Running the full application with Docker

DeepRootGen uses Docker Compose to run its interactive application and the services supporting simulation, calibration and experiment tracking. You do not need Docker to run the simulation model by itself; see Run the model independently.

Prerequisites

Install Git, Docker Engine or Docker Desktop, and the Docker Compose plugin. Ensure the machine has sufficient memory and storage for multiple services.

Clone and configure

git clone https://github.com/Plant-Food-Research-Open/deep-root-gen.git
cd deep-root-gen
docker compose -f docker-compose.yaml config --services
docker compose -f docker-compose.yaml config --variables

Review the values in .env and the Compose file. In particular, verify image version, database connection settings, credentials, host names, object storage settings, and exposed ports. Keep secrets out of published files and replace any development credentials before exposing the stack externally.

Start the main application

docker compose -f docker-compose.yaml up -d
docker compose -f docker-compose.yaml ps

The stack includes the Dash interface, Prefect workflow orchestration, MLflow experiment tracking, PostgreSQL, MinIO object storage and ArangoDB graph storage. To determine the actual host ports for an installation, use docker compose -f docker-compose.yaml ps; do not assume that container ports equal host ports.

Model deployment services (optional)

The repository’s docker-compose.model-deploy.yaml defines additional BentoML services. These are optional and use the same Docker network and back-end services as the primary Compose file.

Service

Host port

Optimisation

5001

Sensitivity analysis

5002

ABC

5003

SNPE

5004

Surrogate modelling

5005

Run both configurations together:

docker compose -f docker-compose.yaml -f docker-compose.model-deploy.yaml up -d
docker compose -f docker-compose.yaml -f docker-compose.model-deploy.yaml ps

The deployed endpoints and readiness depend on their associated BentoML services and their dependencies. A container starting is not equivalent to a model being ready to receive requests.

Logs and shutdown

docker compose -f docker-compose.yaml logs -f
docker compose -f docker-compose.yaml down

For an installation started with both files, supply both files when stopping:

docker compose -f docker-compose.yaml -f docker-compose.model-deploy.yaml down

Do not add -v to down unless deliberately deleting persistent volumes.

Troubleshooting

  • Run docker compose -f docker-compose.yaml config to validate variable expansion and the resulting service definitions.

  • Run docker compose -f docker-compose.yaml ps and logs --tail=100 to diagnose containers that fail or restart.

  • Confirm that all requested host ports are available.

  • Confirm that MLflow and BentoML can reach the relevant storage endpoints using container network names rather than localhost.

  • Use Run the model independently if only a local root architecture simulation is needed.