sbom-nexus/docs/operator-guide.md
tegwick 879012c776
All checks were successful
Build and Publish Container Image / build-and-push (push) Successful in 38s
feat: add controlled source ingestion and replay
Assistant: codex
Assistant-Model: gpt-5.6-sol
Assistant-Session: 01a02b22-9638-76d2-bbff-b7ea1770b118
2026-08-22 23:57:37 +02:00

4.6 KiB

Operator guide

Run locally

make install
make run

The API defaults to 127.0.0.1:8010 and ./sbom-nexus.db. SQLite schema is created automatically for local development.

For PostgreSQL, migrate before starting the API:

export SBOM_NEXUS_DATABASE_URL='postgresql+psycopg://user:password@host/sbom_nexus'
make migrate
make run

When migrations use short-lived login roles, set SBOM_NEXUS_MIGRATION_ROLE to the reviewed durable owner role. Alembic issues SET ROLE before creating or changing any object, so ownership does not become tied to an expiring lease. The production package uses sbom_nexus_owner.

Production deployments should mount a Secret and set SBOM_NEXUS_DATABASE_URL_FILE to its url file rather than exposing the DSN in a manifest or command argument. The same file setting is consumed by both Alembic and the API process.

PostgreSQL never auto-creates tables unless SBOM_NEXUS_AUTO_CREATE=1 is set explicitly. Normal production operation must use Alembic.

Register and ingest a repository

curl -X PUT http://127.0.0.1:8010/repositories/example \
  -H 'Content-Type: application/json' \
  -d '{"checkout_path":"/srv/repos/example","active":true}'

curl -X POST http://127.0.0.1:8010/sbom/example/ingest

The ingest response is terminal: ingested, or skipped with one of no-checkout, no-manifest, ingest-error, source-unavailable, or source-rejected. A skip advances queue fairness but does not advance last_success_at.

Controlled Forgejo source

Production automation uses the contract selected by CUST-WP-0064: Repo Manager projects a canonical Coulomb repository and full commit SHA, Activity Core freezes that source_ref with its bounded target set, and Nexus alone fetches and scans the source. Example projection:

{
  "source_ref": {
    "kind": "forgejo-archive-v1",
    "repository": "coulomb/example",
    "revision": "0123456789abcdef0123456789abcdef01234567",
    "observed_ref": "refs/heads/main",
    "observed_at": "2026-08-22T20:00:00Z"
  }
}

Enable only after the package has migration 0002, bounded ephemeral storage, and Forgejo/DNS egress:

export SBOM_NEXUS_CONTROLLED_SOURCE_ENABLED=true
export SBOM_NEXUS_FORGEJO_BASE_URL=http://forgejo-gitea-http.forgejo.svc.cluster.local:3000
export SBOM_NEXUS_SOURCE_TMP=/var/run/sbom-sources

The runtime defaults are one concurrent scan, 120 seconds each for fetch and scan, 100 MiB compressed, 512 MiB expanded, and 100,000 archive members. The source endpoint accepts no arbitrary URL and requires no credential for public Coulomb repositories. Never substitute a Forgejo administrator token.

Automated ingest and skip requests send the same value in Idempotency-Key and X-Activity-Core-Operation-ID. Nexus persists that operation identity in the snapshot transaction. A retry with the same request replays the original terminal outcome; reuse against a different route, repository, reason, or source reference returns HTTP 409.

Inspect catch-up

curl -s 'http://127.0.0.1:8010/sbom/catch-up?limit=3' | python3 -m json.tool

The default stale threshold is 30 days. Override it globally with SBOM_NEXUS_STALE_DAYS or per query with stale_days during controlled operation.

Import State Hub history

Synchronize the minimum repository identity/path projection first. This command does not call an ingest endpoint or create an SBOM snapshot:

uv run python scripts/sync_repository_projections.py \
  --source-url http://127.0.0.1:8000 \
  --target-url http://127.0.0.1:8010 \
  --host-id bnt-lap001 \
  --dry-run

Remove --dry-run only after the target is deployed. The apply path upserts all active and inactive records, then reads them back and fails on a missing or mismatched active/path projection. Extra target rows are reported but retained.

Run a read-only preview first:

uv run python scripts/import_state_hub.py --dry-run

Then run against a backed-up Nexus database:

uv run python scripts/import_state_hub.py \
  --source-url http://127.0.0.1:8000 \
  --target-url http://127.0.0.1:8010

Imports are idempotent on the State Hub snapshot UUID. Before cutover, compare the source/target repository, snapshot, and entry counts described in the extraction review. The command fails unless every legacy snapshot id, repository, timestamp, entry count, licence group, and direct-production copyleft count reconciles.

Current production limitations

  • Authentication and authorization are not yet integrated.
  • Structured operational metrics, backup/restore proof, and retention policy are required before authority cutover.
  • State Hub and Repo Manager callers have not yet been retargeted.