Ralph provides multiple packaging formats to support different deployment scenarios: Debian packages for Ubuntu, Docker images, and Python packages via setup.py.
Python Package (setup.py)
Package Configuration
Ralph uses setuptools for Python packaging. The main configuration is in setup.py:1:
Requirements
Ralph requires Python 3.10 or higher (setup.py:9):
Entry Points
Ralph provides multiple console scripts and extension points:
Console Scripts:
ralph - Production server
dev_ralph - Development server with additional features
test_ralph - Test runner
validate_ralph - Cross-validation tool
Extension Points:
back_office.transition_action.email_context - Email context for transitions
account.views.get_asset_list_class - Custom asset list views
ralph.cloud_sync_processors - Cloud synchronization processors
See setup.py:37 for all entry points.
Version Generation
Ralph uses a shell script to generate versions based on git tags:
The version format is YYYYMMDD.PATCH (e.g., 20240315.1).
Debian Packages
Ralph provides official Debian packages for Ubuntu using dh_virtualenv, which packages the application and all dependencies into a self-contained virtualenv.
Why dh_virtualenv?
Ralph’s dependency requirements make traditional Debian packaging challenging. Using dh_virtualenv provides a compromise between package availability and maintenance effort by:
- Installing all dependencies into a virtualenv
- Packaging the entire virtualenv as a
.deb file
- Avoiding conflicts with system Python packages
- Simplifying dependency management
Package Details
Package Name: ralph-core
Build Dependencies (debian/control:5):
debhelper (>= 9)
dh-virtualenv
git
libmysqlclient-dev
python3 (>=3.10)
libffi-dev
nodejs (>= 22.0.0)
npm
Runtime Dependencies (debian/control:10):
python3 (>=3.10)
libmysqlclient21
python3-distutils
debconf
Suggested Packages:
Building Packages
Ralph supports two build modes: snapshot (testing) and release (production).
Building Snapshot Packages
Snapshot packages can be built from any commit for testing purposes:
The built package will be placed in the build/ directory with a name like:
This target:
- Builds a Docker container with build dependencies
- Generates a snapshot changelog
- Builds the package
- Copies it to
build/
- Cleans up the container
See Makefile:33 for implementation.
Building Release Packages
Release packages should be built from tagged commits on the ng branch:
See Makefile:28 for implementation.
Release packages are meant for production use. Always build from tagged commits on the ng branch.
Publishing Packages
Packages are published to packagecloud.io/allegro/ralph by Ralph maintainers. Publishing is triggered automatically when a new GitHub release is created.
Docker Images
Ralph provides multiple Docker images for different purposes.
Available Images
ralph: Main application image
Builds:
allegro/ralph:latest
allegro/ralph:<VERSION>
ralph-static-nginx: Static files with nginx
allegro/ralph-static-nginx:latest
allegro/ralph-static-nginx:<VERSION>
inkpy: Additional utility image
See Makefile:40 for build configuration.
Building Snapshot Docker Images
Build snapshot images for testing:
This will:
- Build a snapshot Debian package
- Build Docker images using the snapshot package
- Tag images with snapshot version
See Makefile:57 for implementation.
Publishing Docker Images
See Makefile:73 and Makefile:80 for publishing targets.
Customizing Docker Repository
Set the DOCKER_REPO_NAME variable to push to a different repository:
Package Ecosystem Design
Ralph’s packaging system is designed to be environment-agnostic:
Key Principles
-
No specific environment required: All operations can be performed using tools a regular Ralph developer already has
-
Docker-based builds: Everything runs in Docker containers to avoid platform dependencies
-
Container rebuilds: Containers are rebuilt for each operation to avoid IO performance issues with mounted volumes on certain platforms
-
Official repository: Packages are published to packagecloud.io, owned by Allegro
-
GitHub integration: Package publishing is triggered by creating GitHub releases
Why Docker for Building?
Using Docker for package building provides several benefits:
- Platform independence: Developers on macOS, Windows, or Linux can build packages
- Consistent environment: Same build environment for all maintainers
- No system pollution: Build dependencies don’t affect the host system
- Reproducible builds: Same Dockerfile produces same packages
Makefile Targets
Package Management
Docker Images
Development
See Makefile:1 for all available targets.
Installation from Packages
Installing Debian Package
Installing from PyPI
Installing from Source
Best Practices
-
Use snapshot packages for testing: Never deploy snapshot packages to production
-
Verify package contents: After building, inspect the package to ensure all files are included
-
Test in containers: Use Docker to test packages in clean environments
-
Follow semantic versioning: Version numbers should reflect the release date and patch level
-
Document build dependencies: Keep
debian/control and setup.py requirements in sync
-
Clean builds: Always use
--force-rm with Docker builds to avoid stale layers
Troubleshooting
Build Fails with Missing Dependencies
Ensure all build dependencies are listed in debian/control. Check the build log for specific missing packages.
Version Mismatch
The version is generated from git tags. Ensure you have the latest tags:
Docker Build Timeout
Increase Docker’s memory and CPU allocation in Docker Desktop settings.
Package Installation Fails
Check that all runtime dependencies are installed: