Skip to content

README.md

Latest commit

 

History

History
49 lines (33 loc) · 3.11 KB

README.md

File metadata and controls

49 lines (33 loc) · 3.11 KB
Raw

Audit Module

Retention and pruning of audit-style history tables. It archives old rows to JSON files and then deletes them from Postgres.

This module handles retention. Creating audit records happens elsewhere: ActivityLog (users), AdminAuditInterceptor (admin actions), TransactionAuditService (transactions), and blockchain audit in src/blockchain/audit. See docs/View_Audit_Logs.md.

Files

File Purpose
audit-pruning.service.ts Daily cron: archive then delete old history rows
audit-archive.config.ts Archive storage settings + production guard
audit.module.ts Registers the pruning service (no controller)

Cron job

Schedule Job
@Cron(CronExpression.DAILY) (00:00 server time) AuditPruningService.prune()

prune():

  1. Computes a cutoff of 365 days ago (hard-coded retentionDays).
  2. For transactionHistory and then loginHistory: selects rows with createdAt < cutoff, writes them to archives/<table>-<timestamp>.json, then deleteMany with the same filter.
  3. Tables with no expired rows are skipped (no empty files).

The archive directory is <repo>/archives (resolved from dist/audit/..) and is created if missing.

Configuration

Variable Default Purpose
AUDIT_ARCHIVE_STORAGE local local or s3
AUDIT_ARCHIVE_BUCKET (empty) S3 bucket when storage is s3
AUDIT_ARCHIVE_REGION us-east-1 S3 region

validateAuditStorageConfig() throws if NODE_ENV=production and storage is local, because archives on container disk are lost on redeploy (issue #1055).

Security model

  • No HTTP surface. Archive files contain full login history (IPs, user agents), so treat the archive location as sensitive and restrict access to it.

Gotchas

  • ⚠️ The S3 settings and production guard are not wired in yet. AuditPruningService always writes to local disk and never calls validateAuditStorageConfig(). In production, archived rows are deleted from the DB but only kept on ephemeral disk. Implement S3 upload (and call the guard at bootstrap) before relying on this for compliance.
  • findMany loads all expired rows into memory at once. The first run on a large, never-pruned database can be heavy; consider batching.
  • Archive and delete are not transactional. If the process dies between writeFile and deleteMany, the next run re-archives the same rows, which duplicates them but loses nothing.
  • Both files use // @ts-nocheck, so type errors here are not caught by npm run build.
  • The broader data archival strategy (issue #919) lives in src/archive/ and is exposed at /admin/archive/*.