Server Setup

The backend ships as a three-container Docker Compose stack: the middleman, InfluxDB, and Grafana, all on one bridge network.

Stack

Service	Image / build	Purpose
middleman	built from the repo Dockerfile (python:3.11-slim)	The middleman process, see Middleman
influxdb	${INFLUXDB_IMAGE}	Telemetry, stats, and analysis buckets, see InfluxDB
grafana	${GRAFANA_IMAGE}	Live dashboards, see Grafana

Ports

Port	Maps to	What
${UDP_PORT}/udp	container ${UDP_PORT}	Telemetry UDP ingest from the car
8005	container 8000	Web dashboard and REST API
${INFLUXDB_HTTP_PORT}	container 8086	InfluxDB UI/API
${GRAFANA_HTTP_PORT}	container 3000	Grafana UI

Quick start

cp .env.example .env
# Fill in INFLUX_TOKEN, INFLUX_ORG, GRAFANA_API_KEY, MARPLE_API_TOKEN

docker compose up --build -d

Every setting the Python process reads (config.py) and every setting the Compose file reads come from the same .env, so there's one file to configure per deployment.

.env.example is out of date

The checked-in .env.example uses older variable names (GRAFANA_TOKEN, QUEUE_MAX, MEASUREMENT, UDP_IP) that don't match what config.py actually reads (GRAFANA_API_KEY, INFLUX_QUEUE_SIZE/MARPLE_QUEUE_SIZE/ANALYSIS_QUEUE_SIZE, INFLUX_MEASUREMENT, UDP_HOST). Treat config.py as the source of truth for the middleman's own variable names; the InfluxDB and Grafana container variables (DOCKER_INFLUXDB_INIT_*, GF_*) in .env.example are accurate as-is.

Deployment (CI/CD)

.github/workflows/deploy.yml runs on every push to main: it SSHs into the production host and runs redeploy.sh in /srv/docker/compose, which pulls the latest image and restarts the stack.

Manual update

From the server, update/update.sh rebuilds and restarts without touching InfluxDB/Grafana data:

./update/update.sh          # rebuild middleman only
./update/update.sh --all    # also recreate influxdb + grafana containers

update/docker-compose.yml looks stale

That file builds the middleman image from a different git repository and branch (SoftwareStack2026, feature/telemetry/backendV2, subfolder telemetry) rather than the local build context the root docker-compose.yml uses. It appears to predate this repo's split from a monorepo. update.sh runs against whichever docker-compose.yml sits in its working directory, so confirm which one is actually deployed on the server before relying on this file.

Logging

The middleman container logs with the json-file driver, capped at 10 MB times 3 files, so docker logs won't grow unbounded on a machine that's rarely rebooted.