Files
junhong_cmp_fiber/internal/application/auditarchive/service.go
break 88cc5e96ec
All checks were successful
构建并部署到测试环境(无 SSH) / build-and-deploy (push) Successful in 8m32s
暂存
2026-08-06 09:35:00 +08:00

413 lines
15 KiB
Go

// Package auditarchive 实现统一审计每日冷归档用例。
package auditarchive
import (
"compress/gzip"
"context"
"crypto/sha256"
"fmt"
"io"
"os"
"strconv"
"strings"
"time"
"github.com/bytedance/sonic"
"gorm.io/gorm"
"gorm.io/gorm/clause"
"github.com/break/junhong_cmp_fiber/internal/model"
"github.com/break/junhong_cmp_fiber/pkg/constants"
"github.com/break/junhong_cmp_fiber/pkg/storage"
)
const archivePageSize = 1000
// ObjectStore 是每日归档需要的最小对象存储能力。
type ObjectStore interface {
UploadWithMetadata(context.Context, string, io.Reader, string, map[string]string) error
Stat(context.Context, string) (*storage.ObjectMetadata, error)
Download(context.Context, string) (io.ReadCloser, error)
}
// RetentionAuditWriter 记录月度留存清理的系统审计事实。
type RetentionAuditWriter interface {
WriteRetentionCleanup(context.Context, *gorm.DB, RetentionAudit) error
}
// Service 编排审计归档生成、上传、复核和幂等账本更新。
type Service struct {
db *gorm.DB
store ObjectStore
audit RetentionAuditWriter
instanceID string
location *time.Location
}
// Manifest 是归档对象的完整性清单。
type Manifest struct {
SchemaVersion string `json:"schema_version"`
Source string `json:"source"`
ArchiveDate string `json:"archive_date"`
Timezone string `json:"timezone"`
RangeStart time.Time `json:"range_start"`
RangeEnd time.Time `json:"range_end"`
InstanceID string `json:"instance_id"`
EventCount int64 `json:"event_count"`
ResourceCount int64 `json:"resource_count"`
UncompressedBytes int64 `json:"uncompressed_bytes"`
CompressedBytes int64 `json:"compressed_bytes"`
ObjectKey string `json:"object_key"`
SHA256 string `json:"sha256"`
Revision int `json:"revision"`
GeneratedAt time.Time `json:"generated_at"`
Status string `json:"status"`
}
type archiveLine struct {
Event model.AuditEvent `json:"event"`
Resources []model.AuditEventResource `json:"resources"`
}
type archiveFile struct {
path string
eventCount int64
resourceCount int64
uncompressedBytes int64
compressedBytes int64
sha256 string
}
// NewService 创建统一审计每日冷归档服务。
func NewService(db *gorm.DB, store ObjectStore, instanceID string, audit ...RetentionAuditWriter) (*Service, error) {
location, err := time.LoadLocation(constants.AuditArchiveTimezone)
if err != nil {
return nil, fmt.Errorf("加载审计归档时区失败: %w", err)
}
if strings.TrimSpace(instanceID) == "" {
instanceID = "audit-archive"
}
var auditWriter RetentionAuditWriter
if len(audit) > 0 {
auditWriter = audit[0]
}
return &Service{db: db, store: store, audit: auditWriter, instanceID: instanceID, location: location}, nil
}
// ArchivePreviousDay 归档 Asia/Shanghai 前一完整自然日。
func (s *Service) ArchivePreviousDay(ctx context.Context) error {
now := time.Now().In(s.location)
return s.ArchiveDate(ctx, now.AddDate(0, 0, -1))
}
// ArchiveDate 归档指定 Asia/Shanghai 自然日。
func (s *Service) ArchiveDate(ctx context.Context, archiveDate time.Time) error {
if s.db == nil || s.store == nil {
return fmt.Errorf("审计归档数据库或对象存储未配置")
}
start := time.Date(archiveDate.In(s.location).Year(), archiveDate.In(s.location).Month(), archiveDate.In(s.location).Day(), 0, 0, 0, 0, s.location)
end := start.AddDate(0, 0, 1)
today := time.Now().In(s.location)
todayStart := time.Date(today.Year(), today.Month(), today.Day(), 0, 0, 0, 0, s.location)
if end.After(todayStart) {
return fmt.Errorf("统一审计只能归档已经结束的完整自然日")
}
run, err := s.ensureRun(ctx, start, end)
if err != nil {
return err
}
if run.Status == constants.ArchiveStatusSuccess {
valid, validateErr := s.validateSuccessfulRun(ctx, run)
if validateErr == nil && valid {
return nil
}
}
acquired, err := s.acquireRun(ctx, run)
if err != nil || !acquired {
return err
}
if err := s.execute(ctx, run); err != nil {
s.markFailed(ctx, run.ID, err)
return err
}
return nil
}
func (s *Service) ensureRun(ctx context.Context, start, end time.Time) (*model.LogArchiveRun, error) {
run := model.LogArchiveRun{
Source: constants.AuditArchiveSource, ArchiveDate: start, InstanceID: s.instanceID,
SchemaVersion: constants.AuditArchiveSchemaVersion, Revision: 1,
Status: constants.ArchiveStatusPending, RangeStart: start, RangeEnd: end,
}
result := s.db.WithContext(ctx).Clauses(clause.OnConflict{
Columns: []clause.Column{{Name: "source"}, {Name: "archive_date"}, {Name: "instance_id"}, {Name: "schema_version"}},
DoNothing: true,
}).Create(&run)
if result.Error != nil {
return nil, fmt.Errorf("创建审计归档账本失败: %w", result.Error)
}
if result.RowsAffected == 0 {
if err := s.db.WithContext(ctx).Where(
"source = ? AND archive_date = ? AND instance_id = ? AND schema_version = ?",
constants.AuditArchiveSource, start, s.instanceID, constants.AuditArchiveSchemaVersion,
).First(&run).Error; err != nil {
return nil, fmt.Errorf("读取审计归档账本失败: %w", err)
}
}
return &run, nil
}
func (s *Service) acquireRun(ctx context.Context, run *model.LogArchiveRun) (bool, error) {
revision := run.Revision
if run.Status == constants.ArchiveStatusFailed || run.Status == constants.ArchiveStatusSuccess || run.Status == constants.ArchiveStatusRunning {
revision++
}
now := time.Now()
staleBefore := now.Add(-3 * time.Hour)
result := s.db.WithContext(ctx).Model(&model.LogArchiveRun{}).
Where("id = ? AND (status <> ? OR updated_at < ?)", run.ID, constants.ArchiveStatusRunning, staleBefore).
Updates(map[string]any{
"status": constants.ArchiveStatusRunning, "revision": revision,
"attempt_count": gorm.Expr("attempt_count + 1"), "error_summary": "",
"completed_at": nil, "updated_at": now,
})
if result.Error != nil {
return false, fmt.Errorf("锁定审计归档任务失败: %w", result.Error)
}
if result.RowsAffected == 0 {
return false, nil
}
run.Revision = revision
run.Status = constants.ArchiveStatusRunning
return true, nil
}
func (s *Service) execute(ctx context.Context, run *model.LogArchiveRun) error {
file, err := s.buildArchiveFile(ctx, run.RangeStart, run.RangeEnd)
if err != nil {
return err
}
defer os.Remove(file.path)
dbEvents, dbResources, err := s.databaseCounts(ctx, run.RangeStart, run.RangeEnd)
if err != nil {
return err
}
if dbEvents != file.eventCount || dbResources != file.resourceCount {
return fmt.Errorf("审计归档生成期间数据数量发生变化")
}
objectKey, manifestKey := objectKeys(run.RangeStart, run.Revision)
metadata := archiveMetadata(file, run)
reader, err := os.Open(file.path)
if err != nil {
return fmt.Errorf("打开审计归档临时文件失败: %w", err)
}
uploadErr := s.store.UploadWithMetadata(ctx, objectKey, reader, "application/gzip", metadata)
closeErr := reader.Close()
if uploadErr != nil {
return fmt.Errorf("上传审计归档对象失败: %w", uploadErr)
}
if closeErr != nil {
return fmt.Errorf("关闭审计归档临时文件失败: %w", closeErr)
}
if err := s.verifyObject(ctx, objectKey, file.compressedBytes, metadata); err != nil {
return err
}
generatedAt := time.Now().In(s.location)
manifest := Manifest{
SchemaVersion: constants.AuditArchiveSchemaVersion, Source: constants.AuditArchiveSource,
ArchiveDate: run.RangeStart.In(s.location).Format(time.DateOnly), Timezone: constants.AuditArchiveTimezone,
RangeStart: run.RangeStart, RangeEnd: run.RangeEnd, InstanceID: s.instanceID,
EventCount: file.eventCount, ResourceCount: file.resourceCount,
UncompressedBytes: file.uncompressedBytes, CompressedBytes: file.compressedBytes,
ObjectKey: objectKey, SHA256: file.sha256, Revision: run.Revision,
GeneratedAt: generatedAt, Status: constants.ArchiveStatusSuccess,
}
manifestBytes, err := sonic.Marshal(manifest)
if err != nil {
return fmt.Errorf("序列化审计归档清单失败: %w", err)
}
manifestMetadata := map[string]string{"source": constants.AuditArchiveSource, "data-sha256": file.sha256, "revision": strconv.Itoa(run.Revision)}
if err := s.store.UploadWithMetadata(ctx, manifestKey, strings.NewReader(string(manifestBytes)), "application/json", manifestMetadata); err != nil {
return fmt.Errorf("上传审计归档清单失败: %w", err)
}
if err := s.verifyObject(ctx, manifestKey, int64(len(manifestBytes)), manifestMetadata); err != nil {
return err
}
completedAt := time.Now()
updates := map[string]any{
"status": constants.ArchiveStatusSuccess, "object_key": objectKey, "manifest_key": manifestKey,
"event_count": file.eventCount, "resource_count": file.resourceCount,
"uncompressed_bytes": file.uncompressedBytes, "compressed_bytes": file.compressedBytes,
"sha256": file.sha256, "generated_at": generatedAt, "completed_at": completedAt,
"updated_at": completedAt,
}
if err := s.db.WithContext(ctx).Model(&model.LogArchiveRun{}).Where("id = ?", run.ID).Updates(updates).Error; err != nil {
return fmt.Errorf("更新审计归档成功账本失败: %w", err)
}
return nil
}
func (s *Service) buildArchiveFile(ctx context.Context, start, end time.Time) (*archiveFile, error) {
temp, err := os.CreateTemp("", "audit-events-*.jsonl.gz")
if err != nil {
return nil, fmt.Errorf("创建审计归档临时文件失败: %w", err)
}
path := temp.Name()
failed := true
defer func() {
_ = temp.Close()
if failed {
_ = os.Remove(path)
}
}()
hasher := sha256.New()
gzipWriter := gzip.NewWriter(io.MultiWriter(temp, hasher))
result := &archiveFile{path: path}
var lastID uint
for {
var events []model.AuditEvent
if err := s.db.WithContext(ctx).Where("created_at >= ? AND created_at < ? AND id > ?", start, end, lastID).
Order("id ASC").Limit(archivePageSize).Find(&events).Error; err != nil {
return nil, fmt.Errorf("读取审计归档事件失败: %w", err)
}
if len(events) == 0 {
break
}
ids := make([]uint, 0, len(events))
for i := range events {
ids = append(ids, events[i].ID)
}
var resources []model.AuditEventResource
if err := s.db.WithContext(ctx).Where("audit_event_id IN ?", ids).
Order("audit_event_id ASC, sort_order ASC, id ASC").Find(&resources).Error; err != nil {
return nil, fmt.Errorf("读取审计归档资源失败: %w", err)
}
grouped := make(map[uint][]model.AuditEventResource, len(events))
for i := range resources {
resource := resources[i]
grouped[resource.AuditEventID] = append(grouped[resource.AuditEventID], resource)
}
for i := range events {
eventResources := grouped[events[i].ID]
if eventResources == nil {
eventResources = []model.AuditEventResource{}
}
line, marshalErr := sonic.Marshal(archiveLine{Event: events[i], Resources: eventResources})
if marshalErr != nil {
return nil, fmt.Errorf("序列化审计归档事件失败: %w", marshalErr)
}
line = append(line, '\n')
if _, writeErr := gzipWriter.Write(line); writeErr != nil {
return nil, fmt.Errorf("写入审计归档压缩流失败: %w", writeErr)
}
result.eventCount++
result.resourceCount += int64(len(grouped[events[i].ID]))
result.uncompressedBytes += int64(len(line))
}
lastID = events[len(events)-1].ID
}
if err := gzipWriter.Close(); err != nil {
return nil, fmt.Errorf("关闭审计归档压缩流失败: %w", err)
}
if err := temp.Close(); err != nil {
return nil, fmt.Errorf("关闭审计归档临时文件失败: %w", err)
}
info, err := os.Stat(path)
if err != nil {
return nil, fmt.Errorf("读取审计归档临时文件信息失败: %w", err)
}
result.compressedBytes = info.Size()
result.sha256 = fmt.Sprintf("%x", hasher.Sum(nil))
failed = false
return result, nil
}
func (s *Service) databaseCounts(ctx context.Context, start, end time.Time) (int64, int64, error) {
var eventCount int64
if err := s.db.WithContext(ctx).Model(&model.AuditEvent{}).
Where("created_at >= ? AND created_at < ?", start, end).Count(&eventCount).Error; err != nil {
return 0, 0, fmt.Errorf("统计审计归档事件失败: %w", err)
}
var resourceCount int64
subquery := s.db.Model(&model.AuditEvent{}).Select("id").Where("created_at >= ? AND created_at < ?", start, end)
if err := s.db.WithContext(ctx).Model(&model.AuditEventResource{}).
Where("audit_event_id IN (?)", subquery).Count(&resourceCount).Error; err != nil {
return 0, 0, fmt.Errorf("统计审计归档资源失败: %w", err)
}
return eventCount, resourceCount, nil
}
func (s *Service) validateSuccessfulRun(ctx context.Context, run *model.LogArchiveRun) (bool, error) {
events, resources, err := s.databaseCounts(ctx, run.RangeStart, run.RangeEnd)
if err != nil || events != run.EventCount || resources != run.ResourceCount {
return false, err
}
metadata := map[string]string{
"sha256": run.SHA256, "event-count": strconv.FormatInt(run.EventCount, 10),
"resource-count": strconv.FormatInt(run.ResourceCount, 10), "revision": strconv.Itoa(run.Revision),
}
if err := s.verifyObject(ctx, run.ObjectKey, run.CompressedBytes, metadata); err != nil {
return false, nil
}
manifestMetadata := map[string]string{
"source": constants.AuditArchiveSource, "data-sha256": run.SHA256, "revision": strconv.Itoa(run.Revision),
}
if err := s.verifyObject(ctx, run.ManifestKey, -1, manifestMetadata); err != nil {
return false, nil
}
return true, nil
}
func (s *Service) verifyObject(ctx context.Context, key string, expectedSize int64, expectedMetadata map[string]string) error {
object, err := s.store.Stat(ctx, key)
if err != nil {
return fmt.Errorf("复核归档对象 metadata 失败: %w", err)
}
if expectedSize >= 0 && object.Size != expectedSize {
return fmt.Errorf("归档对象大小复核失败")
}
for name, value := range expectedMetadata {
if object.Metadata[strings.ToLower(name)] != value {
return fmt.Errorf("归档对象 metadata 字段 %s 复核失败", name)
}
}
return nil
}
func (s *Service) markFailed(ctx context.Context, runID uint, archiveErr error) {
summary := []rune(archiveErr.Error())
if len(summary) > 500 {
summary = summary[:500]
}
failedCtx, cancel := context.WithTimeout(context.WithoutCancel(ctx), 5*time.Second)
defer cancel()
_ = s.db.WithContext(failedCtx).Model(&model.LogArchiveRun{}).Where("id = ?", runID).Updates(map[string]any{
"status": constants.ArchiveStatusFailed, "error_summary": string(summary), "updated_at": time.Now(),
}).Error
}
func objectKeys(date time.Time, revision int) (string, string) {
prefix := fmt.Sprintf("audit-archive/v1/%04d/%02d/%02d", date.Year(), date.Month(), date.Day())
name := fmt.Sprintf("audit-events-%s-r%d", date.Format(time.DateOnly), revision)
return prefix + "/" + name + ".jsonl.gz", prefix + "/" + name + ".manifest.json"
}
func archiveMetadata(file *archiveFile, run *model.LogArchiveRun) map[string]string {
return map[string]string{
"schema-version": constants.AuditArchiveSchemaVersion,
"source": constants.AuditArchiveSource,
"archive-date": run.RangeStart.Format(time.DateOnly),
"timezone": constants.AuditArchiveTimezone,
"event-count": strconv.FormatInt(file.eventCount, 10),
"resource-count": strconv.FormatInt(file.resourceCount, 10),
"sha256": file.sha256,
"revision": strconv.Itoa(run.Revision),
}
}