refactor(ops-cleanup): 拆分 executor + table-driven + 提取常量 + 补测试

代码审查反馈： 1. 文件行数超标：ops_cleanup_service.go 594→413 行。拆 opsCleanupPlan / deleteOldRowsByID / truncateOpsTable / isMissingRelationError + counts struct 到 ops_cleanup_executor.go (164 行)。 2. runCleanupOnce 89 行→30 行（table-driven）：用 []opsCleanupTarget 循环替代三组重复的 opsCleanupPlan → runOne → assign。 3. 魔法值提取常量： opsCleanupDefaultSchedule / opsCleanupBatchSize / opsCleanupCronStopTimeout / opsCleanupRunTimeout / opsCleanupHeartbeatTimeout。 ops_settings.go 中 "0 2 * * *" 也统一引用 opsCleanupDefaultSchedule。 4. 补 5 个缺失测试： - Reload 未 Start 时 no-op - Reload 已 Stop 后 no-op - cleanupReloader==nil 时 Update 不 panic - Start 重复调用幂等 - refreshEffectiveBeforeRun 正确更新 snapshot
2026-05-04 13:35:26 +08:00
parent c4598aa9b6
commit d218b6c2aa
4 changed files with 248 additions and 207 deletions
@@ -0,0 +1,164 @@
+package service
+
+import (
+	"context"
+	"database/sql"
+	"fmt"
+	"strings"
+	"time"
+)
+
+const (
+	opsCleanupDefaultSchedule  = "0 2 * * *"
+	opsCleanupBatchSize        = 5000
+	opsCleanupCronStopTimeout  = 3 * time.Second
+	opsCleanupRunTimeout       = 30 * time.Minute
+	opsCleanupHeartbeatTimeout = 2 * time.Second
+)
+
+type opsCleanupTarget struct {
+	retentionDays int
+	table         string
+	timeCol       string
+	castDate      bool
+	counter       *int64
+}
+
+type opsCleanupDeletedCounts struct {
+	errorLogs     int64
+	retryAttempts int64
+	alertEvents   int64
+	systemLogs    int64
+	logAudits     int64
+	systemMetrics int64
+	hourlyPreagg  int64
+	dailyPreagg   int64
+}
+
+func (c opsCleanupDeletedCounts) String() string {
+	return fmt.Sprintf(
+		"error_logs=%d retry_attempts=%d alert_events=%d system_logs=%d log_audits=%d system_metrics=%d hourly_preagg=%d daily_preagg=%d",
+		c.errorLogs,
+		c.retryAttempts,
+		c.alertEvents,
+		c.systemLogs,
+		c.logAudits,
+		c.systemMetrics,
+		c.hourlyPreagg,
+		c.dailyPreagg,
+	)
+}
+
+// opsCleanupPlan 把"保留天数"翻译成具体的清理动作。
+//   - days < 0  → 跳过该项清理（ok=false），保留兼容老数据
+//   - days == 0 → TRUNCATE TABLE（O(1) 全清），truncate=true
+//   - days > 0  → 批量 DELETE 早于 now-N天 的行，cutoff = now - N 天
+func opsCleanupPlan(now time.Time, days int) (cutoff time.Time, truncate, ok bool) {
+	if days < 0 {
+		return time.Time{}, false, false
+	}
+	if days == 0 {
+		return time.Time{}, true, true
+	}
+	return now.AddDate(0, 0, -days), false, true
+}
+
+func opsCleanupRunOne(
+	ctx context.Context,
+	db *sql.DB,
+	truncate bool,
+	cutoff time.Time,
+	table, timeCol string,
+	castDate bool,
+	batchSize int,
+) (int64, error) {
+	if truncate {
+		return truncateOpsTable(ctx, db, table)
+	}
+	return deleteOldRowsByID(ctx, db, table, timeCol, cutoff, batchSize, castDate)
+}
+
+func deleteOldRowsByID(
+	ctx context.Context,
+	db *sql.DB,
+	table string,
+	timeColumn string,
+	cutoff time.Time,
+	batchSize int,
+	castCutoffToDate bool,
+) (int64, error) {
+	if db == nil {
+		return 0, nil
+	}
+	if batchSize <= 0 {
+		batchSize = opsCleanupBatchSize
+	}
+
+	where := fmt.Sprintf("%s < $1", timeColumn)
+	if castCutoffToDate {
+		where = fmt.Sprintf("%s < $1::date", timeColumn)
+	}
+
+	q := fmt.Sprintf(`
+WITH batch AS (
+  SELECT id FROM %s
+  WHERE %s
+  ORDER BY id
+  LIMIT $2
+)
+DELETE FROM %s
+WHERE id IN (SELECT id FROM batch)
+`, table, where, table)
+
+	var total int64
+	for {
+		res, err := db.ExecContext(ctx, q, cutoff, batchSize)
+		if err != nil {
+			if isMissingRelationError(err) {
+				return total, nil
+			}
+			return total, err
+		}
+		affected, err := res.RowsAffected()
+		if err != nil {
+			return total, err
+		}
+		total += affected
+		if affected == 0 {
+			break
+		}
+	}
+	return total, nil
+}
+
+// truncateOpsTable 用 TRUNCATE TABLE 清空指定表，先 SELECT COUNT(*) 取得清空前行数用于 heartbeat。
+func truncateOpsTable(ctx context.Context, db *sql.DB, table string) (int64, error) {
+	if db == nil {
+		return 0, nil
+	}
+	var count int64
+	if err := db.QueryRowContext(ctx, fmt.Sprintf("SELECT COUNT(*) FROM %s", table)).Scan(&count); err != nil {
+		if isMissingRelationError(err) {
+			return 0, nil
+		}
+		return 0, fmt.Errorf("count %s: %w", table, err)
+	}
+	if count == 0 {
+		return 0, nil
+	}
+	if _, err := db.ExecContext(ctx, fmt.Sprintf("TRUNCATE TABLE %s", table)); err != nil {
+		if isMissingRelationError(err) {
+			return 0, nil
+		}
+		return 0, fmt.Errorf("truncate %s: %w", table, err)
+	}
+	return count, nil
+}
+
+func isMissingRelationError(err error) bool {
+	if err == nil {
+		return false
+	}
+	s := strings.ToLower(err.Error())
+	return strings.Contains(s, "does not exist") && strings.Contains(s, "relation")
+}