V3 V1 Workflow Parity Audit v1

Status: active migration audit.

This document tracks workflow parity between the old V1 pages and the V3 production surfaces. V1 was useful as an entity dump; V3 should preserve the operator/user capability while changing the shape to workflow-first pages.

Rule

If a V1 action is still needed, it must land in V3 as one of these patterns:

• Workbench action queue for things an operator should triage now.
• Resource operations panel for actions scoped to one resource.
• Lifecycle/activity tab for what happened and where to pivot next.
• Family local nav for sibling workbenches inside one product family.

Do not add one-off buttons to isolated pages unless they fit one of those patterns.

Parity is not only route/read parity. A V3 replacement is incomplete until the resource lifecycle is visible by intent:

• Setup: create, enroll, bootstrap, configure, grant, attach, publish.
• Update: edit, activate, retire, upgrade, role/scope change, quota change.
• Maintenance: probe, repair, rotate, reconcile, drift review, health check.
• Recovery: retry, resume, rerun, restore, re-enroll, compensate.
• Destructive: release, decommission, detach, remove, delete, revoke.
• Evidence: audit, task, workflow, event, runbook, log, trace, or provider handoff.

When a V1 page exposed an action through an admin button or entity row, the V3 page must make the intent and post-click progress explicit. Hiding the same action behind a new button is not a completed migration.

Current Parity Matrix

V1 surface	V3 target	Parity status	Notes
/allocations	/v3-prod/workloads	mostly covered	V3 unifies compute and app runtimes. Default filter is active. Keep allocation/task/audit events visible from workload detail.
/allocations/{id}	/v3-prod/allocations/{id} → /v3-prod/workloads/{id}	mostly covered	Allocation detail now resolves to the owning workload and preserves legacy panel intent for connect/metrics/events/storage/config. Console, metrics, storage, events/tasks exist. Need continued hardening for null telemetry and slice network/metric gaps found in kind/dev.
/apps/catalog	/v3-prod/apps, /v3-prod/launch/app/{slug}	mostly covered	V3 launch is production-bound. Keep scheduler apps prominent in seed data and visual ordering.
/apps/instances	/v3-prod/workloads	mostly covered	Unified runtime list covers status and open actions. Workload detail now carries app runtime management for members, member operations, version changes, and operation history.
/apps/instances/{id}	/v3-prod/apps/instances/{id} → /v3-prod/workloads/{id}	mostly covered	App-instance detail now resolves to the owning workload and preserves legacy panel intent for connect/metrics/events/storage/config. V3 covers connect/metrics/tasks for app workloads and exposes safe runtime operations: start, stop, restart, repair, decommission, upgrade, rollback, add member, drain member, remove member, member/audit activity, and safe bootstrap/runtime credential posture without private material. Credential rotate/reconcile mutations remain a follow-up.
/apps/artifacts	/v3-prod/apps/artifacts	mostly covered	Project-scoped artifact workbench lists app artifacts, creates publish intents, registers digests, and exposes verify/promote/deprecate/revoke/retire actions. Follow-up: richer artifact detail/activity drawer if operators need per-artifact event history.
/settings/profile	/v3-prod/account/profile	covered	Includes identity/linking shape.
/settings/ssh-keys	/v3-prod/account/security, /v3-prod/access/credentials	covered	Account owns personal key records; Access shows cross-project credential posture.
/settings/team	/v3-prod/access/memberships	covered	V3 splits project/tenant views.
/access/quotas	/v3-prod/access/quotas	covered	Effective project/user allocation headroom is now distinct from entitlements. Platform-wide quota policy editing remains under /v3-prod/platform/config/quotas.
/admin/nodes	/v3-prod/platform/lifecycle, /v3-prod/platform/lifecycle/nodes/{id}	covered for current workflows	Lifecycle workbench and node detail exist. Node detail now has operations for probe, Netdata, bootstrap, re-enroll, agent repair, cert repair, resume/reactivate, MAAS re-image, destructive detach/remove/delete, and slice slot discovery/approval.
/admin/nodes/onboardings/{id}	/v3-prod/platform/lifecycle/onboardings/{id}	covered	Resume/rerun, timeline, install output, MAAS events, summary, and error evidence are present.
/admin/nodes/decommissions/{id}	/v3-prod/platform/lifecycle/decommissions/{id}	covered	Resume/rerun/cancel, timeline, install output, MAAS events, summary, and error evidence are present.
/admin/maas	/v3-prod/platform/lifecycle/maas	mostly covered	Sites, drift, onboarding/decommission summaries visible. The page now explains the site → profile → onboard operator path and links to setup checklist/register-site actions.
/admin/maas/new	/v3-prod/platform/lifecycle/maas/new	covered	Site create workflow exists.
/admin/maas/{siteId}	/v3-prod/platform/lifecycle/maas/{siteId}	mostly covered	Site detail is now workflow-shaped with operations, discovery, profiles, fabric, and workflows tabs. Site settings, credential write, probe, direct one-node onboarding, batch onboarding, candidate ignore/suppress/restore/onboard, profile create/update/disable, and RoCE upsert/enable/disable/delete are available. Needs live kind validation against real MAAS data.
/admin/skus	/v3-prod/platform/config/skus	covered	SKU publish queue now supports create, edit, pricing, activate, and retire.
/admin/os-images	/v3-prod/platform/config/os-images	mostly covered	Compatibility matrix now supports image registration and disable. Edit-in-place is not in the current API contract; use create new image plus disable old image.
/admin/quotas	/v3-prod/platform/config/quotas	covered for current policy	Allocation concurrency policy edit exists. Future tenant/project/user quotas are separate work.
/admin/audit-logs	/v3-prod/platform/evidence	covered	V3 evidence has filters, drawer, and pivots.
/admin/payments/sessions	/v3-prod/platform/finance	mostly covered	Stuck-first session list, interventions, and first-class finance activity stream exist. Session rows open a recovery panel with user pivot, internal balance credit, and refund workflow creation using the payment session as reference. Follow-up: add a true provider-reconcile endpoint if operators need replay rather than credit/refund.
/admin/users	/v3-prod/platform/iam	covered	Governance-first user directory exists.
/admin/users/{id}	/v3-prod/platform/iam/users/{id}	mostly covered	User detail includes roles, membership, credit/refund, evidence/finance pivots.
/admin/telemetry	/v3-prod/platform/ops and future observability surface	partial/deferred	V3 Ops has signals. Netdata edge routing is deferred; app/user observability is future work.
/admin/overview, /admin/ops	/v3-prod/platform/overview, /v3-prod/platform/ops	mostly covered	V3 is workflow-first. Continue validating live read models.

Remaining Follow-Ups

1. Backend resource activity/capability read models: node detail and app runtime detail now have first-class activity/capability read models. Storage bucket detail now includes bucket, attachment, grant, and audit activity in its Events tab. Platform config now exposes config audit/activity rows for SKU, OS image, policy/quota, MAAS profile, and MAAS site changes. Access audit now exposes membership, service-account, credential, entitlement, identity, and connectivity mutation history. Finance now exposes payment session, ledger, refund, and finance-audit lifecycle rows without provider payment references or raw webhook payloads. The UI now exposes the lifecycle operation shape, but some non-node pages may still build activity from facts, tasks, or workflow cards until their durable resource activity stream is available.
2. App runtime credential operations: bootstrap/runtime credential posture is now exposed without surfacing private material. Reconcile, rotate, and disable workflows remain follow-ups until their mutation contracts are safe and audited.
3. MAAS provisioning validation: validate site settings, credentials/probe, direct one-node onboarding, batch onboarding, discovery review/onboard, profile CRUD, RoCE assignment actions, workflow install output, MAAS event pivots, and decommission cancel against real MAAS data in kind.
4. Payment recovery backend follow-up: decide whether stuck sessions need a first-class provider reconcile/replay endpoint. V3 currently supports user credit and refund workflows from the finance workbench.
5. Access/identity write workflows: SSO provider setup, network/security policy changes, service-account rotation/disable, entitlement updates, and scoped quota writes need real contracts before V3 should expose them as mutable controls. Access quota posture is read-only today; keep mutable quota policy under Platform Config unless a scoped quota contract exists.
6. Final UX polish pass after functional parity: some migrated pages still read too much like data-dense admin views. After the remaining workflow validation passes, compare against /v3 mocks and tighten visual hierarchy, navigation return paths, and family-local drill-down behavior.

Backend/Contract Gaps Found During Audit

• V3 app workload detail has stable project_id and app_instance_id for safe runtime operations, stitches version/member/operation-history workflows from existing app instance APIs, and includes first-class credential posture without credential material. It still needs operation capability flags for credential reconcile/rotate/disable.
• Node detail exposes operation capability flags and a unified activity stream from allocations, node tasks, agent lifecycle, MAAS decommissioning, and audit events.
• Storage bucket detail exposes a first-class activity stream from bucket posture, workload attachments, sharing grants, and audit events without surfacing provider credentials.
• Platform config exposes a first-class activity stream from config audit rows covering SKU, OS image, policy/quota, MAAS profile, and MAAS site changes.
• Platform finance exposes a first-class activity stream from payment sessions, ledger entries, refund requests, and finance audit actions while keeping raw provider references out of the read model.
• Non-node resource details should receive first-class lifecycle/audit event streams so activity is not derived from facts, tasks, and workflow cards.