terramatch-microservices

Repository for the Microservices API backend of the TerraMatch service

Requirements:

• Node v20.11.1. Using NVM is recommended.
• Docker
• CDK CLI (install globally)
• AWS SAM CLI
• NX (install globally)
• NestJS (install globally, useful for development)

Building and starting the apps

• Copy .env.local.sample to .env
  • On Linux systems, the DOCKER_HOST value should be unix:///var/run/docker.sock instead of what's in the sample.
• To run all services:
  • nx run-many -t serve
  • The default maximum number of services it can run in parallel is 3. To run all of the services at once, use something like nx run-many --parallel=100 -t serve, or you can cherry-pick which services you want to run instead with nx run-many -t serve --projects user-service jobs-service.
• Some useful targets have been added to the root package.json for service sets. For instance, to run just the services needed by the TM React front end, use npm run fe-services, or to run all use npm run all.
• In .env in your wri-terramatch-website repository, set your BE connection URL correctly by noting the config in .env.local.sample for local development.
  • The NEXT_PUBLIC_API_BASE_URL still points at the PHP BE directly
  • New NEXT_PUBLIC_<SERVICE>_URL values are needed for each service you're running locally. This will typically match the services defined in V3_NAMESPACES in src/generated/v3/utils.ts.

Deployment

Deployment is handled via manual trigger of GitHub actions. There is one for services, and one for the ApiGateway. The ApiGateway only needs to be redeployed if its code changes; it does not need to be redeployed for updates to individual services to take effect.

Once this project is live in production, we can explore continuous deployment to at least staging and prod envs on the staging and main branches.

Creating a new service

• In the root directory: nx g @nx/nest:app apps/foo-service
• Set up the new main.ts similarly to existing services.
  • Make sure swagger docs and the /health endpoint are implemented
  • Pick a default local port that is unique from other services
• In your .env and .env.local.sample, add _PORT for the new service
• In api-gateway-stack.ts, add the new service and namespace to V3_SERVICES
• In your local web repo, follow directions in README.md for setting up a new service.
  • This step can be skipped for services that will not be used by the FE website.
• For deployment to AWS:
  • Add a Dockerfile in the new app directory. A simple copy and modify from user-service is sufficient
  • Add the new service name to the "service" workflow input options in deploy-service.yml
  • Add a new job to deploy-services.yml to include the new services in the "all" service deployment workflow.
    • Make sure to update the check-services step and follow the pattern for the if conditions on the individual service deploy jobs.
  • In AWS:
    • Add ECR repositories for each env (follow the naming scheme from user-service, e.g. terramatch-microservices/foo-service-staging, etc)
      • Set the repo to Immutable
      • After creation, set a Lifecycle Policy. In lower envs, we retain the most recent 2 images, and in prod it's set to 5
    • In CloudWatch, create a log group for each env (follow the naming scheme from user-service, e.g. ecs/foo-service-staging, etc).
      • TODO: the log groups could be created as part of the stack. The ECR repository is needed before the stack runs, so that will need to remain a manual process.

Database work

For now, Laravel is the source of truth for all things related to the DB schema. As such, TypeORM is not allowed to modify the schema, and is expected to interface with exactly the schema that is managed by Laravel. This note is included in user.entity.ts, and should hold true for all models created in this codebase until this codebase can take over as the source of truth for DB schema:

// Note: this has some additional typing information (like width: 1 on bools and type: timestamps on
//   CreateDateColumn) to make the types generated here match what is generated by Laravel exactly.
//   At this time, we want TypeORM to expect exactly the same types that PHP uses by default. Tested
//   by checking what schema gets generated in the test database against the real DB during unit
//   test runs (the only time we let TypeORM modify the DB schema).

This codebase connects to the database running in the wri-terramatch-api docker container. The docker-compose file included in this repo is used only for setting up the database needed for running unit tests in Github Actions.

Testing

To set up the local testing database, run the ./bin/setup-test-database.sh script. This script assumes that the wri-terramatch-api project is checked out in the same parent directory as this one. The script may be run again at any time to clear out the test database records and schema.

setup-jest.ts is responsible for creating the Sequelize connection for all tests. Via the sync command, it also creates database tables according to the schema declared in the entity.ts files in this codebase. Care should be taken to make sure that the schema is set up in this codebase such that the database tables are created with the same types and indices as in the primary database controlled by the Laravel backend.

Factories may be used to create entries in the database for testing. See user.factory.ts, and uses of UserFactory for an example.

To run the tests for a single app/library:

• nx test user-service or nx test common

To run the tests for the whole codebase:

• nx run-many -t test --passWithNoTests

For checking coverage, simply pass the --coverage flag:

• nx test user-service --coverage or nx run-many -t test --passWithNoTests --coverage

For apps/libraries that have tests defined, the coverage thresholds are set for the whole project in jest.preset.js