Files
ai-gateway-go/internal/audit/maintenance.go
T
superidou 9501751792 0.10.1: 安全与业务逻辑加固、新品牌与部署加固
三轮审查修复(60+ 项),相对远端 main(b536672)的关键变更:
- 安全: 数据面 SSRF 拨号防护(防 DNS rebinding)/上游凭据剥离/登录防枚举
  与锁定态统一/可信代理(X-Forwarded-For)限流加固/会话版本失效机制/
  撤销即时传播/弱密钥拒绝启动/脱敏字节级重写(保签名契约)
- 业务逻辑: 裸 body 上传 panic/bootstrap 审计管线卡死/定价通配符优先级/
  全局工具可见性/调度器停机补跑/TOTP 挑战令牌消费顺序/熔断探针语义/
  >4MB 响应 token 计量/管理员重置密码作废会话 等
- 前端: 新 logo(语枢 AI 网关主题)/Provider 凭据异常警示/删除入口/
  后端错误消息透传/localStorage 敏感数据收敛
- 部署: CREDENTIAL_MASTER_KEY 持久化与弱值拒绝/Provider DELETE 接口/
  nginx 安全头/worker 内存限制
- 新增迁移 000029(key_hash 索引)/000030(usage_daily 币种维度)
2026-08-13 10:50:51 +08:00

159 lines
6.2 KiB
Go

package audit
import (
"context"
"errors"
"fmt"
"regexp"
"sort"
"time"
"github.com/jackc/pgx/v5"
"github.com/jackc/pgx/v5/pgxpool"
)
const maintenanceLockID int64 = 6720240816
var auditPartitionPattern = regexp.MustCompile(`^audit_events_(\d{4})(\d{2})$`)
type MaintenanceResult struct {
CreatedPartitions []string `json:"created_partitions"`
DroppedPartitions []string `json:"dropped_partitions"`
DeletedAuditRows int64 `json:"deleted_audit_rows"`
DeletedUsageRows int64 `json:"deleted_usage_rows"`
DeletedTraceRows int64 `json:"deleted_trace_rows"`
}
type Maintenance struct {
pool *pgxpool.Pool
auditRetention time.Duration
usageRetention time.Duration
traceRetention time.Duration
monthsAhead int
}
func NewMaintenance(pool *pgxpool.Pool, auditRetention, usageRetention, traceRetention time.Duration, monthsAhead int) *Maintenance {
return &Maintenance{pool: pool, auditRetention: auditRetention, usageRetention: usageRetention, traceRetention: traceRetention, monthsAhead: monthsAhead}
}
func (m *Maintenance) Run(ctx context.Context, now time.Time) (MaintenanceResult, error) {
var result MaintenanceResult
if m == nil || m.pool == nil {
return result, errors.New("audit maintenance store unavailable")
}
now = now.UTC()
auditCutoff := now.Add(-m.auditRetention)
usageCutoff := now.Add(-m.usageRetention)
tx, err := m.pool.Begin(ctx)
if err != nil {
return result, fmt.Errorf("begin audit maintenance: %w", err)
}
defer func() { _ = tx.Rollback(ctx) }()
if _, err := tx.Exec(ctx, `SELECT pg_advisory_xact_lock($1)`, maintenanceLockID); err != nil {
return result, fmt.Errorf("lock audit maintenance: %w", err)
}
if _, err := tx.Exec(ctx, `LOCK TABLE gateway.audit_events IN ACCESS EXCLUSIVE MODE`); err != nil {
return result, fmt.Errorf("lock audit table: %w", err)
}
partitions, err := listAuditPartitions(ctx, tx)
if err != nil {
return result, err
}
for name := range partitions {
start, ok := auditPartitionMonth(name)
if !ok || start.AddDate(0, 1, 0).After(auditCutoff) {
continue
}
if _, err := tx.Exec(ctx, `DROP TABLE `+pgx.Identifier{"gateway", name}.Sanitize()); err != nil {
return result, fmt.Errorf("drop audit partition %s: %w", name, err)
}
result.DroppedPartitions = append(result.DroppedPartitions, name)
}
if _, err := tx.Exec(ctx, `ALTER TABLE gateway.audit_events DETACH PARTITION gateway.audit_events_default`); err != nil {
return result, fmt.Errorf("detach audit default partition: %w", err)
}
deleted, err := tx.Exec(ctx, `DELETE FROM gateway.audit_events_default WHERE recorded_at < $1`, auditCutoff)
if err != nil {
return result, fmt.Errorf("clean audit default partition: %w", err)
}
result.DeletedAuditRows += deleted.RowsAffected()
from := monthStart(auditCutoff)
through := monthStart(now).AddDate(0, m.monthsAhead+1, 0)
for start := from; start.Before(through); start = start.AddDate(0, 1, 0) {
end := start.AddDate(0, 1, 0)
name := "audit_events_" + start.Format("200601")
if _, exists := partitions[name]; !exists {
statement := fmt.Sprintf(`CREATE TABLE %s PARTITION OF gateway.audit_events FOR VALUES FROM ('%s') TO ('%s')`, pgx.Identifier{"gateway", name}.Sanitize(), start.Format(time.RFC3339), end.Format(time.RFC3339))
if _, err := tx.Exec(ctx, statement); err != nil {
return result, fmt.Errorf("create audit partition %s: %w", name, err)
}
result.CreatedPartitions = append(result.CreatedPartitions, name)
}
statement := `WITH moved AS (DELETE FROM gateway.audit_events_default WHERE recorded_at >= $1 AND recorded_at < $2 RETURNING *) INSERT INTO gateway.audit_events SELECT * FROM moved`
if _, err := tx.Exec(ctx, statement, start, end); err != nil {
return result, fmt.Errorf("move default audit rows into %s: %w", name, err)
}
}
if _, err := tx.Exec(ctx, `ALTER TABLE gateway.audit_events ATTACH PARTITION gateway.audit_events_default DEFAULT`); err != nil {
return result, fmt.Errorf("reattach audit default partition: %w", err)
}
deleted, err = tx.Exec(ctx, `DELETE FROM gateway.audit_events WHERE recorded_at < $1`, auditCutoff)
if err != nil {
return result, fmt.Errorf("apply exact audit retention: %w", err)
}
result.DeletedAuditRows += deleted.RowsAffected()
deleted, err = tx.Exec(ctx, `DELETE FROM gateway.usage_daily WHERE usage_date < $1::date`, usageCutoff.Format("2006-01-02"))
if err != nil {
return result, fmt.Errorf("apply usage retention: %w", err)
}
result.DeletedUsageRows = deleted.RowsAffected()
// M9 Trace 保留:agent_trace_spans 通过 ON DELETE CASCADE 一并清理,
// 防止 trace 表无界增长(每条应用/数字员工请求都会写 trace)。
if m.traceRetention > 0 {
deleted, err = tx.Exec(ctx, `DELETE FROM gateway.agent_traces WHERE started_at < $1`, now.Add(-m.traceRetention))
if err != nil {
return result, fmt.Errorf("apply trace retention: %w", err)
}
result.DeletedTraceRows = deleted.RowsAffected()
}
if err := tx.Commit(ctx); err != nil {
return result, fmt.Errorf("commit audit maintenance: %w", err)
}
sort.Strings(result.CreatedPartitions)
sort.Strings(result.DroppedPartitions)
return result, nil
}
func listAuditPartitions(ctx context.Context, tx pgx.Tx) (map[string]struct{}, error) {
rows, err := tx.Query(ctx, `SELECT child.relname FROM pg_inherits i JOIN pg_class parent ON parent.oid=i.inhparent JOIN pg_namespace n ON n.oid=parent.relnamespace JOIN pg_class child ON child.oid=i.inhrelid WHERE n.nspname='gateway' AND parent.relname='audit_events'`)
if err != nil {
return nil, fmt.Errorf("list audit partitions: %w", err)
}
defer rows.Close()
result := make(map[string]struct{})
for rows.Next() {
var name string
if err := rows.Scan(&name); err != nil {
return nil, fmt.Errorf("scan audit partition: %w", err)
}
if name != "audit_events_default" {
result[name] = struct{}{}
}
}
return result, rows.Err()
}
func auditPartitionMonth(name string) (time.Time, bool) {
match := auditPartitionPattern.FindStringSubmatch(name)
if match == nil {
return time.Time{}, false
}
parsed, err := time.Parse("200601", match[1]+match[2])
return parsed.UTC(), err == nil
}
func monthStart(value time.Time) time.Time {
value = value.UTC()
return time.Date(value.Year(), value.Month(), 1, 0, 0, 0, 0, time.UTC)
}