The repository includes microservices for the NTHU Distributed System course lab. The goal of this project is to introduce a production, realworld microservices backend mono-repo architecture for teaching purpose.
Before going through the following parts, make sure your Docker is running since we are generating/testing/building code inside a Docker container to prevent dependencies from conflicting/missing on your host machine.
The video service serves APIs that accept uploading a video, listing videos, getting a video and deleting a video.
The comment service serves APIs that accept creating a comment under a video, listing comments under a video, updating a comment and deleting a comment.
Many popular tools that are used in the realworld applications are adopted in this project too. For example:
- Use gRPC for defining APIs and synchronous communications between microservices. See the comment module protocol buffer definition for example.
- Use gRPC-gateway to generate a HTTP gateway server that serves RESTful APIs for the gRPC APIs. The purpose of having a HTTP gateway server is that realworld web applications typically do not use gRPC for communication.
- Use PostgreSQL in the comment service and MongoDB in the video service as the DBMS. The microservices architecture allows services to use different databases.
- Use Redis for cache. Realworld backend services use cache to speed up application performance. Redis is one of the most popular caching system and it is easy to learn.
- Use Kafka for asynchronous communications between microservices. Realworld backend services typically rely on message queue systems to accomplish asynchronous communications between microservices.
- Use MinIO storing files. Realworld backend services typically store user uploaded files in cloud storage like Google Cloud Storage or AWS S3. MinIO is a AWS S3 compatible storage system that allows the project to upload files without having a real cloud environment.
- Use OpenTelemetry to collect telemetry data.
- Use Prometheus as the metrics backend.
- Use Kubernetes as the container management system for deployment. Deployment yaml files are in the k8s directory.
Share libraries are wrapped so that they can be extended easily. For example, logs, traces, and metrics can be easily added to the custom share libraries. Share libraries are in the pkg directory.
Some modules use gRPC for communication or use the mockgen library for unit testing.
So there is a need to generate code manually when the code changed.
For generating code for all modules, run make dc.generate.
For generating code for a single module, run make dc.{module}.generate. For example: make dc.video.generate.
We implements unit testing on the DAO and service layers using the ginkgo framework.
To run unit testing for all modules, run make dc.test.
To run unit testing for a single module, run make dc.{module}.test. For example: make dc.video.test.
We use golangci-lint for linting.
To run linting for all modules, run make dc.lint.
To run linting for a single module, run make dc.{module}.lint. For example: make dc.video.lint.
To build docker image, run make dc.image.
The CI/CD runs in Github Actions. See the CI workflow spec and the CD workflow spec for more details.