How eclipse-tractusx/tractusx-edc works
Project type and purpose
• This is not a typical full-stack app but a platform packaging and deploying two microservice-style components of the Eclipse Data Space Connector (EDC):
– Control-Plane (administration, contract negotiation, resource management)
– Data-Plane (data transfer runtime)
• The project provides:
– Java modules and EDC extensions (under edc-controlplane, edc-dataplane, edc-extensions, core, spi)
– Docker images (built via ./gradlew dockerize)
– Helm charts for Kubernetes (under charts/tractusx-connector* and charts/tractusx-connector-memory)
– CI/CD workflows (GitHub Actions in .github/workflows)
File structure
• Top-level packaging:
– core/, spi/ → foundational libraries and SPI interfaces
– edc-controlplane/, edc-dataplane/ → control-plane and data-plane distribution modules (PostgreSQL+Vault, in-memory)
– edc-extensions/ → a catalogue of pluggable feature extensions (agreements, identity, policy, data-plane proxy, token refresh, validators, migrations, etc.)
– charts/ → Helm chart definitions for Kubernetes deployments
– edc-tests/ → automated tests, end-to-end fixtures, and deployment tests
– .github/actions & .github/workflows → build, test, publish Docker images and Helm charts
• Architectural pattern:
– Microservices: control-plane and data-plane run as separate pods
– Plugin/SPI: edc-extensions loaded at runtime via Java’s ServiceExtension SPI
– Infrastructure as Code: Helm charts define k8s Deployment, Service, Ingress, ConfigMap, NetworkPolicy
– CI/CD pipelines automate version bump, code quality scans, Docker and Helm publishes
Main components and their responsibilities
a. Control-Plane Service
– Java process exposing the Management API (contract definitions, negotiations, catalog, assets, EDRs, response channels)
– Loads core-utils, edr-core, json-ld-core, and selected edc-extensions
– Persists state in PostgreSQL
– Manages secrets in Hashicorp Vault (via edc-controlplane-postgresql-hashicorp-vault module)
b. Data-Plane Service
– Java process handling actual data transfer streams
– Can be configured to use Vault for token storage (edc-dataplane-hashicorp-vault) or in-memory runtime for dev (edc-runtime-memory)
c. Helm/Infrastructure Layer
– charts/tractusx-connector-memory: deploys both control-plane (in-memory) and data-plane pods for quick dev
– charts/tractusx-connector: production variant with two separate charts:
· control-plane (deployment-controlplane.yaml, service-controlplane.yaml, ingress-controlplane.yaml, hpa, networkpolicy, serviceaccount)
· data-plane (deployment-dataplane.yaml, service-dataplane.yaml, ingress-dataplane.yaml, hpa)
– ConfigMaps for control-plane and data-plane configuration and custom CA certificates
d. CI/CD Pipeline
– GitHub Actions build Java artifacts, run tests, generate Allure reports, scan code, bump versions
– Actions publish Docker images to registry, publish Helm charts to chart repository
– Deployment tests and upgradeability tests ensure backward compatibility
Relationships and data flow
• Actor (CLI or GUI) → interacts with Control-Plane’s REST API
• Control-Plane → reads/writes to PostgreSQL for metadata and Vault for credentials
• Control-Plane → negotiates contracts, publishes EDRs to Data-Plane via its management endpoint
• Data-Plane → uses EDRs to start data streaming over HTTP/gRPC (or other supported protocols)
• Data consumers/providers → pull/push data via Data-Plane public API
• Helm charts → deploy pods, services, ingresses; they consume Kubernetes Secrets/ConfigMaps to configure Vault and DB connectivity
• CI/CD → upon commit to main, workflows trigger build → Docker image → push → Helm chart package → push
Architectural patterns and design principles
• Separation of concerns: control vs. data plane
• Plugin architecture via Java SPI (org.eclipse.edc.spi.system.ServiceExtension) for extensibility without core changes
• Immutable infrastructure: Docker images built from code, deployed via declarative Helm charts
• Infrastructure modularization: separate charts for memory/dev and prod (Vault+Postgres)
• Automated testing: unit tests in each module, end-to-end tests in edc-tests, deployment smoke tests in CI
Be detailed—capture
• Both runtime options (in-memory vs. prod Vault+Postgres)
• All external dependencies (Vault, Postgres, Docker registry, Helm repo)
• Key extension modules under each plane
• CI/CD artifacts and stages