Labgrid exporter systemd service

Each hw-node (bq, nuc, nemo, …) runs a labgrid-exporter process that publishes its local hardware to the coordinator at 10.0.0.41:20408. We run it as a systemd service under a single host-wide convention so every node looks the same.

Convention

One exporter per host. All paths are fixed:

Path	Purpose
`/etc/labgrid/exporter.yaml`	The exporter resource config. Owned by root, world-readable. Edit this file to change what the exporter publishes.
`/etc/default/labgrid-exporter`	Service knobs: coordinator address, instance name, yaml path, `PATH` (for `ser2net`). Generated by the installer.
`/etc/systemd/system/labgrid-exporter.service`	The unit. The `labgrid-exporter` binary path is baked into `ExecStart` at install time (systemd does not expand environment variables in the executable position). The arguments come from the env file above.

The unit is not templated — there is exactly one labgrid-exporter.service per host. If you need to expose more hardware, add it to /etc/labgrid/exporter.yaml rather than spinning up a second instance.

Installation

There are two scripts in scripts/labgrid-exporter/:

install.sh — the per-node installer. Generic; takes a yaml path plus options.
deploy.sh — orchestrator that pushes the installer to one or more hosts and runs it on each. Two modes:
- Manifest mode (default) — host list comes from a manifest file (nodes.conf next to the script, gitignored). Useful for the routine “deploy to my lab” workflow.
- Ad-hoc mode — single host specified on the command line via --node/--yaml/--bin. Useful for one-off deploys before you’ve added the host to a manifest.

Ad-hoc mode (no manifest needed)

./scripts/labgrid-exporter/deploy.sh \
    --node mini2 \
    --yaml /home/me/lg_mini2_exporter.yaml \
    --bin  /home/me/.local/bin/labgrid-exporter

Anything after a -- separator is forwarded to install.sh as extra arguments — handy for a one-off --coordinator override:

./scripts/labgrid-exporter/deploy.sh \
    --node mini2 --yaml /home/me/exporter.yaml \
    -- --coordinator 10.0.0.41:20408

Manifest mode — first-time setup

Copy the example manifest and edit it for your lab:

cd scripts/labgrid-exporter
cp nodes.conf.example nodes.conf
$EDITOR nodes.conf

Manifest format (pipe-separated, # comments allowed):

# host | yaml-path-on-host | extra-install.sh-args
bq   | /home/me/lg_bq_exporter.yaml   | --bin /home/me/.local/bin/labgrid-exporter
nuc  | /home/me/lg_nuc_exporter.yaml  |
nemo | /home/me/lg_nemo_exporter.yaml | --bin /home/me/venv/bin/labgrid-exporter

The third column forwards extra flags to install.sh. The most common one is --bin <path> when labgrid-exporter isn’t on the sudo login PATH (e.g. installed via uv tool to ~/.local/bin or inside a venv). deploy.sh always passes --stop-manual --force-yaml.

Manifest mode — routine deploy

# From the repo root:
./scripts/labgrid-exporter/deploy.sh             # every host in the manifest
./scripts/labgrid-exporter/deploy.sh bq nemo     # subset (must be in manifest)
./scripts/labgrid-exporter/deploy.sh --dry-run   # print only

# Use a different manifest:
./scripts/labgrid-exporter/deploy.sh --manifest /path/to/other.conf
LG_DEPLOY_MANIFEST=/path/to/other.conf ./scripts/labgrid-exporter/deploy.sh

Each host prompts once for the sudo password. deploy.sh errors out cleanly if a selected host isn’t in the manifest, so a typo in the host name is caught instead of being a silent no-op.

One-off install on a fresh node (without going through deploy.sh):

# On your workstation — push the script + unit to the node:
rsync -av scripts/labgrid-exporter/ <node>:~/labgrid-exporter-install/

# On the node:
ssh -t <node> 'sudo ~/labgrid-exporter-install/install.sh \
    /path/to/your-exporter.yaml --stop-manual'

The installer:

Stops any leftover templated labgrid-exporter@* instances from the prior naming scheme (idempotent — does nothing if none exist).
With --stop-manual, kills any manually launched labgrid-exporter process running as the service user.
Copies the source yaml to /etc/labgrid/exporter.yaml (creating /etc/labgrid/ first). If the file already exists with different content, the install fails unless you also pass --force-yaml, which keeps a timestamped .bak copy.
Writes /etc/systemd/system/labgrid-exporter.service from the template, baking in User= (defaults to $SUDO_USER).
Writes /etc/default/labgrid-exporter with LG_COORDINATOR, LG_EXPORTER_NAME, LG_EXPORTER_YAML, and PATH.
Runs systemctl daemon-reload and enable --now (skip with --no-start).

Concrete examples — the three current production nodes:

# bq
ssh -t bq 'sudo ~/labgrid-exporter-install/install.sh \
    /home/tcollins/dev/dt-fix/lg_adrv9371_zc706_tftp_exporter.yaml \
    --stop-manual'

# nuc
ssh -t nuc 'sudo ~/labgrid-exporter-install/install.sh \
    /home/tcollins/dev/lg-coordinator/lg_fmcdaq3_vcu118_exporter.yaml \
    --stop-manual'

# nemo
ssh -t nemo 'sudo ~/labgrid-exporter-install/install.sh \
    /home/tcollins/lg_adrv9009_zc706_tftp_exporter.yaml \
    --stop-manual'

Installer options

Option	Meaning
`--name NAME`	Instance name registered with the coordinator. Default is `hostname -s`.
`--coordinator ADDR`	`host:port` of the coordinator. Default `10.0.0.41:20408`.
`--user USER`	Service runs as this user. Default `$SUDO_USER`.
`--bin PATH`	Path to `labgrid-exporter`. Default: auto-detect on the service user’s PATH (picks up `~/.local/share/uv/tools/labgrid/bin/labgrid-exporter` from `uv tool install labgrid`).
`--ser2net-path DIR`	Prepend DIR to the service `PATH` so the unit finds a custom `ser2net` (e.g. `$HOME/opt/ser2net-4.6.1/sbin`).
`--no-start`	Install files but don’t `enable --now`.
`--force-yaml`	Overwrite an existing `/etc/labgrid/exporter.yaml` if it differs from the source. A timestamped `.bak` is kept.
`--stop-manual`	Kill any manually launched `labgrid-exporter` running as the service user before installing. Use this on first conversion from a manual setup.

Day-to-day operation

# Edit the resource config and reload:
sudo $EDITOR /etc/labgrid/exporter.yaml
sudo systemctl restart labgrid-exporter

# Live logs:
journalctl -u labgrid-exporter -f

# Status:
systemctl status labgrid-exporter

# Verify the place(s) registered with the coordinator (from any host):
labgrid-client -x 10.0.0.41:20408 places

Restart=on-failure brings the exporter back after a crash, and the unit’s WantedBy=multi-user.target makes it start automatically after a reboot. Re-running install.sh is idempotent — safe to use to pick up a new coordinator address, binary path, or yaml file (--force-yaml if the yaml content has changed).

Migrating from the older templated scheme

Earlier nodes used labgrid-exporter@<place>.service with the yaml referenced in-place from a user home directory. The new installer detects and disables any leftover templated instances automatically, so the migration is a single install.sh invocation per node — no manual cleanup of the old unit needed.