fix(portal-bff): audit writes use raw INSERT (audit_writer has no SELECT for RETURNING) (#121)
## Summary #120 shipped the audit pipeline but the end-to-end path was never smoke-tested against a running Postgres. First click on `/auth/logout` returned 500 with the Pino log: ``` PostgresError code 42501 — permission denied for table events ``` Despite: - ACL on `audit.events` showing `audit_writer=a/audit_owner` (INSERT granted). - `has_table_privilege('audit_writer', 'audit.events', 'INSERT')` returning `t`. - `has_schema_privilege` / `has_type_privilege` all `t`. - A direct psql `INSERT INTO audit.events ...` after `SET LOCAL ROLE audit_writer` **succeeding**. - A psql `INSERT ... RETURNING id` after the same `SET LOCAL ROLE` **failing** with the exact same error. Root cause: Prisma's ORM `tx.auditEvent.create(...)` issues `INSERT ... RETURNING *` to hydrate the returned entity. Postgres requires **SELECT** on every column listed in `RETURNING`. `audit_writer` has INSERT only by ADR-0013 design — RETURNING fails with `code 42501` and the error message reads "permission denied for table events" (no mention of SELECT or RETURNING, which is what made it deeply non-obvious to diagnose). ## Fix `AuditWriter.recordEvent` now issues a parameterised raw INSERT via `tx.$executeRawUnsafe` instead of the ORM `create()`: ```ts await tx.$executeRawUnsafe( `INSERT INTO "audit"."events" (id, event_type, audience, outcome, subject, actor_id_hash, trace_id, payload) VALUES (gen_random_uuid(), $1, $2::"audit"."AuditAudience", $3::"audit"."AuditOutcome", $4, $5, $6, $7::jsonb)`, input.eventType, input.audience, input.outcome, input.subject ?? null, actorIdHash, traceId, payloadJson, ); ``` The role contract per ADR-0013 stays strict: `audit_writer` keeps INSERT only, no SELECT/UPDATE/DELETE/TRUNCATE. The other natural fix (`GRANT SELECT` to `audit_writer`) would have weakened the writer/reader role separation, so we deliberately went the other way. ## Notable choices **`gen_random_uuid()` server-side instead of Prisma's `@default(uuid())` client-side.** The model still declares `@default(uuid())` for any future ORM read or `audit_reader`-side query, but the write path uses the built-in Postgres function. No extension required (Postgres 13+). **Explicit enum and jsonb casts.** Parameters travel as TEXT over the wire; the SQL casts (`$2::"audit"."AuditAudience"`, `$7::jsonb`) ensure Postgres parses them as the right type. Without the casts, the type system rejects the INSERT before privilege check even fires. **Parameterised, not interpolated.** `$executeRawUnsafe` accepts a SQL template with `$1, $2, …` placeholders and a vararg of values — same wire-level parameter binding as a prepared statement, so SQL injection isn't possible even on caller-controlled inputs like `eventType`. The spec pins this with a malicious-input test. **Also fixes an env-sensitivity bug in `auth.controller.spec.ts`.** The test that asserts `session.absoluteExpiresAt == createdAt + 43200000` was reading the default via `readSessionTimeouts()` but didn't override `SESSION_ABSOLUTE_TIMEOUT_SECONDS`. If `apps/portal-bff/.env` has a custom value (as it did during the manual audit debugging), the test failed non-deterministically. Now the test deletes the env var before running and restores it after — same pattern as the other env-touching tests in this file. ## ADR amendment [ADR-0013](docs/decisions/0013-audit-trail-separated-postgres-append-only.md) §"Writer" now carries an **Implementation trap** callout explaining why Prisma's ORM `create()` cannot be used for audit writes (RETURNING requires SELECT, audit_writer has INSERT only). The corresponding Confirmation entry cross-references the callout. Two-commit shape on this PR (code + docs) — the squash-merge will fold them. ## Test plan - [x] `pnpm nx test portal-bff` (clean env) → **144/144 pass**. - [x] `pnpm nx lint portal-bff` → clean. - [x] `pnpm nx build portal-bff` → clean. - [x] Prettier-clean. - [ ] Manual smoke against running BFF + Postgres: - [ ] Sign in → row in `audit.events` with `event_type = 'auth.sign_in'`. - [ ] Sign out → row with `event_type = 'auth.sign_out'`. **The 500 from before is gone.** - [ ] Verify the role contract is still strict : ```sql SET ROLE audit_writer; SELECT * FROM audit.events LIMIT 1; -- should fail "permission denied" UPDATE audit.events SET event_type = 'x'; -- should fail DELETE FROM audit.events; -- should fail ``` --------- Co-authored-by: Julien Gautier <julien.gautier@apf.asso.fr> Reviewed-on: #121
This commit was merged in pull request #121.
This commit is contained in:
@@ -1,6 +1,7 @@
|
||||
---
|
||||
status: accepted
|
||||
date: 2026-04-29
|
||||
last-updated: 2026-05-13
|
||||
decision-makers: R&D Lead
|
||||
tags: [security, observability, data]
|
||||
---
|
||||
@@ -95,6 +96,8 @@ enum AuditOutcome {
|
||||
|
||||
**Writer.** A NestJS `AuditService` exposes a single typed method per event family. Inside a request, the service uses the same Prisma transaction as the business action where applicable; outside a request (background expirations, scheduled jobs), it uses a fresh transaction. Writer connection runs under the `audit_writer` role and has only `INSERT` on `audit.events` — any attempt to `UPDATE` or `DELETE` is rejected by Postgres regardless of code intent.
|
||||
|
||||
> **Implementation trap — Prisma ORM cannot be used for the write.** Prisma's `tx.auditEvent.create(...)` issues `INSERT … RETURNING *` to hydrate the entity it returns. Postgres requires the `SELECT` privilege on every column listed in `RETURNING`, and `audit_writer` has `INSERT` only by design — there is no `SELECT` grant on the writer role. The ORM path therefore fails at runtime with `PostgresError 42501 / "permission denied for table events"`, an error whose message mentions neither `SELECT` nor `RETURNING`. The write path uses **parameterised `$executeRawUnsafe`** with no `RETURNING` clause; the schema-level `id UUID @default(uuid())` from Prisma is replaced server-side with `gen_random_uuid()` in the SQL. This is a deliberate consequence of the role-separation contract and is pinned by a spec test. The alternative — granting `SELECT` on `audit.events` to `audit_writer` — would collapse the writer / reader role separation that the rest of this ADR rests on, so we go the other way.
|
||||
|
||||
**Events emitted in v1.**
|
||||
|
||||
| `event_type` | When | `outcome` |
|
||||
@@ -161,7 +164,7 @@ Hooks for **admin actions** and **sensitive data access** are designed-in: the w
|
||||
- `apps/portal-bff/prisma/schema.prisma` enables the `multiSchema` preview, declares the `audit` schema alongside `public`, and carries the `AuditEvent` model with `AuditAudience` (`workforce | customer`) and `AuditOutcome` (`success | failure | denied`) enums.
|
||||
- The migration `prisma/migrations/*_init_audit_schema/migration.sql` creates `audit.events`, `ALTER`s table + enum types to be owned by `audit_owner`, and re-applies the role grants explicitly: `INSERT` to `audit_writer`, `SELECT` to `audit_reader`, `SELECT, DELETE` to `audit_archiver` (SELECT is needed for archiver to evaluate the `created_at` predicate of "delete older than retention"). No grant of `UPDATE` or `TRUNCATE` to anyone — including the migrator's own login at runtime; only fresh schema migrations amend the table.
|
||||
- The roles themselves and the schema with default privileges are provisioned earlier by `infra/local/init/postgres/01-init.sql` (dev) — production replicates the same SQL via the future on-prem infrastructure ADR.
|
||||
- `apps/portal-bff/src/audit/audit.service.ts` exposes a single `AuditWriter.recordEvent(input)` method. Every write runs in a transaction whose first statement is `SET LOCAL ROLE audit_writer`, so the runtime contract holds even if the BFF connection is otherwise privileged. `trace_id` is auto-resolved from the active OTel span; `actor_id_hash` is read from CLS or accepted as an explicit override (placeholder until ADR-0009 / ADR-0010 land their guards). Failures propagate — no catch-and-swallow, per "blocking writes: no audit ⇒ no action".
|
||||
- `apps/portal-bff/src/audit/audit.service.ts` exposes a single `AuditWriter.recordEvent(input)` method. Every write runs in a transaction whose first statement is `SET LOCAL ROLE audit_writer`, so the runtime contract holds even if the BFF connection is otherwise privileged. The INSERT itself is a **parameterised `$executeRawUnsafe`**, not `tx.auditEvent.create(...)` — see the "Implementation trap" callout in the Writer section above for the RETURNING-requires-SELECT explanation. `trace_id` is auto-resolved from the active OTel span; `actor_id_hash` is read from CLS or accepted as an explicit override (placeholder until ADR-0009 / ADR-0010 land their guards). Failures propagate — no catch-and-swallow, per "blocking writes: no audit ⇒ no action".
|
||||
- BFF connects via the shared `DATABASE_URL` (the role switch is per-transaction). A separate `AUDIT_DATABASE_URL` connection pool is the production hardening, deferred — see "wired as features land" below.
|
||||
- Smoke-tested end to end against the local-dev Postgres: `audit_writer` INSERTs successfully, fails on `UPDATE` and `DELETE`; `audit_archiver` SELECTs + DELETEs successfully.
|
||||
|
||||
|
||||
Reference in New Issue
Block a user