code re-org and bugfix hanging hourly snapshot

internal/tasks/inventoryHelpers.go

@@ -44,10 +44,11 @@ func boolStringFromInterface(value interface{}) string {
 // latestHourlySnapshotBefore finds the most recent hourly snapshot table prior to the given time, skipping empty tables.
 func latestHourlySnapshotBefore(ctx context.Context, dbConn *sqlx.DB, cutoff time.Time) (string, error) {
 	rows, err := dbConn.QueryxContext(ctx, `
-SELECT table_name, snapshot_time
+SELECT table_name, snapshot_time, snapshot_count
 FROM snapshot_registry
-WHERE snapshot_type = 'hourly' AND snapshot_time < ?
+WHERE snapshot_type = 'hourly' AND snapshot_time < ? AND snapshot_count > 0
 ORDER BY snapshot_time DESC
+LIMIT 50
 `, cutoff.Unix())
 	if err != nil {
 		return "", err
@@ -57,17 +58,18 @@ ORDER BY snapshot_time DESC
 	for rows.Next() {
 		var name string
 		var ts int64
-		if scanErr := rows.Scan(&name, &ts); scanErr != nil {
+		var count int64
+		if scanErr := rows.Scan(&name, &ts, &count); scanErr != nil {
 			continue
 		}
 		if err := db.ValidateTableName(name); err != nil {
 			continue
 		}
-		hasRows, err := db.TableHasRows(ctx, dbConn, name)
-		if err != nil {
-			continue
+		// Rely on snapshot_count to avoid costly table scans; fall back to a cheap row check only if count is zero.
+		if count > 0 {
+			return name, nil
 		}
-		if hasRows {
+		if hasRows, _ := db.TableHasRows(ctx, dbConn, name); hasRows {
 			return name, nil
 		}
 	}

internal/tasks/inventoryLifecycle.go

@@ -206,6 +206,8 @@ func isPresent(presence map[string]struct{}, cand lifecycleCandidate) bool {
 	return false
 }
 
+// findDeletionInTables walks ordered hourly tables for a vCenter and returns the first confirmed deletion time
+// (requiring two consecutive misses) plus the time of the first miss for cross-day handling.
 func findDeletionInTables(ctx context.Context, dbConn *sqlx.DB, tables []snapshotTable, vcenter string, cand lifecycleCandidate) (int64, int64) {
 	var lastSeen int64
 	var firstMiss int64

internal/tasks/inventorySnapshots.go

@@ -24,7 +24,8 @@ import (
 )
 
 // RunVcenterSnapshotHourly records hourly inventory snapshots into a daily table.
-func (c *CronTask) RunVcenterSnapshotHourly(ctx context.Context, logger *slog.Logger) (err error) {
+// If force is true, any in-progress marker will be cleared before starting (useful for manual recovery).
+func (c *CronTask) RunVcenterSnapshotHourly(ctx context.Context, logger *slog.Logger, force bool) (err error) {
 	jobCtx := ctx
 	jobTimeout := durationFromSeconds(c.Settings.Values.Settings.HourlyJobTimeoutSeconds, 20*time.Minute)
 	if jobTimeout > 0 {
@@ -40,6 +41,13 @@ func (c *CronTask) RunVcenterSnapshotHourly(ctx context.Context, logger *slog.Lo
 	if err := tracker.ClearStale(staleCtx, "hourly_snapshot", jobTimeout); err != nil {
 		logger.Warn("failed to clear stale cron status", "error", err)
 	}
+	if force {
+		if err := tracker.ClearAllInProgress(staleCtx); err != nil {
+			logger.Warn("failed to clear in-progress flag (force run)", "error", err)
+		} else {
+			logger.Info("force run cleared in-progress marker before starting")
+		}
+	}
 
 	startedAt := time.Now()
 	defer func() {
@@ -51,8 +59,24 @@ func (c *CronTask) RunVcenterSnapshotHourly(ctx context.Context, logger *slog.Lo
 		return err
 	}
 	if skip {
-		logger.Warn("Hourly snapshot skipped because a previous run is still active")
-		return nil
+		if force {
+			logger.Info("Force run requested; clearing in-progress marker and retrying")
+			if err := tracker.ClearAllInProgress(jobCtx); err != nil {
+				logger.Warn("failed to clear in-progress flag for force run", "error", err)
+				return nil
+			}
+			done, skip, err = tracker.Start(jobCtx, "hourly_snapshot")
+			if err != nil {
+				return err
+			}
+			if skip {
+				logger.Warn("Hourly snapshot still marked active after force clear; skipping")
+				return nil
+			}
+		} else {
+			logger.Warn("Hourly snapshot skipped because a previous run is still active", "force", force)
+			return nil
+		}
 	}
 	defer func() { done(err) }()
 
@@ -824,6 +848,7 @@ func (c *CronTask) captureHourlySnapshotForVcenter(ctx context.Context, startTim
 
 	missingCount := 0
 	newCount := 0
+	prevTableName := ""
 
 	for _, inv := range inventoryRows {
 		c.Logger.Debug("checking inventory for deletions", "vm_id", inv.VmId.String, "vm_uuid", inv.VmUuid.String, "name", inv.Name)
@@ -957,49 +982,7 @@ func (c *CronTask) captureHourlySnapshotForVcenter(ctx context.Context, startTim
 		slog.Warn("failed to insert vcenter totals", "vcenter", url, "snapshot_time", startTime.Unix(), "error", err)
 	}
 
-	// Compare with previous snapshot for this vcenter to mark deletions at snapshot time.
-	prevTableName, prevTableErr := latestHourlySnapshotBefore(ctx, dbConn, startTime)
-
-	if prevTableErr != nil {
-		c.Logger.Warn("failed to locate previous hourly snapshot for deletion comparison", "error", prevTableErr, "url", url)
-	}
-
-	prevSnapshotTime := int64(0)
-	if prevTableName != "" {
-		if suffix := strings.TrimPrefix(prevTableName, "inventory_hourly_"); suffix != prevTableName {
-			if ts, err := strconv.ParseInt(suffix, 10, 64); err == nil {
-				prevSnapshotTime = ts
-			}
-		}
-	}
-
-	if prevTableName != "" {
-		moreMissing := c.markMissingFromPrevious(ctx, dbConn, prevTableName, url, startTime, presentSnapshots, presentByUuid, presentByName, inventoryByVmID, inventoryByUuid, inventoryByName)
-		missingCount += moreMissing
-		// Guard against gaps: if previous snapshot is much older than expected, skip "new" detection to avoid false positives when an hourly run was missed.
-		expectedSeconds := int64(durationFromSeconds(c.Settings.Values.Settings.VcenterInventorySnapshotSeconds, time.Hour).Seconds())
-		if HasSnapshotGap(prevSnapshotTime, startTime.Unix(), expectedSeconds) {
-			c.Logger.Info("skipping new-VM detection due to gap between snapshots", "prev_table", prevTableName, "prev_snapshot_unix", prevSnapshotTime, "current_snapshot_unix", startTime.Unix())
-		} else {
-			newCount = countNewFromPrevious(ctx, dbConn, prevTableName, url, presentSnapshots)
-			if newCount > 0 {
-				newRows := listNewFromPrevious(ctx, dbConn, prevTableName, url, presentSnapshots)
-				names := make([]string, 0, len(newRows))
-				for _, r := range newRows {
-					if r.Name != "" {
-						names = append(names, r.Name)
-					} else if r.VmId.Valid {
-						names = append(names, r.VmId.String)
-					}
-				}
-				c.Logger.Info("new VMs since previous snapshot", "prev_table", prevTableName, "count", newCount, "names", names)
-			}
-		}
-		c.Logger.Debug("compared with previous snapshot", "prev_table", prevTableName, "new_since_prev", newCount, "missing_since_prev", missingCount)
-	} else {
-		// No previous snapshot found (or lookup failed).
-		newCount = len(presentSnapshots)
-	}
+	prevTableName, newCount, missingCount = c.compareWithPreviousSnapshot(ctx, dbConn, url, startTime, presentSnapshots, presentByUuid, presentByName, inventoryByVmID, inventoryByUuid, inventoryByName, missingCount)
 
 	// If VM count dropped versus totals and we still haven't marked missing, try another comparison + wider event window.
 	if missingCount == 0 && prevVmCount.Valid && prevVmCount.Int64 > int64(totals.VmCount) {
@@ -1100,3 +1083,63 @@ func (c *CronTask) captureHourlySnapshotForVcenter(ctx context.Context, startTim
 	}
 	return nil
 }
+
+// compareWithPreviousSnapshot cross-checks current vs. previous hourly snapshots:
+// marks deletions, detects new VMs when no gap exists, and returns the previous table name along with new/missing counts.
+func (c *CronTask) compareWithPreviousSnapshot(
+	ctx context.Context,
+	dbConn *sqlx.DB,
+	url string,
+	startTime time.Time,
+	presentSnapshots map[string]InventorySnapshotRow,
+	presentByUuid map[string]struct{},
+	presentByName map[string]struct{},
+	inventoryByVmID map[string]queries.Inventory,
+	inventoryByUuid map[string]queries.Inventory,
+	inventoryByName map[string]queries.Inventory,
+	missingCount int,
+) (string, int, int) {
+	prevTableName, prevTableErr := latestHourlySnapshotBefore(ctx, dbConn, startTime)
+	if prevTableErr != nil {
+		c.Logger.Warn("failed to locate previous hourly snapshot for deletion comparison", "error", prevTableErr, "url", url)
+	}
+
+	prevSnapshotTime := int64(0)
+	if prevTableName != "" {
+		if suffix := strings.TrimPrefix(prevTableName, "inventory_hourly_"); suffix != prevTableName {
+			if ts, err := strconv.ParseInt(suffix, 10, 64); err == nil {
+				prevSnapshotTime = ts
+			}
+		}
+	}
+
+	newCount := 0
+	if prevTableName != "" {
+		moreMissing := c.markMissingFromPrevious(ctx, dbConn, prevTableName, url, startTime, presentSnapshots, presentByUuid, presentByName, inventoryByVmID, inventoryByUuid, inventoryByName)
+		missingCount += moreMissing
+		expectedSeconds := int64(durationFromSeconds(c.Settings.Values.Settings.VcenterInventorySnapshotSeconds, time.Hour).Seconds())
+		// Allow runs as soon as half the normal interval; treat larger gaps as unreliable for "new" detection.
+		if HasSnapshotGap(prevSnapshotTime, startTime.Unix(), expectedSeconds/2) {
+			c.Logger.Info("skipping new-VM detection due to gap between snapshots", "prev_table", prevTableName, "prev_snapshot_unix", prevSnapshotTime, "current_snapshot_unix", startTime.Unix())
+		} else {
+			newCount = countNewFromPrevious(ctx, dbConn, prevTableName, url, presentSnapshots)
+			if newCount > 0 {
+				newRows := listNewFromPrevious(ctx, dbConn, prevTableName, url, presentSnapshots)
+				names := make([]string, 0, len(newRows))
+				for _, r := range newRows {
+					if r.Name != "" {
+						names = append(names, r.Name)
+					} else if r.VmId.Valid {
+						names = append(names, r.VmId.String)
+					}
+				}
+				c.Logger.Info("new VMs since previous snapshot", "prev_table", prevTableName, "count", newCount, "names", names)
+			}
+		}
+		c.Logger.Debug("compared with previous snapshot", "prev_table", prevTableName, "new_since_prev", newCount, "missing_since_prev", missingCount)
+	} else {
+		newCount = len(presentSnapshots)
+	}
+
+	return prevTableName, newCount, missingCount
+}