nomadnet-pages

A collection of Nomadnet .mu pages and supporting Python scripts, organised by server.

Structure

nomadnet-pages/
├── pingplotter/
│   ├── pages/      # .mu pages → deploy to ~/.nomadnetwork/storage/pages/
│   └── scripts/    # .py dependencies called by the .mu pages
├── ascii-webcam/
│   └── pages/      # .mu pages → deploy to ~/.nomadnetwork/storage/pages/
├── lora-rnode/
│   ├── pages/      # .mu pages → deploy to ~/.nomadnetwork/storage/pages/
│   └── scripts/    # background daemons required by the pages
└── jpeg-webcam/
    └── pages/      # .mu pages → deploy to ~/.nomadnetwork/storage/pages/

Each subdirectory is a self-contained Nomadnet server. Pages and scripts for one server are never mixed with another.

Pages — pingplotter

Page	Description
index.mu	Landing page
ping-graph.mu	Live ping RTT graph to selected Reticulum nodes
mtr-graph.mu	MTR traceroute graph
announces.mu	Reticulum announce monitor
announce-rate.mu	Announce rate over time
iface-traffic.mu	Interface traffic stats
rnpath.mu	Reticulum path lookup
rnstatus.mu	Reticulum interface status
neofetch.mu	System info via neofetch
netio.mu	Network I/O stats
pingtest.mu	Simple ping test
services.mu	Systemd service status
sysinfo.mu	Basic system info
sysinfo2.mu	Extended system info
uptime.mu	System uptime
testpage.mu	Test page

Dependencies: ~/.venvs/pingtools/ venv with plotille and typing_extensions. announce-listener.py daemon must be running for announce-rate.mu.

Pages — ascii-webcam

Page	Description
index.mu	Landing page
webcam.mu	Live webcam image rendered as ASCII art
webcam-colour.mu	Live webcam image as thermal-style colour ASCII art

Dependencies: fswebcam (sudo apt install fswebcam), Pillow in ~/nomadnet-env. User must be in the video group.

Pages — lora-rnode

Page	Description
index.mu	Landing page
sysinfo.mu	Basic system info
sysinfo2.mu	Extended system info
rnstatus.mu	Reticulum interface status
retconfig.mu	Displays ~/.reticulum/config content
noise-monitor-braille.mu	RF noise floor and channel load graphs over time
rf-noise.mu	Live RF noise floor reading
rf-rssi.mu	Live RSSI from most recent LoRa packet
rssi-monitor.mu	Per-node RSSI/SNR from announces, plotted over time
names.mu	Lists known nodes with resolved human-readable names

Dependencies: ~/.venvs/rns-tools/ venv with plotille, typing_extensions and rns. Daemons noise-logger.py, rssi-logger.py and names-resolver.py must be running for graph and names pages.

Pages — jpeg-webcam

Page	Description
index.mu	Landing page with links to live webcam JPEG files
status.mu	Reticulum interface status via rnstatus
sysinfo.mu	Basic system info

Dependencies: fswebcam (sudo apt install fswebcam). User must be in the video group. A cron job writes JPEGs every 2 minutes to ~/.nomadnetwork/storage/files/ for Nomadnet to serve via /file/ links.

Server setup & reliability

Notes on running nomadnet itself as an always-on systemd service, distilled from comparing the pingplotter and webcam (ascii-webcam/jpeg-webcam) servers and fixing the startup-reliability issues found on each.

Architecture: embedded vs. shared rnsd

Reticulum can run two ways, pick one per server:

• Embedded (pingplotter): nomadnet starts its own Reticulum instance directly. Simplest option if nomadnet is the only Reticulum-using process on the box.
• Shared instance via rnsd (webcam server): a separate rnsd.service owns the actual interfaces and Reticulum Transport; nomadnet and any companion daemon connect to it as clients over a local Unix domain socket. Needed when more than one local app requires Reticulum. This is the default behavior when share_instance = Yes in ~/.reticulum/config (the default) — whichever process starts first becomes the shared instance.

Either way, nomadnet lives in a venv, conventionally at ~/nomadnet-env/:

python3 -m venv ~/nomadnet-env
~/nomadnet-env/bin/pip install rns nomadnet

Running the TUI headless: tmux

nomadnet with no arguments is a full-screen TUI — it needs a real pty, so it can't run as a plain systemd service with no terminal. The pattern is to have systemd launch it detached inside tmux, so you can attach to the same running session later:

tmux attach -t nomadnet

(Detach without killing it: Ctrl-b then d.) Use the default tmux socket, not a custom -L socket — a custom socket means remembering to pass -L <name> on every attach, and forgetting it produces a confusing "no sessions" error against the wrong (default) socket for no benefit.

systemd units

rnsd.service (shared-instance architecture only):

[Unit]
Description=Reticulum Network Stack Daemon
After=network.target

[Service]
Type=simple
User=user
ExecStart=/home/user/nomadnet-env/bin/rnsd
Restart=on-failure
RestartSec=5

[Install]
WantedBy=multi-user.target

Note Type=simple means systemd considers this "started" the instant the process forks — not when Reticulum Transport is actually ready. That matters for nomadnet.service below.

nomadnet.service — the reliable version:

[Unit]
Description=Nomadnet
# Shared-instance architecture only:
After=network.target rnsd.service
Requires=rnsd.service
StartLimitIntervalSec=300
StartLimitBurst=10

[Service]
Type=forking
User=user
# Shared-instance architecture: fast-path gate on rnsd's socket existing.
# Embedded architecture: use `ExecStartPre=/bin/sleep 10` instead (just letting
# the network/system settle after boot — there's no companion socket to poll).
ExecStartPre=/bin/bash -c 'for i in $(seq 1 30); do ss -xl 2>/dev/null | grep -q "@rns/default" && exit 0; sleep 0.5; done'
ExecStart=/usr/bin/tmux new-session -d -s nomadnet '/home/user/nomadnet-env/bin/nomadnet'
ExecStop=-/usr/bin/tmux kill-session -t nomadnet
Restart=always
RestartSec=5

[Install]
WantedBy=multi-user.target

Why each piece matters, found by actually reproducing the failures with real VM reboots (not just systemctl restart):

• ExecStop=-... (leading -): if the tmux session already died for any reason, the stop command exits nonzero; without the - that gets logged as a hard failure for what's really a no-op.
• Restart=always, not Restart=on-failure, plus a widened StartLimitBurst: on the shared-instance server, nomadnet occasionally exits cleanly (status=0) very shortly after systemd launches it — even after rnsd's shared-instance socket is confirmed already listening. Timing tests showed this is genuinely non-deterministic (sometimes fails within 10s of an rnsd restart, sometimes succeeds within 7s), and it isn't a tmux-state or environment issue either — it looks like an occasional hiccup in the RNS shared-instance handshake itself (rnsd's own log separately shows an unrelated, self-recovering RuntimeError: dictionary changed size during iteration inside RNS/Transport.py, suggesting some thread-safety fragility in the underlying library around connection churn). Because the failure is a clean exit, Restart=on-failure does not retry it — the service just sits dead until someone notices. Restart=always means systemd just tries again a few seconds later regardless of exit code, and it reliably succeeds on the next attempt. The widened StartLimitBurst/StartLimitIntervalSec stop a run of quick retries during boot from permanently locking the unit into start-limit-hit.

Passwordless sudo for systemctl (optional, for remote administration)

Scope it — never grant blanket passwordless sudo:

# /etc/sudoers.d/systemctl-nopasswd
user ALL=(root) NOPASSWD: /usr/bin/systemctl

Always validate before installing (a bad sudoers file can lock out all sudo access):

visudo -cf /etc/sudoers.d/systemctl-nopasswd
chown root:root /etc/sudoers.d/systemctl-nopasswd
chmod 0440 /etc/sudoers.d/systemctl-nopasswd
visudo -c

Note this doesn't cover journalctl (different binary — add it too if wanted), and systemctl edit <anything> opens an editor as root, so this is a convenience grant, not a hard security boundary.

New-install checklist

1. Create the venv, install rns/nomadnet.
2. Pick embedded vs. shared-instance architecture.
3. Install the unit(s) above — default tmux socket, Restart=always, widened start-limit, tolerant ExecStop.
4. systemctl daemon-reload && systemctl enable --now nomadnet (and rnsd first, if applicable).
5. Test with a real reboot, not just systemctl restart — then journalctl -u nomadnet -b (and -u rnsd -b) to confirm a clean start, or at least a successful auto-retry.
6. Confirm tmux attach -t nomadnet reaches the running session.

Licence

MIT — see LICENSE