Multi-WAN sites
Experimental: this setup depends on the configuration of the site's network. It has not been tested beyond the example below.
At a multi-WAN (wide area network) site, a router combines several Starlink
terminals into one network. Every terminal answers at 192.168.100.1 on its
own connection. A Nexus Telemetry Fleet collector runs in Docker for each
terminal, on that terminal’s virtual local area network (VLAN).
Requirements
| Requirement | Detail |
|---|---|
| Switch | A managed switch with each terminal’s cable on its own VLAN |
| Router | Each terminal’s VLAN also reaches its WAN on the site’s router, through an access port to each WAN port or a trunk to VLAN-tagged WAN interfaces |
| Trunk | 1 switch port with every terminal VLAN tagged, cabled to the collector host |
| Terminals | In bypass mode. A terminal behind its own Starlink router requires a different container setup. That setup is not tested |
| Host | A Linux host with Docker and Docker Compose. Checked with Docker Engine 29.8 and Docker Compose 5.5 |
| Fleet server | The host reaches the fleet server over its own network, through the site’s router |
| Token | 1 enrolment token with Max uses at least the number of terminals, from the Add telemetry collector window. See Tokens for many collectors for a token with more than one use. To place every collector in the site’s group, set Group when generating the token. See Groups for group membership |
Router modes
Each collector reads its terminal on the terminal’s VLAN, before the router. Load balancing, failover and bonding make no difference to the readings. The collector reads a terminal on standby in the same way as an active one. When no connection is available, each collector keeps reading its terminal. It sends the readings when a connection is available again.
Unsupported setups
The setup does not work for:
- Terminals joined to the router over Wi-Fi, or with no Ethernet port.
- Terminals cabled straight into the router with no switch. Insert a managed switch first.
Container
Each terminal has 1 container. It joins the terminal’s VLAN and the Docker
network out, which reaches the fleet server through the host. At start, it
sets 192.168.100.2 on the VLAN as root with NET_ADMIN. On that VLAN,
192.168.100.1 reaches that terminal only. The container then drops every
capability and runs the collector as the unprivileged user nexus. A restart
sets the address again.
Docker refuses two networks on the same subnet. The Compose file creates the
VLAN networks with no Docker addressing, enable_ipv4: false. The container
sets the address itself.
Compose file
The file covers 2 terminals. For each further terminal, add a service, a VLAN network and a volume with the next number.
name: nf-multi
x-collector: &collector
image: registry.nexustelemetry.com/nf-collector:latest
restart: always
env_file: collector.env
user: root
cap_add: [NET_ADMIN]
entrypoint: ["sh", "-c", "ip addr add 192.168.100.2/24 dev terminal && exec setpriv --reuid=nexus --regid=nexus --init-groups --inh-caps=-all --bounding-set=-all nf-collector"]
x-vars: &vars
NF_SERVER: <your-address>
NF_SOURCES: starlink,ping
services:
t1:
<<: *collector
networks:
out: {}
vlan101: { interface_name: terminal }
environment: { <<: *vars, NF_DEVICE_NAME: Terminal 1 }
volumes: ["t1-data:/var/lib/nexus-fleet"]
t2:
<<: *collector
networks:
out: {}
vlan102: { interface_name: terminal }
environment: { <<: *vars, NF_DEVICE_NAME: Terminal 2 }
volumes: ["t2-data:/var/lib/nexus-fleet"]
networks:
out: {}
vlan101: { driver: macvlan, enable_ipv4: false, driver_opts: { parent: eth0.101 } }
vlan102: { driver: macvlan, enable_ipv4: false, driver_opts: { parent: eth0.102 } }
volumes:
t1-data: {}
t2-data: {}Save the token in collector.env.
NF_TOKEN=<token>| Value | Set it to |
|---|---|
parent |
The host’s trunk interface and the terminal’s VLAN ID, for example eth0.101. Docker creates the VLAN interface |
<your-address> |
The fleet server’s address, as in the Add telemetry collector window |
NF_DEVICE_NAME |
The terminal’s name on the dashboard. Required, because a token with more than one use names no terminal |
<token> |
The enrolment token. Every collector enrols with it and receives its own device certificate |
NF_SOURCES leaves out the connection source. Every collector in this
setup connects out through the host. The connection source would describe
the site’s connection, not the terminal’s. See
Collector configuration for
NF_SOURCES.
Run
-
Save the Compose file as
compose.yaml, in one folder on the host withcollector.env. -
Start the containers in that folder.
docker compose up -d
Each terminal appears on the dashboard as its own collector.
Warning: each volume holds that collector’s device certificate. A collector whose volume is deleted enrols as a new collector with a new record.
Readings per terminal
| Reading | Per terminal |
|---|---|
| Terminal status, history and obstructions | Yes |
| Ping to the terminal | Yes |
| Ping to the gateway | No. Gateway in Reachability reaches the Docker network’s gateway, not the site’s router |
| Ping to the internet | No. It uses the connection the site’s router chooses |
| On backup state | Not reported |
Update
To update, pull the new image in the folder and start the containers again. The dashboard cannot update a container. Each collector starts from the device certificate in its volume and reconnects to the same record with the same history.
docker compose pulldocker compose up -dRemove
To run the collectors again later on this host, stop and remove the containers and networks. Docker keeps the volumes. The collectors reconnect to the same records when they start again.
docker compose downPurge
To remove a site permanently, decommission each collector on the dashboard first, while it is online. See Retire and revoke for decommissioning. Then remove the containers and the volumes.
docker compose down -v