatedAt       string         `json:"updated_at,omitempty"`
-	UpdatedByUserID int64          `json:"updated_by_user_id,omitempty"`
-	Extra           map[string]any `json:"extra,omitempty"`
+	RetentionDays int `json:"retention_days"`
+	// RequestRetentionDays controls usage_logs; 0 keeps request history indefinitely.
+	// A nil value from older clients preserves the existing setting.
+	RequestRetentionDays *int           `json:"request_retention_days"`
+	Source               string         `json:"source,omitempty"`
+	UpdatedAt            string         `json:"updated_at,omitempty"`
+	UpdatedByUserID      int64          `json:"updated_by_user_id,omitempty"`
+	Extra                map[string]any `json:"extra,omitempty"`
 }
 
 type OpsAlertRuntimeSettings struct {

backend/internal/service/request_log_retention_test.go

@@ -0,0 +1,109 @@
+package service
+
+import (
+	"context"
+	"encoding/json"
+	"errors"
+	"testing"
+	"time"
+
+	"github.com/Wei-Shaw/sub2api/internal/config"
+	"github.com/Wei-Shaw/sub2api/internal/pkg/logger"
+	"github.com/stretchr/testify/require"
+)
+
+func TestRequestLogRetention_RuntimePolicy(t *testing.T) {
+	for _, tt := range []struct {
+		name        string
+		raw         string
+		enabled     bool
+		wantDays    int
+		wantCleanup bool
+	}{
+		{"legacy uses deployment", `{"retention_days":30}`, true, 90, true},
+		{"short window", `{"request_retention_days":7}`, true, 7, true},
+		{"long window extends dedup", `{"request_retention_days":730}`, true, 730, true},
+		{"forever skips request and dedup deletion", `{"request_retention_days":0}`, true, 0, false},
+		{"works without aggregation", `{"request_retention_days":14}`, false, 14, true},
+		{"legacy disabled stays disabled", `{"retention_days":30}`, false, 0, false},
+	} {
+		t.Run(tt.name, func(t *testing.T) {
+			settings := newRuntimeSettingRepoStub()
+			settings.values[SettingKeyOpsRuntimeLogConfig] = tt.raw
+			repo := &dashboardAggregationRepoTestStub{}
+			svc := NewDashboardAggregationService(repo, nil, &config.Config{DashboardAgg: config.DashboardAggregationConfig{
+				Enabled:   tt.enabled,
+				Retention: config.DashboardAggregationRetentionConfig{UsageLogsDays: 90, UsageBillingDedupDays: 365, HourlyDays: 180, DailyDays: 730},
+			}})
+			svc.settingRepo = settings
+			now := time.Date(2026, 9, 15, 12, 0, 0, 0, time.UTC)
+			svc.maybeCleanupRetention(context.Background(), now)
+			if tt.wantCleanup {
+				require.Equal(t, 1, repo.cleanupUsageCalls)
+				require.Equal(t, now.AddDate(0, 0, -tt.wantDays), repo.cleanupUsageCutoff)
+				require.Equal(t, now.AddDate(0, 0, -max(365, tt.wantDays)), repo.cleanupDedupCutoff)
+			} else {
+				require.Zero(t, repo.cleanupUsageCalls)
+				require.Zero(t, repo.cleanupDedupCalls)
+			}
+			if !tt.enabled {
+				require.Zero(t, repo.cleanupAggregateCalls)
+			}
+			// Read the updated policy at the next rolling cleanup.
+			settings.values[SettingKeyOpsRuntimeLogConfig] = `{"request_retention_days":180}`
+			next := now.Add(dashboardAggregationRetentionInterval)
+			svc.maybeCleanupRetention(context.Background(), next)
+			require.Equal(t, next.AddDate(0, 0, -180), repo.cleanupUsageCutoff)
+		})
+	}
+}
+
+func TestRequestLogRetention_ReadFailureSkipsDeletion(t *testing.T) {
+	for _, raw := range []string{`{broken`, `{"request_retention_days":-1}`, `{"request_retention_days":3651}`} {
+		settings := newRuntimeSettingRepoStub()
+		settings.values[SettingKeyOpsRuntimeLogConfig] = raw
+		repo := &dashboardAggregationRepoTestStub{}
+		svc := &DashboardAggregationService{repo: repo, settingRepo: settings}
+		svc.runScheduledRetention()
+		require.Zero(t, repo.cleanupUsageCalls)
+		require.Zero(t, repo.cleanupDedupCalls)
+		require.Nil(t, svc.lastRetentionCleanup.Load())
+	}
+	settings := newRuntimeSettingRepoStub()
+	settings.getValueFn = func(string) (string, error) { return "", errors.New("database unavailable") }
+	svc := &DashboardAggregationService{repo: &dashboardAggregationRepoTestStub{}, settingRepo: settings}
+	svc.runScheduledRetention()
+	require.Nil(t, svc.lastRetentionCleanup.Load())
+}
+
+func TestRuntimeLogConfig_RequestRetentionCompatibility(t *testing.T) {
+	settings := newRuntimeSettingRepoStub()
+	settings.values[SettingKeyOpsRuntimeLogConfig] = `{"level":"info","retention_days":7,"request_retention_days":180}`
+	svc := &OpsService{settingRepo: settings}
+	require.NoError(t, logger.Init(logger.InitOptions{Level: "info", Format: "json", Output: logger.OutputOptions{ToStdout: true}}))
+	t.Cleanup(logger.Sync)
+	legacy := defaultOpsRuntimeLogConfig(nil)
+	legacy.RequestRetentionDays = nil
+	updated, err := svc.UpdateRuntimeLogConfig(context.Background(), legacy, 1)
+	require.NoError(t, err)
+	require.Equal(t, 180, *updated.RequestRetentionDays)
+	for _, days := range []int{0, 1, 3650} {
+		updated.RequestRetentionDays = &days
+		saved, err := svc.UpdateRuntimeLogConfig(context.Background(), updated, 1)
+		require.NoError(t, err)
+		require.Equal(t, days, *saved.RequestRetentionDays)
+		var persisted OpsRuntimeLogConfig
+		require.NoError(t, json.Unmarshal([]byte(settings.values[SettingKeyOpsRuntimeLogConfig]), &persisted))
+		require.Equal(t, days, *persisted.RequestRetentionDays)
+	}
+	for _, days := range []int{-1, 3651} {
+		updated.RequestRetentionDays = &days
+		before := settings.values[SettingKeyOpsRuntimeLogConfig]
+		_, err := svc.UpdateRuntimeLogConfig(context.Background(), updated, 1)
+		require.ErrorContains(t, err, "request_retention_days")
+		require.Equal(t, before, settings.values[SettingKeyOpsRuntimeLogConfig])
+	}
+	reset, err := svc.ResetRuntimeLogConfig(context.Background(), 1)
+	require.NoError(t, err)
+	require.Equal(t, 90, *reset.RequestRetentionDays)
+}

backend/internal/service/wire.go

@@ -381,8 +381,9 @@ func ProvideGrokTokenProvider(
 }
 
 // ProvideDashboardAggregationService 创建并启动仪表盘聚合服务
-func ProvideDashboardAggregationService(repo DashboardAggregationRepository, timingWheel *TimingWheelService, lockCache LeaderLockCache, db *sql.DB, cfg *config.Config) *DashboardAggregationService {
+func ProvideDashboardAggregationService(repo DashboardAggregationRepository, timingWheel *TimingWheelService, lockCache LeaderLockCache, db *sql.DB, cfg *config.Config, settingRepo SettingRepository) *DashboardAggregationService {
 	svc := NewDashboardAggregationService(repo, timingWheel, cfg)
+	svc.settingRepo = settingRepo
 	svc.SetLeaderLock(lockCache, db)
 	svc.Start()
 	return svc

docs/log-retention.md

@@ -0,0 +1,21 @@
+# 滚动日志保留设置
+
+管理员在 **运维监控 → 系统日志 → 运行时日志配置** 中设置保留期，然后点击「保存并应用」。
+
+| 日志 | 可选范围 | 执行时间 |
+| --- | --- | --- |
+| 运维系统日志 | 1–3650 天 | 运维设置中的数据清理计划，须开启「启用数据清理」 |
+| 请求日志（请求使用明细） | 1–3650 天，或永久保留 | 每 6 小时清理一次，保存后在下一次清理生效 |
+
+下拉框提供 1、3、7、14、30、60、90、180、365 天，也可以选择「自定义天数」。按天清理只删除早于截止时间的记录；截止时间上的记录保留。请求日志按月分区时，也会清理截止日期所在分区中的过期记录。
+
+## 配置兼容
+
+- 设置保存在 `ops_runtime_log_config`：`retention_days` 控制运维系统日志，`request_retention_days` 控制请求日志，后者为 `0` 时表示永久保留。
+- 没有请求日志设置时，沿用 `dashboard_aggregation.retention.usage_logs_days`（默认 90 天）；原来禁用仪表盘聚合的部署继续保留请求日志，不自动开启删除。
+- 显式保存请求日志保留天数后，即使仪表盘聚合关闭，滚动清理仍会运行。
+- 旧客户端没有传入 `request_retention_days` 时，保留已有值。「重置默认值」恢复部署配置。
+- 计费去重记录的保留期至少与请求日志一样长；请求日志永久保留时，也停止清理计费去重记录。
+- 读取请求日志保留设置失败或设置无效时，跳过清理，避免回退到更短的保留期。
+
+运维错误日志和指标继续使用运维设置中的各自保留策略。本页的运维保留期针对数据库中的系统日志；磁盘日志文件仍使用部署配置中的轮转策略。

frontend/src/api/admin/ops.ts

@@ -822,6 +822,7 @@ export interface OpsRuntimeLogConfig {
   caller: boolean
   stacktrace_level: 'none' | 'error' | 'fatal'
   retention_days: number
+  request_retention_days: number
   source?: string
   updated_at?: string
   updated_by_user_id?: number

frontend/src/i18n/locales/en/admin/ops.ts

@@ -41,8 +41,14 @@ export default {
         stacktraceThreshold: 'Stacktrace threshold',
         samplingInitial: 'Sampling initial',
         samplingThereafter: 'Sampling thereafter',
-        retentionDays: 'Retention days',
-        retentionDaysHint: 'Applied by the scheduled data-cleanup job.',
+        retentionDays: 'Operations log retention days',
+        requestRetentionDays: 'Request log retention days',
+        requestRetentionDaysHint: 'Request usage records are pruned every 6 hours. Changes apply on the next cleanup. Keeping records forever uses increasing storage.',
+        retentionDaysInvalid: 'Keep operations logs for 1–3650 days; keep request logs for 1–3650 days or choose Forever.',
+        retentionDaysOption: '{days} days',
+        retentionDaysCustom: 'Custom days',
+        retentionForever: 'Forever',
+        retentionDaysHint: 'Applied on the data-cleanup schedule when cleanup is enabled in Operations Settings.',
         caller: 'caller',
         sampling: 'sampling',
         persistAccessLogs: 'Store access logs in database',
@@ -69,6 +75,7 @@ export default {
         logDetails: 'Log Details',
         loadFailed: 'Failed to load system logs',
         runtimeConfigActive: 'Runtime log configuration is active',
+        runtimeConfigLoadFailed: 'Failed to load log configuration. Refresh and try again.',
         runtimeConfigSaveFailed: 'Failed to save log configuration',
         resetRuntimeConfigConfirm: 'Reset to startup configuration (env/yaml) and apply immediately?',
         runtimeConfigReset: 'Reset to startup log configuration',

frontend/src/i18n/locales/zh/admin/ops.ts

@@ -41,8 +41,14 @@ export default {
         stacktraceThreshold: '堆栈阈值',
         samplingInitial: '采样初始条数',
         samplingThereafter: '后续采样间隔',
-        retentionDays: '保留天数',
-        retentionDaysHint: '由定时数据清理任务执行。',
+        retentionDays: '运维日志保留天数',
+        requestRetentionDays: '请求日志保留天数',
+        requestRetentionDaysHint: '请求使用明细每 6 小时滚动清理一次，保存后在下次清理生效。永久保留会持续占用存储空间。',
+        retentionDaysInvalid: '运维日志须保留 1–3650 天；请求日志须保留 1–3650 天或选择永久保留。',
+        retentionDaysOption: '{days} 天',
+        retentionDaysCustom: '自定义天数',
+        retentionForever: '永久保留',
+        retentionDaysHint: '按运维设置中启用的数据清理计划执行。',
         caller: '调用方',
         sampling: '采样',
         persistAccessLogs: '将访问日志写入数据库',
@@ -69,6 +75,7 @@ export default {
         logDetails: '日志详情',
         loadFailed: '加载系统日志失败',
         runtimeConfigActive: '运行时日志配置已生效',
+        runtimeConfigLoadFailed: '日志配置加载失败，请刷新后重试。',
         runtimeConfigSaveFailed: '保存日志配置失败',
         resetRuntimeConfigConfirm: '确定要重置为启动配置（env/yaml）并立即应用吗？',
         runtimeConfigReset: '已重置为启动日志配置',

frontend/src/views/admin/ops/components/LogRetentionSelect.vue

@@ -0,0 +1,43 @@
+<script setup lang="ts">
+import { computed, ref } from 'vue'
+import { useI18n } from 'vue-i18n'
+import Select from '@/components/common/Select.vue'
+
+const props = defineProps<{
+  modelValue: number
+  label: string
+  allowForever?: boolean
+}>()
+const emit = defineEmits<{ 'update:modelValue': [value: number] }>()
+const { t } = useI18n()
+const presets = [1, 3, 7, 14, 30, 60, 90, 180, 365]
+const custom = ref(false)
+const options = computed(() => [
+  ...(props.allowForever ? [{ value: 0, label: t('admin.ops.systemLogs.retentionForever') }] : []),
+  ...presets.map(days => ({ value: days, label: t('admin.ops.systemLogs.retentionDaysOption', { days }) })),
+  { value: 'custom', label: t('admin.ops.systemLogs.retentionDaysCustom') }
+])
+const selection = computed(() => custom.value || !(presets.includes(props.modelValue) || (props.allowForever && props.modelValue === 0)) ? 'custom' : props.modelValue)
+
+function select(value: string | number | boolean | null | undefined) {
+  custom.value = value === 'custom'
+  if (typeof value === 'number') emit('update:modelValue', value)
+}
+</script>
+
+<template>
+  <div class="space-y-2">
+    <Select :model-value="selection" :options="options" :aria-label="label" @update:model-value="select" />
+    <input
+      v-if="selection === 'custom'"
+      :value="modelValue"
+      type="number"
+      min="1"
+      max="3650"
+      step="1"
+      class="input"
+      :aria-label="`${label}: ${t('admin.ops.systemLogs.retentionDaysCustom')}`"
+      @input="emit('update:modelValue', ($event.target as HTMLInputElement).valueAsNumber)"
+    />
+  </div>
+</template>

frontend/src/views/admin/ops/components/OpsSystemLogTable.vue

@@ -2,6 +2,7 @@
 import { computed, onMounted, reactive, ref, watch } from 'vue'
 import { useMediaQuery } from '@vueuse/core'
 import { useI18n } from 'vue-i18n'
+import LogRetentionSelect from './LogRetentionSelect.vue'
 import { opsAPI, type OpsRuntimeLogConfig, type OpsSystemLog, type OpsSystemLogSinkHealth } from '@/api/admin/ops'
 import Pagination from '@/components/common/Pagination.vue'
 import Select from '@/components/common/Select.vue'
@@ -38,6 +39,7 @@ const health = ref<OpsSystemLogSinkHealth>({
 })
 
 const runtimeLoading = ref(false)
+const runtimeConfigLoaded = ref(false)
 const runtimeSaving = ref(false)
 const runtimeConfig = reactive<OpsRuntimeLogConfig>({
   level: 'info',
@@ -47,7 +49,8 @@ const runtimeConfig = reactive<OpsRuntimeLogConfig>({
   sampling_thereafter: 100,
   caller: true,
   stacktrace_level: 'error',
-  retention_days: 30
+  retention_days: 30,
+  request_retention_days: 90
 })
 
 const filters = reactive({
@@ -229,6 +232,7 @@ const fetchHealth = async () => {
 
 const loadRuntimeConfig = async () => {
   runtimeLoading.value = true
+  runtimeConfigLoaded.value = false
   try {
     const cfg = await opsAPI.getRuntimeLogConfig()
     runtimeConfig.level = cfg.level
@@ -239,14 +243,23 @@ const loadRuntimeConfig = async () => {
     runtimeConfig.caller = cfg.caller
     runtimeConfig.stacktrace_level = cfg.stacktrace_level
     runtimeConfig.retention_days = cfg.retention_days
+    runtimeConfig.request_retention_days = cfg.request_retention_days ?? 90
+    runtimeConfigLoaded.value = true
   } catch (err: any) {
     console.error('[OpsSystemLogTable] Failed to load runtime log config', err)
+    appStore.showError(t('admin.ops.systemLogs.runtimeConfigLoadFailed'))
   } finally {
     runtimeLoading.value = false
   }
 }
 
 const saveRuntimeConfig = async () => {
+  if (!runtimeConfigLoaded.value || runtimeLoading.value) return
+  if (!Number.isInteger(runtimeConfig.retention_days) || runtimeConfig.retention_days < 1 || runtimeConfig.retention_days > 3650 ||
+      !Number.isInteger(runtimeConfig.request_retention_days) || runtimeConfig.request_retention_days < 0 || runtimeConfig.request_retention_days > 3650) {
+    appStore.showError(t('admin.ops.systemLogs.retentionDaysInvalid'))
+    return
+  }
   runtimeSaving.value = true
   try {
     const saved = await opsAPI.updateRuntimeLogConfig({ ...runtimeConfig })
@@ -258,6 +271,7 @@ const saveRuntimeConfig = async () => {
     runtimeConfig.caller = saved.caller
     runtimeConfig.stacktrace_level = saved.stacktrace_level
     runtimeConfig.retention_days = saved.retention_days
+    runtimeConfig.request_retention_days = saved.request_retention_days ?? 90
     appStore.showSuccess(t('admin.ops.systemLogs.runtimeConfigActive'))
   } catch (err: any) {
     console.error('[OpsSystemLogTable] Failed to save runtime log config', err)
@@ -282,6 +296,7 @@ const resetRuntimeConfig = async () => {
     runtimeConfig.caller = saved.caller
     runtimeConfig.stacktrace_level = saved.stacktrace_level
     runtimeConfig.retention_days = saved.retention_days
+    runtimeConfig.request_retention_days = saved.request_retention_days ?? 90
     appStore.showSuccess(t('admin.ops.systemLogs.runtimeConfigReset'))
     await fetchHealth()
   } catch (err: any) {
@@ -422,9 +437,14 @@ onMounted(async () => {
         </label>
         <label class="text-xs text-gray-600 dark:text-gray-300">
           {{ t('admin.ops.systemLogs.retentionDays') }}
-          <input v-model.number="runtimeConfig.retention_days" type="number" min="1" max="3650" class="input mt-1" />
+          <LogRetentionSelect v-model="runtimeConfig.retention_days" :label="t('admin.ops.systemLogs.retentionDays')" class="mt-1" />
           <span class="mt-1 block text-[11px] text-gray-500 dark:text-gray-400">{{ t('admin.ops.systemLogs.retentionDaysHint') }}</span>
         </label>
+        <label class="text-xs text-gray-600 dark:text-gray-300">
+          {{ t('admin.ops.systemLogs.requestRetentionDays') }}
+          <LogRetentionSelect v-model="runtimeConfig.request_retention_days" :label="t('admin.ops.systemLogs.requestRetentionDays')" allow-forever class="mt-1" />
+          <span class="mt-1 block text-[11px] text-gray-500 dark:text-gray-400">{{ t('admin.ops.systemLogs.requestRetentionDaysHint') }}</span>
+        </label>
         <div class="md:col-span-2 xl:col-span-6">
           <div class="grid gap-3 lg:grid-cols-[minmax(0,1fr)_auto] lg:items-end">
             <div class="flex flex-wrap items-center gap-x-4 gap-y-2">
@@ -442,7 +462,7 @@ onMounted(async () => {
               </label>
             </div>
             <div class="flex flex-wrap items-center gap-2 lg:justify-end">
-              <button type="button" class="btn btn-primary btn-sm" :disabled="runtimeSaving" @click="saveRuntimeConfig">
+              <button type="button" class="btn btn-primary btn-sm" :disabled="runtimeSaving || runtimeLoading || !runtimeConfigLoaded" @click="saveRuntimeConfig">
                 {{ runtimeSaving ? t('common.saving') : t('admin.ops.systemLogs.saveAndApply') }}
               </button>
               <button type="button" class="btn btn-secondary btn-sm" :disabled="runtimeSaving" @click="resetRuntimeConfig">

frontend/src/views/admin/ops/components/__tests__/OpsSystemLogTable.spec.ts

@@ -2,6 +2,7 @@ import { beforeEach, describe, expect, it, vi } from 'vitest'
 import { defineComponent } from 'vue'
 import { flushPromises, mount } from '@vue/test-utils'
 import OpsSystemLogTable from '../OpsSystemLogTable.vue'
+import LogRetentionSelect from '../LogRetentionSelect.vue'
 import enLocale from '@/i18n/locales/en'
 import zhLocale from '@/i18n/locales/zh'
 
@@ -9,6 +10,8 @@ const mockListSystemLogs = vi.fn()
 const mockCleanupSystemLogs = vi.fn()
 const mockGetSystemLogSinkHealth = vi.fn()
 const mockGetRuntimeLogConfig = vi.fn()
+const mockUpdateRuntimeLogConfig = vi.fn()
+const mockShowError = vi.fn()
 
 vi.mock('@/api/admin/ops', () => ({
   opsAPI: {
@@ -16,12 +19,13 @@ vi.mock('@/api/admin/ops', () => ({
     cleanupSystemLogs: (...args: any[]) => mockCleanupSystemLogs(...args),
     getSystemLogSinkHealth: (...args: any[]) => mockGetSystemLogSinkHealth(...args),
     getRuntimeLogConfig: (...args: any[]) => mockGetRuntimeLogConfig(...args),
+    updateRuntimeLogConfig: (...args: any[]) => mockUpdateRuntimeLogConfig(...args),
   },
 }))
 
 vi.mock('@/stores', () => ({
   useAppStore: () => ({
-    showError: vi.fn(),
+    showError: (...args: any[]) => mockShowError(...args),
     showSuccess: vi.fn(),
   }),
 }))
@@ -60,6 +64,7 @@ const runtimeConfig = {
   caller: true,
   stacktrace_level: 'error',
   retention_days: 30,
+  request_retention_days: 90,
 }
 
 const sinkHealth = {
@@ -93,6 +98,7 @@ describe('OpsSystemLogTable host support', () => {
     mockCleanupSystemLogs.mockResolvedValue({ deleted: 1 })
     mockGetSystemLogSinkHealth.mockResolvedValue(sinkHealth)
     mockGetRuntimeLogConfig.mockResolvedValue(runtimeConfig)
+    mockUpdateRuntimeLogConfig.mockImplementation(async config => config)
   })
 
   it('renders the host and sends it with list and cleanup filters', async () => {
@@ -150,3 +156,60 @@ describe('OpsSystemLogTable host support', () => {
     expect(locale.admin.ops.systemLogs.host).toBe('Host')
   })
 })
+
+
+describe('rolling log retention settings', () => {
+  const mountTable = () => mount(OpsSystemLogTable, {
+    global: { stubs: { Select: SelectStub, Pagination: PaginationStub } }
+  })
+  const saveButton = (wrapper: ReturnType<typeof mountTable>) => wrapper.findAll('button').find(button => button.text() === 'admin.ops.systemLogs.saveAndApply')!
+
+  beforeEach(() => {
+    vi.clearAllMocks()
+    mockListSystemLogs.mockResolvedValue({ items: [], total: 0, page: 1, page_size: 20 })
+    mockGetSystemLogSinkHealth.mockResolvedValue(sinkHealth)
+    mockGetRuntimeLogConfig.mockResolvedValue(runtimeConfig)
+    mockUpdateRuntimeLogConfig.mockImplementation(async config => config)
+  })
+
+  it('loads and saves separate operations and request retention windows', async () => {
+    const wrapper = mountTable()
+    await flushPromises()
+    const fields = wrapper.findAllComponents(LogRetentionSelect)
+    expect(fields.map(field => field.props('modelValue'))).toEqual([30, 90])
+    fields[0]!.vm.$emit('update:modelValue', 7)
+    fields[1]!.vm.$emit('update:modelValue', 180)
+    await saveButton(wrapper).trigger('click')
+    await flushPromises()
+    expect(mockUpdateRuntimeLogConfig).toHaveBeenCalledWith(expect.objectContaining({ retention_days: 7, request_retention_days: 180 }))
+    wrapper.unmount()
+  })
+
+  it('supports forever and rejects fractional or empty custom days', async () => {
+    const wrapper = mountTable()
+    await flushPromises()
+    const requestField = wrapper.findAllComponents(LogRetentionSelect)[1]!
+    requestField.vm.$emit('update:modelValue', 0)
+    await saveButton(wrapper).trigger('click')
+    await flushPromises()
+    expect(mockUpdateRuntimeLogConfig).toHaveBeenCalledWith(expect.objectContaining({ request_retention_days: 0 }))
+    mockUpdateRuntimeLogConfig.mockClear()
+    for (const value of [1.5, Number.NaN, -1, 3651]) {
+      requestField.vm.$emit('update:modelValue', value)
+      await saveButton(wrapper).trigger('click')
+    }
+    expect(mockUpdateRuntimeLogConfig).not.toHaveBeenCalled()
+    expect(mockShowError).toHaveBeenCalledWith('admin.ops.systemLogs.retentionDaysInvalid')
+    wrapper.unmount()
+  })
+
+  it('does not overwrite saved retention with defaults after a load failure', async () => {
+    mockGetRuntimeLogConfig.mockRejectedValue(new Error('unavailable'))
+    const wrapper = mountTable()
+    await flushPromises()
+    expect(saveButton(wrapper).attributes('disabled')).toBeDefined()
+    await saveButton(wrapper).trigger('click')
+    expect(mockUpdateRuntimeLogConfig).not.toHaveBeenCalled()
+    wrapper.unmount()
+  })
+})