Multicast Docker Harness

The examples/multicast-docker folder is a reproducible multicast harness built around nextmini_py. It launches Postgres, controller, one source node, and two receiver nodes in a dedicated Docker network.

This page focuses on running the harness. For lifecycle semantics and packet-level behavior, see Example: Multicast Flow Lifecycle and Multicast integration guide.

The source process creates a group, installs a star DAG from source to all receivers, waits for receiver readiness files, and sends a tensor payload. Each receiver joins the group, receives the payload losslessly, and writes a local artifact (receiver-<node>.bin). Shared metadata and outputs are written under examples/multicast-docker/artifacts.

Run

From the repository root:

cd examples/multicast-docker
docker compose up --build

The default run uses:

• source node id 1
• receiver node ids 2,3
• group label demo-multicast
• automatic tensor generation when no --tensor-path is provided

Verification

Watch runtime logs:

cd examples/multicast-docker
docker compose logs -f controller source receiver_a receiver_b

Then run the verification script from the repository root:

uv run examples/multicast-docker/scripts/verify_received_files.py

A passing run prints VERIFICATION PASSED: All receivers got identical data.

You can also inspect generated artifacts:

cd examples/multicast-docker
ls -lah artifacts
cat artifacts/group-info.json
cat artifacts/tensor-metadata.json

Cleanup

cd examples/multicast-docker
docker compose down --volumes

Remove generated receiver payloads and metadata if you want a fully clean workspace:

cd examples/multicast-docker
rm -f artifacts/receiver-*.bin artifacts/group-info.json artifacts/tensor-metadata.json artifacts/receiver-ready-*.json