Event modeling

Goal, Make this schema alive :

Described using Event Modeling All the post-its on the schema exists in the codebase

Using NATS Jetstream and NestJS

• Code : Full API of NATS JetStream in javascript(ish)
• What are Event Collaboration And Event Sourcing ?
• Anatomy of a NATS application
• NestJS NATS JetStream microservice library

Concepts

And folder structure too

Feature :

The whole schema : One folder per event model containing everything needed for the flow

Adapter

Rows on the top : For each UI/Trigger on the schema an adapter.

NestJS controllers to adapt source protocol to domain's needs

• HTTP REST API
• Nats consumers :

  • A consumer is a stateful view of a stream

  • trigger side effects (think saga)

  • keep read models up to date

  • dedicated consumer to limit blast radius if any destination has an issue

Command

Blue post-its :

• Business logic : handles command, emits events.
• WARNING : Create, Update, Delete on the database is forbidden here !
• Better to use with one Command by file to make code more atomic.
• Consistency on the application : Infinite message deduplication in JetStream
• Publish domain events on nats subjects (JetStream or Core)

Projection

Green Post it :

• Consume events to manage read model's state
• Build needed read model(s) for this feature.
• Can be used for versionning and event.specversion management.
• Build a materialised view with an history using nats keystore

Model

All DTOs are here :

• Shared accross all features on all domains and all teams
• One folder per domain (bottom rows on the schema)
• Mainly declarative

Event

Orange Post its

• An event describe a finished action.
• A Domain Event describes an interesting action that occurred in the past and has an impact within the domain.
• Convention used : CloudEvents, a specification for describing event data in a common way

Stream

The row on the bottom :

• Where the events are gonna be stored or will be consumed.
• Conservation, deduplication, rate limit, ordering, ... are configured here
• Definition of a stream
• Multi Tenancy using Accounts

Entity

Our core domain

• A data model with an ID.
• Used to derivate other DTO : event, command, query results, ...
• Centralise validation rules using class validator with JSON Schema or AsyncAPI schema
• Modelina is the official AsyncAPI SDK to generate data models

Run it

docker run -d --name nats -p 4222:4222 -p 8222:8222 -p 6222:6222 nats --jetstream -m 8222

rush install
ts-node feature/hotel-booking/backend/main.ts 

curl --request PUT \
 --url http://127.0.0.1:3000/mutation/book-room \
 --header 'Content-Type: application/json' \
 --data '{
    "data": {
      "room": { "id" : 1},
      "client": { "id": 1},
      "date": "2022-12-12"
    }
  }'

curl --request PUT \
 --url http://127.0.0.1:3000/mutation/cleanup \
 --header 'Content-Type: application/json' \
 --data '{
    "data": {
      "room": { "id" : 1},
      "date": "2022-12-12"
    }
  }'

curl --request PUT \
 --url http://127.0.0.1:3000/mutation/checkin \
 --header 'Content-Type: application/json' \
 --data '{
    "data": {
      "room": { "id" : 1 },
      "client": { "id": 1},
      "date": "2022-12-13"
    }
  }'

nats stream view room-booking --subject='booking.1.2022-12-12.1_2022-12-12'
nats stream view room-cleanup --subject='cleanup.1.2022-12-12.1_2022-12-12'

TODO :

• Documentation
  • AsyncAPI request and sub
  • Websocket example
  • NATS cli command doc to debug (or docker GUI ?)
• Event|Command validation
  • Use CloudEvent as base event interface on driver
  • Use CloudEvent as base command interface
  • Middleware validation
• Idempotency
  • publish options per event's emit
  • Fix Expect last subject sequence not working (poc root)
• Shared state with materialized views
  • allow access to nats JetStream KeyStore from client
  • mapper to nats Keystore table.id.attributes = bucket.key.value
  • watch in projection to compare with state
  • handle state modification automatically
• QOS connection lost
  • Implement heartbeat https://github.com/nats-io/nats.deno/blob/main/jetstream.md#heartbeats
  • Audit log stream with 14 days memory, store command = event type, date, subject, id
  • Correlation ID shared from start of process
• Auth and security
  • configure nats accounts
  • Set websocket access limits solutions
• Publishing
  • publish event
  • send message
  • simplify NatsJetStreamClient+Proxy
• Subscribing
  • Pull Subscription
  • One specific consumer per destination
  • consumer
  • message handler with nats
  • auto stream creation