Skip to content

README.md

Latest commit

 

History

History
95 lines (77 loc) · 4.72 KB

README.md

File metadata and controls

95 lines (77 loc) · 4.72 KB
Raw

appstack — MRO Spare Parts Store (application layer)

Internal warehouse spare-parts store running on the appstack-infra AWS architecture (ECS-on-EC2, RDS, OpenSearch, SQS, S3/CloudFront, ALB). Developed and tested locally with Docker Compose against the same components — real OpenSearch and real SQS (LocalStack), not substitutes — so the whole thing runs on a laptop with zero AWS cost.

Run

cd app
docker compose up --build           # first run pulls images + builds services

Open the admin panel: http://localhost:8090 (admin@appstack.local / Admin123!)

Reset everything (re-seed DB, re-create queues/buckets): docker compose down -v Stop: docker compose down

What's running (16 containers)

Group Components Maps to (AWS)
Gateway nginx :8090 ALB path-routing + CloudFront/S3 SPA origin
Services auth :8081, catalog :8082, search :8083, inventory :8084, orders :8085, notification :8086, suppliers :8087, integration :8088 ECS tasks behind the ALB
Workers search-sync-worker, pdf-ingest-worker ECS worker tasks
Data Postgres :5432, OpenSearch :9200 RDS, OpenSearch domain
Messaging/Storage LocalStack :4566 (SQS + S3) SQS queues, S3 ingest bucket
Monitoring Prometheus :9090, Grafana :3000 self-managed on the EC2 host

All 8 microservices from the architecture diagram are live.

The demo flow (all clickable in the panel at :8090)

  1. Search — full-text part search served by OpenSearch.
  2. Catalog — browse parts (Postgres).
  3. Bulk upload — upload a supplier CSV. It PUTs to the S3 ingest bucket → ObjectCreated event → pdf-ingest SQS → pdf-ingest-worker bulk-upserts to Postgres and publishes price-sync → new parts appear in Search within seconds.
  4. Requisitions — raise a basket, admin approves → stock is deducted atomically via the shared ledger → search stock updates live.
  5. Low stock / Alerts — inventory-service low-stock report; notification-service background monitor raises low-stock alerts.
  6. Suppliers — vendor master data.
  7. Grafana (:3000) — per-service request rate, p95 latency, 5xx, up targets.

Architecture notes

  • .NET 8 Minimal API, Dapper + Npgsql (schema via db/init/*.sql, no EF migrations).
  • JWT issued by auth-service, validated by every service with a shared key; roles admin / staff.
  • One origin via the nginx gateway → no CORS; path rules mirror the ALB.
  • DB + SQS integration only between services (no service-to-service HTTP), matching the intended design. Canonical stock = parts.stock_qty, shared across catalog/ inventory/search.
  • price-sync SQS (catalog/inventory/orders publish → search-sync-worker → OpenSearch) and pdf-ingest SQS (S3 event → pdf-ingest-worker → Postgres) are the two pipelines from the sqs_messaging module.
  • Every Dockerfile is multi-stage and reused as-is for ECR/ECS.

On AWS (implemented & validated)

The same architecture runs on real AWS — the local pieces map 1:1:

Local AWS
nginx gateway (path routing + SPA origin) CloudFront (S3 SPA origin + ALB API origin → one origin, no CORS) + ALB path rules
Docker Postgres db/init entrypoint db-migrate one-shot ECS task applies the same SQL to RDS
LocalStack S3 ObjectCreated:* real S3 event → pdf-ingest SQS (no suffix filter — bulk import is CSV)
docker compose services ECS-on-EC2 services (ecs_services module)

notification-service polls the DB (locally and on AWS) — it can't share the price-sync queue (competing consumers would steal messages from search-sync-worker); in production this becomes SNS fan-out or its own queue.

CI/CD — one-merge GitOps deploy (.github/workflows/apply.yml)

A merge to main runs: terraform apply → publish SPA to S3 + CloudFront invalidate → build & push 11 images to ECR → db-migrate → roll the 10 ECS services → print the site URL. Tear down with destroy.yml (confirm=destroy-appstack). The standalone app-deploy.yml (build/push + optional ECS rollout, workflow_dispatch) remains as a manual image-refresh path.

Layout

app/
  docker-compose.yml
  db/init/                 01 schema+seed … 07 integration   (applied in order)
  services/<name>/         Program.cs + Dockerfile per service/worker
  gateway/nginx.conf       ALB + CloudFront stand-in
  frontend/index.html      admin SPA (S3/CloudFront origin)
  localstack/init/         create SQS queues + S3 bucket/notification
  monitoring/              prometheus.yml + grafana provisioning/dashboards
  sample-data/             supplier-catalog.csv