Home › Documentation › Usage
Usage
gnulte-doctor and read authorised use.
Quick start
sudo gnulte # guided menu: target -> conditions -> run
sudo gnulte -t 192.168.1.50 -l 300 -j 30 --duration 90 # 300ms latency + 30ms jitter, 90s auto-stop
sudo gnulte -r 192.168.1.0/24 -w 192.168.1.100 --profile throttle --duration 600
gnulte-lan # six-screen live watch (auto-discovers subnet)
gnulte-lan -t 192.168.1.20 --duration 60 # one host, up close
gnulte-scan --deep # sweep + in-Go port scan + OS fingerprint
gnulte-scan --watch 5 # live re-scan: ▲ new ▼ gone ~ changed
gnulte-scan -T # tabbed browser: Devices / Log / Summary
gnulte-devices -s # instant ARP inventory + optional sweep
Recipe: shape one host and keep the evidence
A labelled run that auto-stops, records the traffic and files a report:
sudo gnulte -t 192.168.1.50 -l 250 -j 40 -p 2 \
--duration 300 --note gaming-latency-test-3 \
--capture session.pcap
The dashboard shows per-second ping results while the run lasts and a live
netem live line straight from tc -s qdisc; when it ends,
a session summary (min/avg/p95/max, jitter, loss distribution, a rough E-model MOS)
is printed and the HTML report lands in ~/GNULTE Reports/.
Recipe: watch first, then test
The tools are designed to hand off to each other. Start a watch, pick the noisy
host, and press ⏎ to open gnulte -t <ip> in a
terminal window with that host already loaded:
gnulte-lan # discover and watch the whole LAN
# ... find the host, press ⏎ to test it, or x to save the watch to the device store
The device store is shared: gnulte reads it with w in
the target picker, and gnulte-scan marks devices it already knows with
*.
The LAN-watch screens and keys
| Screen | Shows |
|---|---|
| 1 hosts | every host with live down/up speed, packet rate, latency and a health grade; one stable colour per IP |
| 2 talkers | top talkers ranked with bars and peer counts |
| 3 flows | per-pair conversations (needs the root capture socket) |
| 4 neighbours | the ARP/neighbour table |
| 5 map | the live interconnection map: devices as nodes, live pairs as edges (root) |
| 6 history | the session's alarms beside today's recorded up/down transitions |
Keys: ↑↓ browse ·
⏎/g hand off to gnulte -t <ip> ·
Tab/d detail ·
1–6 (or t/f/n/m) screens ·
x save the watch to the shared store ·
s sort · a alarm-only · o settings ·
h help · q quit.
Recipe: motion detection with jq
Several tools emit machine-readable feeds. gnulte-scan --watch N --json
prints one JSON object per sweep — a {"kind":"baseline"} object, then
{"kind":"delta"} objects with the movement — while the narrative goes to
stderr:
gnulte-scan --watch 5 --json | jq 'select(.kind=="delta")'
gnulte-lan --json | jq . # JSON-lines: baseline, per-second rates + alerts
gnulte-top --json # one bandwidth sample as JSON
gnulte-trace --json # the finished path table as JSON
gnulte-doctor --json # the pre-flight report as JSON
Recipe: find what a slow network is doing to a path
sudo gnulte-trace 1.1.1.1 # live hop-by-hop table with loss per hop
sudo gnulte-trace -m 20 --html trace.html # bounded, with a self-contained report
Ending a run
The first Ctrl+C restores the screen and lets the shutdown finish —
previous qdisc and ip_forward value restored, ARP caches repaired. A
second Ctrl+C (or a grace-period timeout) forces the exit if a teardown
really is stuck. Do not kill -9 a shaping run: it leaves targets
mid-spoof.
Per-tool flags and options: the tool directory and the command reference.