Installation and first boot
On this page
Before you travel
Three things have to be true before the unit is worth unpacking. Every install that has gone badly went badly because one of them was assumed rather than checked.
- An address is reserved. Either a DHCP reservation against the unit’s MAC, printed on the base, or a static address agreed with the network team. The unit must not float.
- SNMP is reachable from the subnet the unit will sit on to every subnet the printers sit on. See network requirements; this is the part that needs somebody else’s approval and therefore the part that delays the install.
- The community string or SNMPv3 credentials are known. Not “probably public”. Half of the fleets we adopt have been changed at some point and nobody wrote it down.
Take with you: the unit, its supply, a short patch lead for the service port, a laptop with an Ethernet socket or a known-good adapter, and a label maker.
Mounting and cabling
- Clip the unit to the DIN rail or fix the wall bracket. Keep the LEDs visible and the cable run away from the cabinet door hinge.
- Patch the site LAN into eth0 — the socket nearer the power inlet. Label that cable now.
- Leave eth1 empty for the moment. It is the service port and it is permanently 192.168.77.1/24; patching it into the site LAN puts a rogue DHCP-free gateway address onto the network and confuses everyone for a week.
- Apply power. The power LED lights immediately; the status LED blinks blue while the unit boots, for 40–70 seconds.
Reaching the service port
Everything below is done from a laptop on the service port, not over the LAN. The unit has no address on the site network until you give it one.
- Patch the laptop into eth1.
- Set the laptop to 192.168.77.50/24, no gateway. The unit does not run a DHCP server on the service port — deliberately, so that a mis-patched cable cannot hand out addresses on the customer LAN.
- Open https://192.168.77.1/. The certificate is self-signed on a new unit; accept it for this session only.
- Sign in with the commissioning credentials printed on the base. The wizard forces a change before it will do anything else.
If the page does not open, the serial console is the fallback: micro-USB, 115200 8N1, and the same login. From there mluctl net show tells you what the unit thinks its interfaces are doing.
The first-boot wizard
Six screens, in this order. Nothing here cannot be changed afterwards in the console, but doing it now saves a second visit.
| # | Screen | What to enter |
|---|---|---|
| 1 | Administrator password | Replaces the commissioning password. Twelve characters minimum; the unit refuses anything that contains the serial number. |
| 2 | Site identity | Site name and reference. This is what appears on every export and every alert, so use the name the billing desk uses, not the building’s nickname. |
| 3 | LAN addressing | Static address, mask, gateway, or DHCP if a reservation exists. Static is preferred; see addressing. |
| 4 | Time and locale | NTP server and time zone. Meter readings are stamped in UTC internally and displayed in the site zone. |
| 5 | SNMP credentials | One default community or SNMPv3 identity, plus any per-subnet exceptions. Additional credentials can be added later. |
| 6 | First scan | The ranges to discover. Leave “scan now” ticked only if the port table has already been signed off. |
Finishing the wizard restarts the collection service. The unit is then reachable on the LAN address you gave it, and the service port keeps working in parallel — it always does, which is why a lost LAN address is an inconvenience rather than an outage.
Acceptance checks
Do not leave site until all five pass. They take four minutes and save a return trip.
- The unit answers on its LAN address from a machine that is not on the service port.
- Discovery found the expected number of devices. Compare against the site’s own list, not against what the console found — the point is to catch what it missed.
- At least one meter reading exists for every discovered device: mluctl meter read --all --verbose and look for zero failures.
- Time is correct and syncing: mluctl status shows the NTP source and an offset in milliseconds, not seconds.
- A configuration backup exists off the unit. mluctl backup create --to usb, then take the stick with you.
Replacing a failed unit
The order below keeps device identities and therefore keeps the meter history attached to them. Getting it wrong does not lose data, but it creates a second set of device records and the reports silently split in two.
- Take the latest configuration backup, from the old unit if it still answers (mluctl backup create), otherwise from the scheduled copy.
- Install and power the replacement, but do not run a scan in the wizard. Untick it.
- Restore: mluctl backup restore <file>. This brings back device records, credentials, thresholds, release settings and the site identity.
- Give the new unit the same LAN address as the old one. Release printing points at that address from every workstation, and changing it means touching every workstation.
- Only now run discovery. It matches existing records by serial number and MAC, and adds nothing that was already known.
- Confirm continuity: mluctl meter history <device> --last 30d should show readings from before the swap.