nl6 — network device simulator

📖 Documentation: https://labmonkeys-space.github.io/nl6/

A scalable network and infrastructure simulator that exposes realistic SNMP v2c/v3, SSH, and HTTPS REST interfaces for testing network management software, monitoring systems, and automation tools. nl6 can simulate tens of thousands of network devices, GPU servers, storage systems, and Linux servers — each with its own IP address via Linux TUN interfaces and network namespaces.

Highlights

• Runs 30,000+ simulated devices on a single host — see Scaling.
• 28 device types across 8 categories (core / edge routers, DC and campus switches, firewalls, servers, NVIDIA DGX/HGX GPU servers, and enterprise storage) — see Device types.
• Multi-protocol per device: SNMP v2c/v3 (MD5/SHA1 auth, DES/AES128 privacy), SSH with VT100 terminal emulation, and HTTPS REST — see SNMP reference and Web API.
• Realistic dynamic metrics: CPU / memory / temperature on 100-point sine waves; full IF-MIB counter cycling (octets plus per-direction unicast / multicast / broadcast packet counts, errors, discards) with per-device error-scenario tuning (clean / typical / degraded / failing); per-GPU DCGM metrics — see SNMP reference → Dynamic IF-MIB counters and GPU simulation.
• Self-reporting version: ./nl6 -version, GET /api/v1/version, and a hero-kicker (vX.Y.Z) in the web UI all report the running build — no source checkout needed to identify a deployed simulator.
• Per-device flow export (NetFlow v5 / v9, IPFIX, sFlow v5) with per-device source IPs — see Flow export.
• Per-device SNMPv2c trap / INFORM export — central Poisson scheduler with a global rate cap, a user-overridable JSON catalog, and per-device UDP source IPs. Suited to OpenNMS trapd scale testing. Configure with -trap-collector <host:port>; full flag list and catalog schema in CLAUDE.md → "SNMP trap export".
• Per-device UDP syslog export (RFC 5424 / RFC 3164) — central Poisson scheduler with a global rate cap, user-overridable JSON catalog, and per-device UDP source IPs. Ships six generic entries (interface up/down, auth success/failure, config change, system restart) spanning local7 and authpriv; select format with -syslog-format 5424|3164. Suited to OpenNMS syslogd scale testing — configure with -syslog-collector <host:port>; full flag list and catalog schema in CLAUDE.md → "UDP syslog export".

Status & scale

Stable — SNMP v2c/v3, SSH, HTTPS REST (storage APIs), NetFlow v5/v9 and IPFIX, TUN-per-device scaling with opensim network-namespace isolation, web UI, REST control plane.

Experimental — sFlow v5 (synthesised from FlowCache records with a fixed sampling_rate; suitable for collector-plumbing validation, not link-utilisation benchmarking — see Flow export reference → sFlow caveat).

Tested scale — up to 30,000 concurrent simulated devices on a single host. Toolchain — Go 1.26 or later; canonical version pinned in go/go.mod.

Quick start

git clone https://github.com/labmonkeys-space/nl6.git
cd nl6/go/nl6 && go build -o nl6 .

# Auto-create 5 devices starting at 192.168.100.1
sudo ./nl6 -auto-start-ip 192.168.100.1 -auto-count 5

Then query any device:

snmpget -v2c -c public 192.168.100.1 1.3.6.1.2.1.1.1.0
ssh simadmin@192.168.100.1                            # password: simadmin
gnmic -a 192.168.100.1:9339 --skip-verify capabilities

Per-device exports (flow + trap + syslog in a single create call):

# Boot without any export CLI flags — the subsystems are always-on.
sudo ./nl6

# Create 10 devices that all emit IPFIX flows, SNMPv2c traps, and
# RFC 5424 syslog to one collector. Any of the three blocks is optional.
curl -X POST http://localhost:8080/api/v1/devices \
  -H 'Content-Type: application/json' \
  -d '{
    "start_ip": "10.0.0.1",
    "device_count": 10,
    "flow":   {"collector": "192.168.1.10:4739", "protocol": "ipfix"},
    "traps":  {"collector": "192.168.1.10:162",  "mode":     "trap"},
    "syslog": {"collector": "192.168.1.10:514",  "format":   "5424"}
  }'

# Inspect what each subsystem has attached.
curl http://localhost:8080/api/v1/flows/status   | jq '.data.collectors'
curl http://localhost:8080/api/v1/traps/status   | jq '.collectors'
curl http://localhost:8080/api/v1/syslog/status  | jq '.collectors'

Full walkthrough: Getting started → Quick start. Container deployment: Getting started → Docker.

Documentation map

The docs site has four top-level sections:

• Getting Started — build, first run, Docker.
• Operations — scaling, network namespace, flow export, troubleshooting.
• Reference — architecture, CLI flags, web API, device types, SNMP, flow export, resource files, GPU simulation.
• GPU simulation — NVIDIA DCGM OID layout, per-GPU metrics, and the pollaris / parser integration notes (formerly plans/).

Reference content that used to live in this README now lives in the docs site. A bare README.md on GitHub is intentional: the site is the canonical home.

Contributing

Contributions are welcome. Two project policies apply to every patch:

1. Sign off every commit (Developer Certificate of Origin). All commits must carry a Signed-off-by: trailer certifying the DCO. Use -s on every commit:

git commit -s -m "your commit message"

A DCO-check gate will fail any PR whose commits are missing the sign-off trailer.

Suggested workflow

1. Fork labmonkeys-space/nl6.
2. Create a feature branch.
3. Make your changes and add / update tests.
4. Run make check-tidy && make build && make test locally.
5. git commit -s each commit.
6. gh pr create --repo labmonkeys-space/nl6 --base main.

Cutting a release. Maintainers: see RELEASING.md for the tag-driven release workflow and the short post-tag verification checklist.

License

Licensed under the Apache License, Version 2.0. See LICENSE for details.

Heritage

Originally forked from saichler/l8opensim; see commit history for upstream attribution.

---

nl6 — simulate networks, test at scale, develop with confidence.