Rails upgrade

[kylefritz]

Progress Update (March 1, 2026 — audit hardening pass)

Completed a best-practices cleanup/audit pass after the Rails 7.2 + Solid Queue migration work.

What was tightened

• Release task correctness
  • fixed Sentry deploy URL/release handling in lib/tasks/release.rake.
• Gem hygiene
  • removed stale minitest < 6 pin.
  • upgraded rack-mini-profiler to Rack-3-compatible ~> 4.0.
• Rails defaults + queue bootstrap reliability
  • moved config.load_defaults to 7.2.
  • hardened solid_queue:bootstrap to validate all queue schema tables, not just sentinel tables.
• Queue UX/observability
  • moved demo job to dedicated demo queue.
  • reduced Mission Control log noise in dev by only wiring MissionControl::Jobs.logger in production.
• Frontend dependency/code cleanup
  • pruned unused ActionCable JS wiring (app/javascript/channels/*).
  • refreshed Rails JS deps (kept actively-used packages current).
  • removed dead react-stripe-elements TS shim.
• Dead/legacy config cleanup
  • removed obsolete PG coder initializer used only for pre-7.1.
  • removed stale pre-cutover queue comments/requires in development.rb.
  • centralized Solid Queue retention policy in app config and removed duplicate queue cleanup path from cleanup:trim_analytics.
  • removed redundant RAILS_LOG_TO_STDOUT from app.json (logging now explicitly configured in env files).

Commit map (audit pass)

• 23b7339 Fix release task Sentry deploy release identifier
• 5e31e47 Remove legacy minitest version pin
• 2b51e4e Load Rails 7.2 framework defaults
• f45bfed Validate full Solid Queue schema during bootstrap
• dd4e658 Upgrade rack-mini-profiler to Rack 3 compatible release
• 452d3e8 Move queue demo job onto dedicated demo queue
• 6a2be35 Limit Mission Control logger wiring to production
• b263644 Prune unused ActionCable JS and refresh Rails frontend deps
• 6df9102 Remove obsolete PG coder compatibility initializer
• e918717 Drop dead react-stripe-elements TypeScript shim
• 1ca077a Remove stale pre-cutover queue config comments
• dde144f Centralize Solid Queue retention and trim redundant env config

Validation snapshots

• Rails tests passing: 215 runs, 0 failures
• Bun tests passing: 46 pass, 0 fail
• Typecheck passing: bun run typecheck

Safety tag

• Added and pushed tag: before_audit -> 88b8864c0a1665109f35e9102e625266bac65e90

---

Progress Update (March 1, 2026)

Completed in this PR branch since the initial plan post:

• Phase 3 (Solid Queue cutover)
  • switched worker process to Solid Queue (bin/jobs), removed Sidekiq runtime wiring, added Mission Control Jobs admin link and demo job trigger.
  • added Solid Queue cleanup in cleanup:trim_analytics.
• Phase 4 (Redis removal)
  • switched production ActionCable adapter to async.
  • removed Redis runtime initializer and Redis dependency from app/runtime docs.
• Phase 5A (React ecosystem modernization)
  • migrated Stripe checkout from react-stripe-elements to @stripe/react-stripe-js / @stripe/stripe-js.
  • removed Material UI usage in builder components and removed @material-ui/* dependencies.
• Phase 5B (React 18 core)
  • upgraded to React 18 / ReactDOM 18 and migrated entrypoints from ReactDOM.render to createRoot.
• Phase 6 (hardening + docs)
  • updated README/runbook notes for Solid Queue and async ActionCable.
  • stabilized app TypeScript checks with tsconfig.app.json and app typing fixes (bun run typecheck now passes).

Recent commit map:

• 371cc9d Stabilize app TypeScript checks after React 18 migration
• c090201 Update README for Solid Queue and async Action Cable
• cb39cde Remove unused react-is dependency
• fb5bc86 Upgrade React to 18 and migrate entrypoints to createRoot
• 9819633 Replace Material UI builder components with local styled UI
• c5ea276 Migrate checkout Stripe integration to stripe-js Elements
• 2a1fce5 Remove Redis runtime dependency and use async Action Cable
• e7a4153 Bootstrap Solid Queue tables during setup and release

Validation snapshots on this branch:

• Rails tests passing (213 runs, 0 failures)
• Bun tests passing (46 pass, 0 fail)
• App typecheck passing (bun run typecheck)

---

Motzi Upgrade Plan: Rails + React + Queue Stack

1) Goals

• Upgrade Rails to the newest version that is realistic without a major rewrite.
• Upgrade React to the newest version that is realistic without a major rewrite.
• Remove Sidekiq + Redis from the core app stack.
• Keep admin workflows stable (ActiveAdmin is the highest-risk area).
• Ship incrementally with production-safe checkpoints.

2) Recommended Targets

Primary target (recommended)

• Rails: 7.2.x
• ActiveAdmin: 3.4.x (or latest compatible 3.x)
• React: 18.x
• Queue backend: solid_queue (database-backed ActiveJob)
• ActionCable backend: async (if real-time is unused) or solid_cable (if durability/multi-instance needed)

3) Current State Snapshot

• Rails pinned to 6.1: gem 'rails', '~> 6.1.7.3' in Gemfile.
• ActiveAdmin pinned to 2.13.1 and lockfile constrains railties < 7.1.
• React is 16.x with ReactDOM.render entrypoints.
• JS deps include blockers for React 18:
  • react-stripe-elements (legacy)
  • @material-ui/core v4 (supports React 16/17)
  • testing libs pinned for pre-18 behavior
• Sidekiq + Redis wired in:
  • queue adapter config (ShopConfig.shop.queue_adapter)
  • sidekiq initializer
  • /sidekiq route mount
  • worker process in Procfile and Procfile.dev
  • sidekiq/deploy hook in release task
  • Redis cable adapter in production config/cable.yml

4) Scope and Non-Goals

In scope

• Dependency and framework upgrades.
• Queue backend migration and Redis removal.
• Admin and ordering flow compatibility fixes.
• Test and deploy pipeline updates required by upgrades.

Out of scope (for this cycle)

• UI redesign / broad component rearchitecture.
• Business logic refactors unrelated to compatibility.
• Rails 8 / React 19 migration.

5) Delivery Strategy

• Use commits and ship in phases.
• Each phase must be deployable and rollback-safe.
• Treat admin paths and checkout/email flows as release gates.
• Do not combine Rails major hop + React major hop + queue migration in one PR.

6) Work Plan (Detailed)

Phase 0: Baseline, Safety Nets, and Branching

Objectives

• Establish a stable baseline and avoid regressions.

Tasks

• Create upgrade branch: upgrade-rails-react-queue.
• Capture baseline versions:
  • bundle list | rg 'rails|activeadmin|sidekiq|redis'
  • cat package.json | rg 'react|material|stripe|testing-library'
• Run and record baseline checks:
  • bundle exec rails test
  • bun test
  • bun run typecheck
• Add/confirm CI jobs for Ruby + JS checks.
• Document critical manual flows for QA (see section 9).

Exit criteria

• Baseline test pass/fail documented.
• Known failing tests listed with owner (if any).
• Manual QA script agreed.

---

Phase 1: ActiveAdmin Upgrade (Unlock Rails 7.1+)

Why first

• Current ActiveAdmin lock blocks Rails > 7.0 (railties < 7.1).
• Upgrading ActiveAdmin first isolates admin-specific breakage before Rails hops.

Tasks

• Update gems:
  • activeadmin to latest 3.x compatible with target Rails.
  • align ransack, inherited_resources, kaminari if needed. we are not concerned about ransack search safety since it's admin only; ok to patch.
• Resolve DSL and rendering differences in:
  • app/admin/menus.rb
  • app/admin/users.rb
  • app/admin/bakery.rb
  • config/initializers/active_admin.rb
• Verify custom components:
  • GonFooter view factory registration.
• Re-run admin integration/unit tests and add targeted regression tests where behavior changed.

Risk hotspots

• member_action custom flows and redirects.
• custom controller do ... overrides (especially destroy batch logic).
• partial rendering from ActiveAdmin contexts.

Exit criteria

• Admin pages load without runtime errors.
• Critical admin actions pass manual QA:
  • open menu
  • test email
  • copy-from menu
  • user batch delete safeguards
  • pickup list page

---

Phase 2: Rails Upgrade Path (6.1 -> 7.0 -> 7.1 -> 7.2)

Strategy

• Upgrade one minor train at a time; run full test suite at each hop.
• Keep config.load_defaults changes separate from gem version bump when possible.

Phase 2A: Rails 6.1 -> 7.0

Tasks:

• bump Rails gems to ~> 7.0.x.
• run bin/rails app:update and apply only required config changes.
• resolve deprecations and keyword argument changes.
• review:
  • Devise compatibility
  • Sprockets/jsbundling integration
  • Zeitwerk/autoload edge cases

Exit criteria:

• app boots, tests run, admin + order checkout flow verified.

Phase 2B: Rails 7.0 -> 7.1

Tasks:

• bump Rails to ~> 7.1.x.
• regenerate/update framework defaults file for 7.1 and adopt toggles explicitly.
• verify ActiveStorage queue names and job config.
• verify puma, sentry, newrelic, paper_trail, ahoy_* compatibility.

Exit criteria:

• no production-blocking deprecations for 7.1.

Phase 2C: Rails 7.1 -> 7.2

Tasks:

• bump Rails to ~> 7.2.x.
• apply 7.2 defaults cautiously (separate commit).
• verify assets pipeline + esbuild output and digesting.
• rerun full test + manual QA.

Exit criteria:

• production candidate stable on Rails 7.2.

---

Phase 3: Replace Sidekiq/Redis with DB-backed Queue

Recommended implementation

• Use solid_queue for ActiveJob.
• Keep a worker process, but no Redis dependency.

Tasks

• Add solid_queue gem and run installer/migrations.
• Add mission_control-jobs gem for web UI (replaces Sidekiq Web):
  • mount MissionControl::Jobs::Engine at /jobs in routes.
  • protect behind admin auth (same guard as the old /sidekiq mount).
  • provides queue inspection, job retry/discard, pause/unpause, recurring task management.
• Switch queue adapter:
  • remove dynamic adapter switch in config/shop.yml and set explicit adapter by env.
  • set config.active_job.queue_adapter = :solid_queue in production.
• Keep existing queue names (default, mailers, active_storage) or map them explicitly in Solid Queue config.
• Update process model:
  • replace Sidekiq worker command in Procfile / Procfile.dev with Solid Queue worker command.
• Remove Sidekiq-specific code:
  • config/initializers/sidekiq.rb
  • require 'sidekiq/web' and route mount /sidekiq (replaced by /jobs Mission Control mount)
  • require "sidekiq/deploy" and Sidekiq::Deploy.mark! call in release task
  • gem 'sidekiq' and gem 'redis' (after cable migration decision)

Cutover plan

• Pause scheduler jobs briefly during deploy window.
• Drain Sidekiq queues to zero before cutover (no in-flight loss).
• Deploy Solid Queue worker process.
• Resume scheduler jobs and watch queue drain metrics.

Exit criteria

• deliver_later and perform_later jobs execute reliably.
• no Redis requirement for background jobs.

---

Phase 4: Remove Redis Dependency from ActionCable

Decision point

• If real-time features are not actively used, set production cable adapter to async and remove Redis.
• If real-time or multi-instance cable is needed, migrate to solid_cable.

Tasks

• Audit actual channel usage (currently only boilerplate channel files appear present).
• Option A (likely): set config/cable.yml production adapter to async.
• Option B: install/configure solid_cable.
• Remove config/initializers/redis.rb if unused globally.
• Remove Heroku Redis addon from infra config once queue + cable no longer depend on it.

Exit criteria

• No runtime attempts to connect to Redis in production.
• Cable behavior accepted for chosen adapter strategy.

---

Phase 5: React Upgrade Path (16 -> 18)

Why phased

• React 18 is blocked by current Stripe and Material UI dependencies.
• Upgrade ecosystem first, then React core.

Phase 5A: Dependency modernization (while app still stable)

Tasks:

• Stripe:
  • replace react-stripe-elements with @stripe/react-stripe-js + @stripe/stripe-js.
  • refactor payment components from injectStripe/StripeProvider to hooks + Elements API.
  • update tests to mock new Stripe API.
• Material UI:
  • migrate @material-ui/core / @material-ui/icons to MUI v5+ (@mui/material, @mui/icons-material).
  • update styling API usages (makeStyles/legacy imports where needed).
• Testing stack:
  • upgrade @testing-library/react and peers to React 18 compatible versions.
  • align react-test-renderer with target React.

Exit criteria:

• JS tests and typecheck pass before React core bump.

Phase 5B: React 18 core migration

Tasks:

• bump react and react-dom to 18.x.
• replace ReactDOM.render entrypoints with createRoot.
• validate event behavior, state batching, and test async timing changes.
• optionally enable StrictMode after functional parity is confirmed.

Exit criteria:

• all entrypoints render via createRoot.
• checkout/menu/builder/credits flows verified manually and via tests.

---

Phase 6: Hardening, Deploy, and Cleanup

Tasks

• Remove dead code and old shims.
• Update docs:
  • README.md (remove Redis requirement)
  • deployment notes / runbooks
  • scheduler and worker commands
• Performance and reliability checks:
  • job latency
  • mail throughput
  • admin response times
• Final dependency audit and lockfile cleanup.

Exit criteria

• Production stable for at least 1 release cycle.
• No Sidekiq/Redis references remain in app/runtime config.

7) PR Breakdown (Suggested)

1. PR-01: Baseline tests + upgrade harness docs.
2. PR-02: ActiveAdmin 3.x upgrade and admin fixes.
3. PR-03: Rails 7.0.
4. PR-04: Rails 7.1 + defaults.
5. PR-05: Rails 7.2 + defaults.
6. PR-06: Solid Queue introduction (keep Sidekiq still present, dark launch if needed).
7. PR-07: Switch queue adapter, remove Sidekiq runtime.
8. PR-08: ActionCable Redis removal (async or solid_cable).
9. PR-09: Stripe migration to modern React Stripe SDK.
10. PR-10: MUI migration + test tooling updates.
11. PR-11: React 18 bump + createRoot entrypoint migration.
12. PR-12: Docs/infrastructure cleanup and final remove-redis pass.

8) Risk Register

High risk

• ActiveAdmin upgrade regressions in custom admin actions and partials.
• Payment flow migration from react-stripe-elements.
• Queue cutover timing (in-flight jobs).

Mitigations:

• keep admin + checkout smoke tests mandatory for every phase.
• schedule queue cutover during low traffic window.
• drain and verify before removing old worker.

Medium risk

• Rails defaults toggles changing subtle behavior.
• test timing changes on React 18.

Mitigations:

• apply defaults in isolated commits.
• update flaky tests intentionally, not ad hoc.

9) Manual QA Checklist (Release Gates)

• Admin:
  • login/logout
  • dashboard load
  • menu CRUD + copy-from
  • open menu for orders
  • send test menu email
  • user batch delete behavior
  • pickup list view
• Customer:
  • view menu
  • add/remove cart items
  • place order (credit card + credit flow)
  • update order
  • receive confirmation email
• Jobs:
  • SendWeeklyMenuJob enqueues and processes
  • reminder tasks execute and send mail
  • ActiveStorage analysis/purge jobs execute

10) Definition of Done

• Rails on 7.2.x.
• ActiveAdmin on 3.x, all admin critical flows passing.
• React on 18.x with no ReactDOM.render usage.
• Sidekiq and Redis removed from app runtime dependencies.
• Background jobs running on Solid Queue in production.
• README/runbooks updated and deployment stable.