up
Some checks failed
api-governance / spectral-lint (push) Has been cancelled
Docs CI / lint-and-preview (push) Has been cancelled
oas-ci / oas-validate (push) Has been cancelled
SDK Publish & Sign / sdk-publish (push) Has been cancelled
Policy Lint & Smoke / policy-lint (push) Has been cancelled
Policy Simulation / policy-simulate (push) Has been cancelled
devportal-offline / build-offline (push) Has been cancelled

This commit is contained in:
StellaOps Bot
2025-11-26 20:23:28 +02:00
parent 4831c7fcb0
commit d63af51f84
139 changed files with 8010 additions and 2795 deletions

View File

@@ -1,64 +1,58 @@
# Exception Governance Migration Guide
Status: Draft (2025-11-26) — aligns with Excititor VEX/exception flows and Doc task DOCS-EXC-25-007.
> **Imposed rule:** All exceptions must be time-bound, tenant-scoped, and auditable; legacy perpetual suppressions are prohibited after cutover.
## Why this migration
- Retire legacy suppressions/waivers and move to policy/VEX-first exceptions with provenance.
- Provide auditable, reversible exceptions with rollout/rollback plans and notifications.
This guide explains how to migrate from legacy suppressions/notifications to the unified Exception Governance model in Excititor and Console.
## Target state
- Exceptions are recorded as:
- VEX statements (OpenVEX) referencing findings and components.
- Policy overrides (simulator) with scope, expiry, and rationale.
- Optional authority approval attestation (DSSE) per exception bundle.
- Persistence: Excititor stores exception records with tenant, source, scope, expiry, approver, evidence links.
- Propagation:
- Graph overlays surface exception status on nodes/edges.
- Notify emits `exception.created/expired` events.
- CLI/Console consume exception status via `/exceptions` API.
## 1. What changes
- **Unified exception object:** replaces ad-hoc suppressions. Fields: `tenant`, `scope` (purl/image/component), `vuln` (CVE/alias), `justification`, `expiration`, `owner`, `evidence_refs`, `policy_binding`, `status` (draft/staged/active/expired).
- **Two-phase activation:** `draft → staged → active` with policy simulator snapshot; rollbacks produce a compensating exception marked `supersedes`.
- **Notifications:** move from broad email hooks to route-specific notifications (policy events, expiring exceptions) using Notify service templates.
- **Auditability:** each lifecycle change emits Timeline + Evidence Locker entries; exports include DSSE attestation of the exception set.
## Migration steps
1) **Freeze legacy inputs**
- Disable new legacy suppressions in UI/CLI.
- Mark existing suppressions read-only; export CSV/NDJSON snapshot.
2) **Export legacy suppressions**
- Run `exc suppressions export --format ndjson --with-rationale --with-expiry`.
- Store exports with SHA256 and DSSE envelope in `exceptions/export-YYYYMMDD/`.
3) **Transform to VEX/policy overrides**
- Convert each suppression to OpenVEX statement:
- `status`: `not_affected` or `under_investigation`.
- `justification`: map legacy reason → VEX justification code.
- `impact`: optional; include nearest safe version if known.
- Generate policy override record with:
- `scope` (component, environment, service).
- `expiresAt` (carry forward or set 30/90d).
- `rationale` from legacy note.
4) **Import**
- Use Excititor `/exceptions/import` (or CLI `exc exceptions import`) to load transformed NDJSON.
- Verify import report: counts by status, rejected items, conflicts.
5) **Notify & rollout**
- Notify downstream systems: Graph overlays refresh, Scanner/Vuln Explorer respect VEX, Policy Engine caches reload.
- Announce change freeze window and rollback plan.
## 2. Migration phases
1. **Inventory legacy suppressions**
- Export current suppressions and notification rules (per tenant) to NDJSON.
- Classify by scope: package, image, repo, tenant-wide.
2. **Normalize and enrich**
- Map each suppression to the unified schema; add `expiration` (default 30/90 days), `owner`, `justification` (use VEX schema categories when available).
- Attach evidence references (ticket URL, VEX claim ID, scan report digest) where missing.
3. **Create staged exceptions**
- Import NDJSON via Console or `stella exceptions import --stage` (CLI guide: `docs/modules/cli/guides/exceptions.md`).
- Run policy simulator; resolve conflicts flagged by Aggregation-Only Contract (AOC) enforcement.
4. **Activate with guardrails**
- Promote staged → active in batches; each promotion emits Timeline events and optional Rekor-backed attestation bundle (if Attestor is enabled).
- Configure Notify templates for expiring exceptions (T14/T3 days) and denied promotions.
5. **Decommission legacy paths**
- Disable legacy suppression writes; keep read-only for 30 days with banner noting deprecation.
- Remove legacy notification hooks after confirming staged/active parity.
## Rollback plan
- Keep legacy suppressions export.
- If import fails or downstream errors:
- Disable new exception endpoints feature flag.
- Re-enable legacy suppressions (read-only → writable) temporarily.
- Investigate rejected items; re-run transform/import after fix.
## 3. Data shapes
- **Import NDJSON record (minimal):** `{ tenant, vuln, scope:{type:'purl'|'image'|'component', value}, justification, expiration, owner }
- **Export manifest:** `{ generated_at, tenant, count, sha256, aoc_enforced, source:'migration-legacy-suppressions' }`
- **Attestation (optional):** DSSE over exception set digest; stored alongside manifest in Evidence Locker.
## Notifications
- When migrating: send summary to tenants with counts per status and upcoming expiries.
- Ongoing: Notify events `exception.created`, `exception.expired`, `exception.rejected` with tenant + scope.
## 4. Rollback plan
- Keep legacy suppressions read-only for 30 days.
- If a promotion batch causes regressions, mark affected exceptions `expired` and re-enable corresponding legacy suppressions for that tenant only.
- Emit `rollback_notice` Timeline events and Notify operators.
## Validation checklist
- [ ] Legacy suppressions export stored with SHA256 + DSSE.
- [ ] Transform script maps all legacy reasons → VEX justification codes.
- [ ] Import dry-run produces zero rejects or documented rejects with reasons.
- [ ] Overlay/Console shows exceptions on affected components.
- [ ] Rollback tested on staging.
## 5. Air-gap considerations
- Imports/exports are file-based (NDJSON + manifest); no external calls required.
- Verification uses bundled attestations; Rekor proofs are optional offline.
- Console shows AOC badge when Aggregation-Only Contract limits apply; exports record `aoc=true`.
## References
- Excititor module: `docs/modules/excititor/architecture.md`, `vex_observations.md`.
- Policy simulator: `docs/modules/policy/architecture.md` (POLICY-ENGINE-30-001..003).
- Graph overlays: `docs/modules/graph/architecture-index.md`.
## 6. Checklists
- [ ] All legacy suppressions exported to NDJSON per tenant.
- [ ] Every exception has justification, owner, expiration.
- [ ] Policy simulator run and results attached to exception batch.
- [ ] Notify templates enabled for expiring/denied promotions.
- [ ] Legacy write paths disabled; read-only banner present.
- [ ] Attestation bundle stored (if Attestor available) and Evidence Locker entry created.
## 7. References
- `docs/modules/excititor/architecture.md`
- `docs/modules/excititor/implementation_plan.md`
- `docs/modules/cli/guides/exceptions.md`
- `docs/security/export-hardening.md`
- `docs/policy/ui-integration.md`