Every lab opens locked, so a stray click cannot move nodes: press the lock in the toolbar (or "Unlock to edit" in the hint at the top) to edit. The hint goes away for good once you have unlocked a lab.
A lab in a folder that is not one of your workspaces is read-only: the backend only writes inside workspaces. The UI says so and offers Add folder as workspace, which adds the lab's folder and reopens the lab editable. Workspaces can also be managed from the folder browser.
| Keys | Opens |
|---|---|
| Ctrl+P | everything: labs, nodes, actions, and search inside the lab. Context actions (run on nodes, live traffic, running configs) come first while a lab runs. |
| Ctrl+` (the key can be changed in Settings → General, e.g. for a German layout) | Show or hide the bottom panel; with nothing open in it, start a terminal |
| Ctrl+Shift+` (same key as above) | Another terminal: your normal shell in the lab's folder (on your host when netlab-ui runs in a container). The + in the panel does the same. |
| Ctrl+I | AI agents panel (connect your agent over MCP, review its proposals) |
Run Command on Nodes… (right-click a deployed lab, or the stacked-cards
button in the terminal dock) opens a dock tab that sends one command to any
mix of nodes, groups and all, in parallel:
- Auto (the default) sends
show …to each device's CLI (netlab connect --show: vtysh, Cli, sr_cli…) and everything else to its shell (netlab exec; pipes work); Linux hosts always get the shell. Shell / Show force one. - Nodes selected on the canvas are preselected;
allmeans the running nodes. - Patterns, for labs too big to click through. Type into Run on and press Enter:
r1-r3orr1-3(a numeric range),r[1-3,5](the list form shown on result cards),leaf*/h?/pc#(*any text,?one character,#digits, as in the bulk-link dialog),leaf01-leaf16(zero-padded), or a regular expression such as(spine|border)-[ab](must match the whole name). Patterns can be mixed with names and groups, a line under the field shows how many nodes they match, and the backend does the matching, so a range over a hundred thousand nodes is one request. A name that exists always wins over reading it as a range. The AI assistant'srun_show_commandtakes the same patterns. - Quick chips offer your last commands plus the ones that fit this lab
(
show ip ospf neighboronly if it runs OSPF…). One click runs one; Shift-click puts it in the command box to edit first. - Answers appear side by side, and identical answers are merged
(
r[1-3]) so the odd node stands out. Non-zero exits and CLI error replies (% Unknown command) are flagged red. - ↑/↓ browse the command history, and ↻ re-runs a command.
- Save session as script stores the commands next to the topology in
<topology>.netlab-ui-scripts.json, so the script is part of the lab. The ⏵ scripts menu replays one; Export transcript downloads the session as Markdown.
Ctrl+P searches netlab's transformed topology as well as labs, nodes and actions. That includes addresses netlab assigned from its pools and modules a node inherits from a group.
| Type | Finds |
|---|---|
10.1.0.5 |
the interface holding it, or whose subnet contains it |
10.1.0.0/16 |
every address inside the prefix |
as 65001 |
BGP AS |
10 |
VLAN id, AS or node id |
module:ospf, vlan:, vrf:, group:, device:eos, role: |
that kind only (module: alone lists all modules) |
Picking a hit spotlights its nodes: everything else on the canvas is dimmed. The spotlight bar has one chip per module, so you can switch the canvas between "who runs OSPF / BGP / VLANs"; press Esc to show everything again.
Running Configs & Changes… (right-click a deployed lab → Inspect, or the 📜 button in the Lenses panel) shows what the devices actually run.
- A snapshot of every node's
show running-configis taken automatically after each successful deploy, restart ornetlab initial. You can also take one by hand. - The node list shows drift: lines added and removed since the chosen snapshot. Volatile header and timestamp lines are ignored.
- A side-by-side diff compares any snapshot with the live configuration, or
two snapshots, e.g. to see what a
netlab initialor a hand edit changed.
Snapshots live in <lab>/.netlab-ui/configs/<topology>/; the newest 20 are
kept. The generated (not running) configuration per module is still in the
node editor's Config tab.
Guided tours (Lenses → ▶ tour button) are now also exercises. A step can have:
- a task for the student,
- checks:
netlab validatetests from the lab'svalidate:block, picked from a list with their descriptions, - a hint that is shown after a failed check.
In presentation mode the student presses Check my work. Only that step's
tests run, and the step passes when all of them pass; failures show netlab's
own message and the hint. A progress bar counts passed steps; it is kept per
browser, and clicking it starts over. Everything is stored in the lab's
<topology>.netlab-teaching.json, so a lab and its exercise travel together
(e.g. via Copy Lab to Workspace…).
Click a link on a deployed lab to see its traffic chart (clab-ui's link panel). The Traffic lens (Lenses tab) shows every link at once: links are coloured and thickened by load and labelled with their rate. A link that is down turns red and dashed; one that is dropping or erroring right now turns red. The panel lists all links, busiest and broken ones first. Counters are read from the containers every 3 seconds.
- Right-click a link → Take Link Down / Bring Link Up. This pulls the
cable (
ip link set … downin the container), so the peer loses carrier and routing protocols react as they would to a real failure. - Right-click a link → Link Impairments sets netem delay, jitter, loss,
rate and corruption per endpoint. The form now shows what is actually
applied, read back from
tc. - In the Traffic lens, each link row has a ⚡ menu: down/up, 100 ms delay, 5 % loss, 1 Mb/s cap, or clear (applied to both ends).
Impairments need the lab host's sch_netem kernel module (sudo modprobe sch_netem if the UI reports that netem is missing).
The Wireshark items on a link's (or node's) context menu open a capture dialog with three ways to capture:
- Download a pcap of 10 s to 5 min. It captures both directions and keeps VLAN tags. No Edgeshark or tcpdump is needed: the backend enters the node's network namespace itself.
- Live in your own Wireshark: copy the
curl -sN '…/api/lab/capture/pcap?…&seconds=0' | wireshark -k -i -command. - Wireshark in the browser via Edgeshark. If Edgeshark isn't running, the
dialog offers Install Edgeshark: it starts netlab's own
edgesharktool (the same containerstools: [ edgeshark ]would start), pulls the Wireshark image, and checks the containers stay up (internet needed, a few minutes the first time). Edgeshark's capture port 5001 and the Wireshark web page are published on the UI's own bind address (127.0.0.1by default), not on every interface. SetNETLAB_APP_CAPTURE_BINDto change it.
In a regular node shell, ⇄ Sync input mirrors your typing into every other shell that has sync switched on (like tmux's synchronize-panes).
netlab has three providers, and every lab can mix them (provider: on a node
overrides the lab's). The UI works with all three. A lab without containers
gets its canvas straight from netlab's transformed topology, so VMs show
netlab's real interface names (Ethernet1, GigabitEthernet0/1). Settings →
Environment has the same table as below; hover a mark there for the reason.
| Feature | containerlab | libvirt VMs | external devices |
|---|---|---|---|
| Draw/edit, validate, preview configs, deploy diff | ✅ | ✅ | ✅ |
| Deploy / destroy | ✅ | ✅ ¹ | ◐ configs only |
| Web shell, run on many nodes, running configs | ✅ | ✅ (SSH) | ✅ |
| Start / stop / restart / pause a node | ✅ | ✅ virsh |
— |
| Link down / up | ✅ | ✅ domif-setlink |
— |
| Live traffic and errors | ✅ | ◐ ² | — |
| Packet capture | ✅ | ◐ ² | — |
| Link impairment (delay, loss, rate) | ✅ | ◐ ² via netlab tc |
— |
| Node logs | ✅ | — | — |
| Image manager | ✅ | — (Vagrant boxes) | — |
¹ Needs KVM, libvirt and Vagrant on the backend host. The container image has
none of them; run the backend natively, or use the UI-only image with a host
netlab install.
² Link state works on every VM link. Counters, capture and impairment work on
LAN links only: netlab builds point-to-point VM links as UDP tunnels with no host
interface (the same limit netlab capture has).
Where something isn't possible for a node, the UI says why instead of failing.
- External tools (Ctrl+P → External tools, or a lab's context menu):
Graphite, SuzieQ, NUTS, Cisco NSO and Edgeshark from netlab's
tools:. Switch one on and netlab starts it with every deploy. For a deployed lab, see whether it runs, open its web UI, connect to its CLI (netlab connect suzieq) or start and stop it. The commands are netlab's own. - Monitoring (Ctrl+P → Monitoring, or a lab's context menu): turns on
the netlab
monitoringplugin shipped with the UI. It shows the lab's health against what the topology defines (BGP sessions, OSPF and IS-IS adjacencies up vs expected, and which are missing), how every node is collected, and opens the Grafana dashboards. Fault tests flap one or more links on a fixed schedule (cycles, seconds down, seconds up) and measure each cycle: how fast the lab noticed, how many sessions went down, and how long recovery took. Results are saved with the lab for comparison. Link outages and fault tests show as shaded bands on the dashboards (toggles at the top of each dashboard). - Reports (Ctrl+P → Reports): every netlab report as a table, rendered
HTML or text, with download in each format netlab offers (
.md,.html, text) and Open for the HTML version. The HTML is sandboxed: scripts don't run and it doesn't get the UI's origin. - Custom configs: the node editor's Configuration tab lists the
templates netlab would find for
config: [name](lab directory,~/.netlab,/etc/netlab) and warns when one has no variant for the node's device. New template creates<name>/<device>.j2with a commented starter and opens it. - Containerlab tarball (a deployed lab's context menu, or Ctrl+P):
netlab clab tarballdownloadsclab.config.ymlplus the devices' current configs, to run with plain containerlab elsewhere. - Setup helpers (Settings → Environment):
- Check my setup runs
netlab testclab|libvirt|podman|grpc: a real deploy of a tiny lab, checked and removed again. Stopping it halfway still tears everything down. - Install software runs
netlab install(Ansible, containerlab, libvirt, GraphViz…) on the backend host. It needs root or password-less sudo. - Build routing-daemon containers runs
netlab clab build(BIRD, dnsmasq…). - Vagrant box recipes show
netlab libvirt config <device>.
- Check my setup runs