Docker¶

Run your workflow using Docker. Only Docker is required — no Python or cwltool installation needed, as the orchestration container includes everything.

Prerequisites¶

Docker

1. Unzip and (optionally) pre-pull tool images¶

unzip my_pipeline.crate.zip -d my_workflow
cd my_workflow

# Optional: pre-download tool container images so the first run doesn't stall.
# Without this, cwltool will pull images on-the-fly as needed.
bash prefetch_images.sh

2. Edit the Job File¶

Open workflows/<name>_job.yml and replace file path placeholders with your actual data paths. Use /data/... paths (e.g. /data/my_brain.nii.gz) to match the volume mount in the run command below.

3. Build the Orchestration Container¶

This builds a container with cwltool and your workflow files inside. It does not include your data or the tool containers — those are handled separately at runtime.

docker build -t my-pipeline .

4. Run the Workflow¶

docker run --rm \
  -v /var/run/docker.sock:/var/run/docker.sock \
  -v /path/to/data:/data \
  -v /path/to/output:/output \
  my-pipeline

Volume Mounts¶

Mount	Purpose
`-v /var/run/docker.sock:...`	Gives the orchestration container access to Docker so cwltool can launch tool containers
`-v /path/to/data:/data`	Makes your input data available inside the container at `/data` — job file paths should use `/data/...`
`-v /path/to/output:/output`	Where cwltool writes workflow results (output directory)

Additional Options¶

# Show usage info
docker run my-pipeline --help

# Run with verbose output
docker run --rm -v ... my-pipeline --verbose

Troubleshooting¶

Docker not found — Ensure Docker is running: docker --version

Permission denied on Docker — Add your user to the docker group: sudo usermod -aG docker $USER (log out and back in)

Validation errors — Validate the workflow: cwltool --validate workflows/<name>.cwl