git.stella-ops.org/docs/ops/authority-backup-restore.md

# Authority Backup & Restore Runbook

## Scope
- **Applies to:** StellaOps Authority deployments running the official `ops/authority/docker-compose.authority.yaml` stack or equivalent Kubernetes packaging.
- **Artifacts covered:** MongoDB (`stellaops-authority` database), Authority configuration (`etc/authority.yaml`), plugin manifests under `etc/authority.plugins/`, and signing key material stored in the `authority-keys` volume (defaults to `/app/keys` inside the container).
- **Frequency:** Run the full procedure prior to upgrades, before rotating keys, and at least once per 24 h in production. Store snapshots in an encrypted, access-controlled vault.

## Inventory Checklist
| Component | Location (compose default) | Notes |
| --- | --- | --- |
| Mongo data | `mongo-data` volume (`/var/lib/docker/volumes/.../mongo-data`) | Contains all Authority collections (`AuthorityUser`, `AuthorityClient`, `AuthorityToken`, etc.). |
| Configuration | `etc/authority.yaml` | Mounted read-only into the container at `/etc/authority.yaml`. |
| Plugin manifests | `etc/authority.plugins/*.yaml` | Includes `standard.yaml` with `tokenSigning.keyDirectory`. |
| Signing keys | `authority-keys` volume -> `/app/keys` | Path is derived from `tokenSigning.keyDirectory` (defaults to `../keys` relative to the manifest). |

> **TIP:** Confirm the deployed key directory via `tokenSigning.keyDirectory` in `etc/authority.plugins/standard.yaml`; some installations relocate keys to `/var/lib/stellaops/authority/keys`.

## Hot Backup (no downtime)
1. **Create output directory:** `mkdir -p backup/$(date +%Y-%m-%d)` on the host.
2. **Dump Mongo:**
   ```bash
   docker compose -f ops/authority/docker-compose.authority.yaml exec mongo \
     mongodump --archive=/dump/authority-$(date +%Y%m%dT%H%M%SZ).gz \
     --gzip --db stellaops-authority
   docker compose -f ops/authority/docker-compose.authority.yaml cp \
     mongo:/dump/authority-$(date +%Y%m%dT%H%M%SZ).gz backup/
   ```
   The `mongodump` archive preserves indexes and can be restored with `mongorestore --archive --gzip`.
3. **Capture configuration + manifests:**
   ```bash
   cp etc/authority.yaml backup/
   rsync -a etc/authority.plugins/ backup/authority.plugins/
   ```
4. **Export signing keys:** the compose file maps `authority-keys` to a local Docker volume. Snapshot it without stopping the service:
   ```bash
   docker run --rm \
     -v authority-keys:/keys \
     -v "$(pwd)/backup:/backup" \
     busybox tar czf /backup/authority-keys-$(date +%Y%m%dT%H%M%SZ).tar.gz -C /keys .
   ```
5. **Checksum:** generate SHA-256 digests for every file and store them alongside the artefacts.
6. **Encrypt & upload:** wrap the backup folder using your secrets management standard (e.g., age, GPG) and upload to the designated offline vault.

## Cold Backup (planned downtime)
1. Notify stakeholders and drain traffic (CLI clients should refresh tokens afterwards).
2. Stop services:
   ```bash
   docker compose -f ops/authority/docker-compose.authority.yaml down
   ```
3. Back up volumes directly using `tar`:
   ```bash
   docker run --rm -v mongo-data:/data -v "$(pwd)/backup:/backup" \
     busybox tar czf /backup/mongo-data-$(date +%Y%m%d).tar.gz -C /data .
   docker run --rm -v authority-keys:/keys -v "$(pwd)/backup:/backup" \
     busybox tar czf /backup/authority-keys-$(date +%Y%m%d).tar.gz -C /keys .
   ```
4. Copy configuration + manifests as in the hot backup (steps 3–6).
5. Restart services and verify health:
   ```bash
   docker compose -f ops/authority/docker-compose.authority.yaml up -d
   curl -fsS http://localhost:8080/ready
   ```

## Restore Procedure
1. **Provision clean volumes:** remove existing volumes if you’re rebuilding a node (`docker volume rm mongo-data authority-keys`), then recreate the compose stack so empty volumes exist.
2. **Restore Mongo:**
   ```bash
   docker compose exec -T mongo mongorestore --archive --gzip --drop < backup/authority-YYYYMMDDTHHMMSSZ.gz
   ```
   Use `--drop` to replace collections; omit if doing a partial restore.
3. **Restore configuration/manifests:** copy `authority.yaml` and `authority.plugins/*` into place before starting the Authority container.
4. **Restore signing keys:** untar into the mounted volume:
   ```bash
   docker run --rm -v authority-keys:/keys -v "$(pwd)/backup:/backup" \
     busybox tar xzf /backup/authority-keys-YYYYMMDD.tar.gz -C /keys
   ```
   Ensure file permissions remain `600` for private keys (`chmod -R 600`).
5. **Start services & validate:**
   ```bash
   docker compose up -d
   curl -fsS http://localhost:8080/health
   ```
6. **Validate JWKS and tokens:** call `/jwks` and issue a short-lived token via the CLI to confirm key material matches expectations. If the restored environment requires a fresh signing key, follow the rotation SOP in [`docs/11_AUTHORITY.md`](../11_AUTHORITY.md) using `ops/authority/key-rotation.sh` to invoke `/internal/signing/rotate`.

## Disaster Recovery Notes
- **Air-gapped replication:** replicate archives via the Offline Update Kit transport channels; never attach USB devices without scanning.
- **Retention:** maintain 30 daily snapshots + 12 monthly archival copies. Rotate encryption keys annually.
- **Key compromise:** if signing keys are suspected compromised, restore from the latest clean backup, rotate via OPS3 (see `ops/authority/key-rotation.sh` and `docs/11_AUTHORITY.md`), and publish a revocation notice.
- **Mongo version:** keep dump/restore images pinned to the deployment version (compose uses `mongo:7`). Driver 3.5.0 requires MongoDB **4.2+**—clusters still on 4.0 must be upgraded before restore, and future driver releases will drop 4.0 entirely. citeturn1open1

## Verification Checklist
- [ ] `/ready` reports all identity providers ready.
- [ ] OAuth flows issue tokens signed by the restored keys.
- [ ] `PluginRegistrationSummary` logs expected providers on startup.
- [ ] Revocation manifest export (`dotnet run --project src/Authority/StellaOps.Authority`) succeeds.
- [ ] Monitoring dashboards show metrics resuming (see OPS5 deliverables).