Where to Contribute
Screwdriver has a modular architecture, and the various responsibilities are split up into separate repos.
Check out the architecture diagram to see the overall workflow of Continuous Delivery using Screwdriver. The next few sections will help lay out where different code repositories fit.
The screwdriver repo is the core of screwdriver, providing the API endpoints for everything that screwdriver does. The API is based on the hapijs framework and is implemented in node as a series of plugins.
Build bookends allow a user to create setup and teardown steps for builds.
The API can also send notifications to users. notifications-base is the base class for defining the behavior between Screwdriver and notifications plugins, like email notifications and slack notifications.
The API can also uploading code coverage reports and/or test results. coverage-bookend defines the relationship between Screwdriver and coverage bookends. coverage-base is the base class for defining the behavior between Screwdriver coverage bookend plugins, like coverage-sonar.
The launcher performs step execution and housekeeping internal to build containers. This is written in Go and mounted into build containers as a binary.
- sd-cmd: A Go-based CLI for sharing binaries which provides a single interface for executing a versioned command (via remote binary, docker image, or habitat package) during a Screwdriver build
- sd-step: A Shared Step allows people to use the same packages and commands in all build containers, regardless of build environment
- meta-cli: A Go-based CLI for reading/writing information from the metadata
An executor is used to manage build containers for any given job. Several implementations of executors have been created. All are designed to follow a common interface. Executor implementations are written in node:
- executor-base: Common interface
- executor-docker: Docker implementation
- executor-j5s: Jenkins implementation
- executor-k8s: Kubernetes implementation
- executor-k8s-vm: Kubernetes VM implementation
- executor-nomad: Nomad implementation
The executor router is a generic executor plugin that routes builds to a specified executor.
The object models provide the definition of the data that is saved in datastores (analogous to databases). This is done in two parts:
A datastore implementation is used as the interface between the API and a data storage mechanism. There are several implementations written in node around a common interface.
- datastore-base: Base class defining the interface for datastore implementations
- datastore-sequelize: MySQL, PostgreSQL, SQLite3, and MS SQL implementations
- datastore-dynamodb: DynamoDB implementation, with dynamic-dynamodb being used to create the datastore tables for it
The Artifact Store (not to be confused with the datastores mentioned above) is used for saving log outputs, shared steps, templates, test coverage, and any artifacts that are generated during a build. The log service is a Go tool for reading logs from the Launcher and uploading them to the store. The artifact-bookend is used for uploading artifacts to the store.
Source Code Management
An SCM implementation is used as the interface between the API and an SCM. There are several implementations written in nodejs around a common interface.
- scm-base: Common interface
- scm-bitbucket: Bitbucket implementation
- scm-github: Github implementation
- scm-gitlab: Gitlab implementation
Templates are snippets of predefined code that people can use to replace a job definition in their
screwdriver.yaml. A template contains a series of predefined steps along with a selected Docker image.
- templates: A repo for all build templates
- template-main: The CLI for validating and publishing job templates
- template-validator: A tool used by the API to validate a job template
Node module for validating and parsing user’s
The Ember-based user interface of Screwdriver.
- circuit-fuses: Wrapper to provide a node-circuitbreaker with callback interface
- client: Simple Go-based CLI for accessing the Screwdriver API
- gitversion: Go-based tool for updating git tags on a repo for a new version number
- keymbinatorial: Generates the unique combinations of key values by taking a single value from each keys array
- toolbox: Repository for handy Screwdriver-related scripts and other tools
- hashr: Wrapper module for generating ids based on a hash
Adding a New Screwdriver Repo
We have some tools to help start out new repos for screwdriver:
- generator-screwdriver: Yeoman generator that bootstraps new repos for screwdriver
- eslint-config-screwdriver: Our ESLint rules for node-based code. Included in each new repo as part of the bootstrap process
If you create a new repo, please come back and edit this page so that others can know where your repo fits in.
Screwdriver.cd Tests and Examples
The organization screwdriver-cd-test contains various example repos/screwdriver.yamls and acceptance tests for Screwdriver.cd.