A self-contained, reproducible Stack4Things (IoTronic) environment for demos and hands-on training: the full control plane (Keystone, IoTronic conductor + WAMP agent, Horizon UI), six virtual Lightning-Rod boards, and ready-made demos — plugins, web services over WSTUN, metrics (InfluxDB + Grafana), federated learning (Flower), fleets and on-demand boards.
Everything runs in Docker on a single machine. One command brings it up, one command populates the demos, and the whole thing survives restarts, reboots and container re-creation (all state lives in named volumes; startup is idempotent and ordered by health checks).
git clone https://github.com/ProSoDiAC/stack4things-demo-lab.git
cd stack4things-demo-lab
./lab init <ip-or-hostname-users-will-use> # optional, defaults to this machine's first IP
./lab up # pulls images (≈ 9 GB the first time), builds 3 small helper images, starts 20 containers,
# waits until healthy (≈ 1 min once the images are cached)
./lab bootstrap # creates boards, plugins, metrics stream, web services, FL deps (≈ 3 min, downloads PyTorch)
./lab verify # health report (self-heals what it can) -> "RESULT : READY"Then open http://<host>/horizon and log in as admin / s4t (domain Default).
| Control plane | Keystone, MariaDB, RabbitMQ, Crossbar (WAMP over TLS), WSTUN, IoTronic conductor + agent, Horizon with lab panels |
| Boards | board-alpha … board-zeta: six virtual Lightning-Rods, all online and registered after ./lab bootstrap |
| Plugins | HelloName (synchronous), EnvironmentalDemo (asynchronous, streams a real sensor dataset), fl-client-heart, fl-client-pm |
| Web services | weather API, WoT "Fritzing" UI + API, nginx demo, SSH — exposed through WSTUN public ports |
| Metrics | board → metrics-gateway → InfluxDB → Grafana, embedded in Horizon (IoT → Metrics) |
| Federated learning | Flower server + live dashboard, 3 boards training heart-disease / predictive-maintenance models (IoT → Federated Learning) |
| Fleets & on-demand boards | create fleets, run plugins on whole fleets, and spin up a new Lightning-Rod container from the Boards page |
| Tools | ./lab CLI, backup/restore, self-healing verifier, end-to-end test suite |
| Service | URL | User / password |
|---|---|---|
| Horizon | http://<host>/horizon |
admin / s4t (domain Default) |
| Grafana | http://<host>:3000 |
admin / admin (anonymous viewer enabled for the embedded dashboards) |
| Lightning-Rod UI | http://<host>:1474 … :1479 (alpha … zeta) |
me / arancino |
| SSH in boards (over WSTUN) | port shown in IoT → Services | root / arancino |
These are demo credentials baked into the configuration. Do not expose the lab on an untrusted network (see Security).
- Linux host, Docker Engine ≥ 24 and the Compose v2 plugin (tested with Docker 28.0 / Compose 2.33)
- ≈ 8 GB RAM, ≈ 20 GB free disk (images ≈ 9 GB, plus volumes), internet access on first start (image pulls, Python wheels for the FL clients, the sample dataset)
python3≥ 3.8 on the host (standard library only) for./lab bootstrap/verify;curl- Free host ports:
80, 3000, 8080, 8181, 50001-50100, 1474-1479(public) and3306, 5000, 5672, 15672, 8086, 8090-8093, 8812(loopback only by default)
| Command | What it does |
|---|---|
./lab init [HOST] |
create .env from .env.example, set REPO_ROOT and the host address used in printed URLs |
./lab up |
build + start everything, wait for health checks |
./lab bootstrap [--only boards plugins fl metrics services verify] |
populate the demos; idempotent, safe to re-run |
./lab verify [--check-only] |
check every layer (containers, HTTP, boards, plugins, metrics, WSTUN services, FL) and heal |
./lab urls |
print URLs and credentials |
./lab ps / logs [svc] / restart [svc] / down |
usual Compose helpers |
./lab backup [DIR] / ./lab restore DIR |
DB dump + all volumes; see operations |
./lab reset |
destroy containers and data (asks for confirmation) |
- docs/DEMOS.md — walk-through of each demo, with the commands and what to look at in Horizon
- docs/ARCHITECTURE.md — services, network, volumes, ports, how persistence works, and what the lab patches
- docs/OPERATIONS.md — configuration, security, backup/restore, upgrades, troubleshooting, tests
- lab-ops/ — helper scripts for manually managed Lightning-Rod containers
tests/run-all.sh # ~2 min: stack health, Horizon login + every panel, metrics E2E,
# fleets (37 checks), federated learning (heart + PM), dynamic board lifecycle
tests/run-all.sh quick # skips the FL training and dynamic-board testslab CLI entry point
docker-compose.yml the whole stack (single file, pinned image digests, health checks, restart policies)
.env.example configuration (copy to .env; `./lab init` does it)
conf/ IoTronic conductor/agent, Horizon, Crossbar configuration
entrypoints/ startup scripts for keystone, conductor, agent, UI and Lightning-Rod containers
docker/ Dockerfiles of the three helper images (metrics gateway, FL control, LR log proxy)
patches/ lab modifications mounted over the upstream images (Horizon panels, conductor fix, Grafana)
plugins/ HelloName and EnvironmentalDemo plugin sources
experiments/ metrics, federated learning, multi-board/fleets, web services (WSTUN), networking helpers
vendor/ small upstream sources the demos need (Flower server, weather server) — see vendor/README.md
scripts/ bootstrap.py, verify-demo.py, backup/restore
tests/ end-to-end test suite
lab-ops/ scripts for manual Lightning-Rod containers
Based on the Stack4Things Docker deployment from the
AssemblingSmartCPS repositories and the upstream
MDSLab images (mdslab/iotronic-*, mdslab/lrod), which this repository
pulls unmodified (pinned by digest) and customises through bind-mounted patches. See
vendor/README.md for the few third-party sources included.
Licensed under the Apache License 2.0.