mendaomn / yaci

:crystal_ball: Yet Another CI

Yet Another CI

In the quest of understanding a bit better the world of Continuous Integration, I started wondering how would a simple CI work.

What better way to learn something new than to get one's hands dirty and code an automation server from scratch!

In this project I'll be building:

• an Automation Server, exposing RESTful endpoints to create and run pipelines
• a Command Line Interface, acting as client, to make it easy to actually use the CI in a real project

Architecture

An authenticated client can

• upload a Makefile with a name which will represent a pipeline
• run a previously-uploaded pipeline referring it by name

The automation server can

• store Makefiles
• run make -f {pipeline file name}
• store outputs of executions (exit code, stdout and stderr)

Usage

• Run a pipeline by name: POST /executions -D {"name": "speed-staging"} --> executionID
• Upload a pipeline: POST /pipelines -D {"name": "speed-staging", "file": %FILEBLOB%
• List available pipelines GET /pipelines --> speed-staging, speed-prod, cas-test, cas-prod
• Check pipeline status: GET /executions?id=executionID --> (OK/KO, stdoud, stderr)

CLI

automation-server <command> <args>

  commands:
    run             pipeline-name
      --watch, -w     runs check in watch mode
    upload          pipeline-name file-path
      --run, -r       run the pipeline right after uploading it
      --watch, -w     used along with --run, runs check in watch mode
    list
    check           execution-id
      --watch, -w     watch mode