Testing

Layer	Tooling	State
Load and performance	k6, in tests/load/k6	substantial
Frontend unit	Jest via Nx (nx run-many --target=test)	5 spec files
Frontend end-to-end	Playwright, in frontend/web/e2e	5 suites for Module Federation behaviour
Backend unit and integration	none yet	being added

Load testing with k6

The k6 suite is the most complete testing asset in the repository. tests/load/k6/config.js centralizes the service definitions, both direct paths (/api/v1/... for port-forwarded services) and gateway paths (the Ocelot upstream templates), plus the load profiles below.

Profile	Shape (virtual users over time)	Thresholds
smoke	1 VU for 1 min	p95 below 500 ms, errors below 1%
load	ramp to 10, hold 1 min, ramp down	p95 below 800 ms, errors below 5%
stress	100, 200, 300 VUs (port-forward) or 100, 300, 500 VUs (gateway)	p95 below 2 s, errors below 10%
spike	2, then a jump to 100 for 2 min, then back to 2	p95 below 2 s, p99 below 5 s, errors below 10%
soak	20 VUs for about 10 min	p95 below 800 ms, p99 below 1.2 s

Scripts:

Script	Purpose
tests/load/k6/gateway-smoke-test.js	10 VUs for 30 s through the gateway; p95 below 2 s, errors below 5%. Used by the CI workflow
tests/load/k6/catalog-test.js, tests/load/k6/basket-test.js, tests/load/k6/ordering-test.js, tests/load/k6/discount-test.js	Per-service scenarios
tests/load/k6/stress-test.js, tests/load/k6/spike-test.js, tests/load/k6/soak-test.js	Profile-driven tests with custom summaries
tests/load/k6/workflow-shopping.js	End-to-end journey (browse, view product, add to basket, view basket, checkout, order history) with custom workflow_duration, workflow_success and workflow_failed metrics; p95 workflow below 10 s
tests/load/k6/workflow-website.js	The journey as the website performs it
tests/load/k6/verify-failures-test.js	Checks that failures are detected and reported

Runner scripts (run-all-tests.sh, run-with-gateway.sh, smart-port-forward.sh) handle port-forwarding or targeting a gateway URL. tests/load/k6/push-metrics.sh pushes results to a Prometheus Pushgateway for the Grafana k6 dashboard.

cd tests/load/k6
k6 run gateway-smoke-test.js                                   # local gateway on :8010
AWS_GATEWAY_URL=http://<nlb-host> k6 run -e TEST_TYPE=stress stress-test.js

In CI, .github/workflows/performance-test.yml runs on demand only. The weekly schedule was removed because the workflow does not yet provision the databases and the Discount service, so scheduled runs only failed. docs/dependency-management.md records this decision.

Frontend tests

• Unit (Jest, via Nx): host/src/app/app.spec.tsx, the store’s route test and utility tests (productUtils, formatUtils), and checkout’s formatUtils test. CI runs nx affected --target=test --passWithNoTests, so only projects touched by the PR run.
• End to end (Playwright): frontend/web/playwright.config.ts starts the host with --devRemotes=store,account,checkout on port 4200 and runs Chromium. The suites cover what is specific to micro-frontends:
  • module-federation.spec.ts: each remote loads dynamically, remotes load in sequence, load errors are handled, and state survives navigation.
  • cross-microfrontend.spec.ts: navigation and shared state across remotes.
  • deep-linking.spec.ts and browser-history.spec.ts: URLs inside a remote survive reloads and back/forward.
  • error-handling.spec.ts: an unavailable remote degrades gracefully.

The Playwright suite is not wired into CI yet.

Backend tests

dotnet test Ecommerce.sln runs 97 tests (86 unit, 11 integration), all under tests. They use xUnit, NSubstitute and Shouldly, with shared settings in tests/Directory.Build.props and package versions in Directory.Packages.props.

Project	Tests	What it covers
tests/unit/Common.Mediator.Tests	8	Dispatch, pipeline-behaviour ordering, and a DI regression test with ValidateScopes = true
tests/unit/Common.Api.Tests	12	Exception to RFC 7807 mapping (400 / 404 / 409 / 500), no stack traces outside Development
tests/unit/Basket.Application.Tests	25	Discount application, gRPC failure fallback, empty basket for new users, checkout validation
tests/unit/Catalog.Application.Tests	14	Not-found handling, paging clamps
tests/unit/Ordering.Application.Tests	27	Validators, not-found handling in handlers
tests/integration/Ordering.IntegrationTests	7	Migrations on a fresh and a legacy SQL Server (Testcontainers), fail-fast startup, consumer idempotency via the MassTransit test harness
tests/integration/Discount.IntegrationTests	4	Coupons survive restart, idempotent seeding, concurrent startups (Testcontainers PostgreSQL)

End-to-end smoke test

scripts/smoke/e2e.sh runs against a running Compose stack and drives the whole system through the gateway: list and fetch products, create a basket (the discount is applied over gRPC), reject an invalid checkout with 400 while keeping the basket, check out (202), then poll Ordering until the order produced by the RabbitMQ consumer appears. It also asserts the error contracts (404 problem details, empty basket for a new user, clamped paging).

cp .env.example .env
docker compose up -d --build
bash scripts/smoke/e2e.sh

Next layers

1. Contract checks for the .proto and the integration-event types, so publisher and consumer cannot drift apart.