trails/openspec/specs/infrastructure/spec.md
Ullrich Schäfer b6908fb288
Fix picomatch audit vulnerability and complete security hardening verification
Override picomatch to 4.0.4 to resolve high-severity ReDoS (GHSA-c2c7-rcm5-vvqj),
verify all security hardening tasks (gitleaks, pnpm audit, Docker non-root,
Caddy headers, scanner blocking), sync specs, and archive the change.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
2026-03-26 00:31:27 +01:00

4.5 KiB

ADDED Requirements

Requirement: Terraform Hetzner provisioning

Infrastructure SHALL be provisioned on Hetzner Cloud using Terraform with the Hetzner provider.

Scenario: Provision server

  • WHEN terraform apply is run
  • THEN a Hetzner CX21 server (2 vCPU, 4 GB RAM, 40 GB SSD) is created with Docker installed

Requirement: Docker Compose deployment

All services SHALL be deployed via Docker Compose on the Hetzner server.

Scenario: Start all services

  • WHEN docker compose up -d is run on the server
  • THEN the Journal, Planner, BRouter, PostgreSQL, and Garage containers start and are reachable

Requirement: Service configuration

Each service SHALL be configured via environment variables defined in Docker Compose, with security best practices including non-root execution and security headers.

Scenario: Journal configuration

  • WHEN the Journal container starts
  • THEN it reads DOMAIN, DATABASE_URL, PLANNER_URL, S3_ENDPOINT, and S3_BUCKET from environment variables

Scenario: Planner configuration

  • WHEN the Planner container starts
  • THEN it reads BROUTER_URL and DATABASE_URL from environment variables

Scenario: Caddy security headers

  • WHEN Caddy proxies a request
  • THEN it adds HSTS, X-Content-Type-Options, X-Frame-Options, Referrer-Policy, and Permissions-Policy headers

Scenario: Caddy scanner blocking

  • WHEN a request matches known scanner paths (.env, .git, wp-config, etc.)
  • THEN Caddy returns 403 without forwarding to the application

Requirement: PostgreSQL with PostGIS

The database SHALL be PostgreSQL with the PostGIS extension for spatial queries.

Scenario: PostGIS available

  • WHEN the PostgreSQL container starts
  • THEN the PostGIS extension is available and can be enabled with CREATE EXTENSION postgis

Requirement: BRouter segment management

The infrastructure SHALL support downloading and updating Germany RD5 segments from brouter.de.

Scenario: Download segments

  • WHEN the segment download script is run
  • THEN Germany RD5 files (E5_N45, E5_N50, E10_N45, E10_N50) are downloaded to the segments volume

Scenario: Weekly segment update

  • WHEN the weekly cron job runs
  • THEN RD5 segments are updated from brouter.de and the BRouter container is restarted

Requirement: CI/CD pipeline

GitHub Actions SHALL build, deploy, and security-scan both apps on push to main.

Scenario: Push triggers deployment

  • WHEN code is pushed to the main branch
  • THEN GitHub Actions builds Docker images, pushes to ghcr.io/trails-cool/, and deploys to the Hetzner server

Scenario: Gitleaks scan

  • WHEN a PR is opened
  • THEN gitleaks scans for committed secrets

Scenario: Dependency audit

  • WHEN CI runs
  • THEN pnpm audit checks for high/critical vulnerabilities

Requirement: Backup strategy

The infrastructure SHALL include daily backups of the PostgreSQL database.

Scenario: Daily backup

  • WHEN the daily backup cron runs
  • THEN a PostgreSQL dump is uploaded to the Hetzner Storage Box

Requirement: Domain and TLS

The infrastructure SHALL configure DNS and TLS for trails.cool and planner.trails.cool.

Scenario: HTTPS access

  • WHEN a user navigates to https://trails.cool
  • THEN the connection is secured with a valid TLS certificate

Requirement: Sentry error tracking

The system SHALL enrich Sentry events with user and session context, use route-aware tracing, and prevent source maps from being served to clients.

Scenario: Journal error includes user context

  • WHEN an authenticated Journal user triggers an error
  • THEN the Sentry event SHALL include the user's ID and username

Scenario: Journal error without user context

  • WHEN an unauthenticated visitor triggers an error
  • THEN the Sentry event SHALL have no user context (Sentry.setUser(null))

Scenario: Planner error includes session ID

  • WHEN an error occurs during a Planner session
  • THEN the Sentry event SHALL include a session_id tag with the active session ID

Scenario: Route-level performance traces

  • WHEN a user navigates between routes in either app
  • THEN Sentry SHALL create a transaction span named after the route pattern (e.g., /routes/:id)

Scenario: Source maps not served to clients

  • WHEN a client requests a .map file from the production server
  • THEN the server SHALL return 404 (source maps are uploaded to Sentry during build, not shipped in the bundle)