Cables and power

Configuration problems are visible and cable problems are not. That is why people spend an afternoon in the firewall while a badly crimped plug hangs in the meter cupboard. Look at this layer first, not last.

The cable test

RouterOS can measure a cable itself, with no extra equipment:

/interface ethernet cable-test ether5

You get a status per pair and, for a break or a short, a distance in metres. It is not a precision instrument, but the difference between two metres and twenty-eight metres tells you whether to look in the patch cabinet or at the desk.

Two limitations: it does not work on every switch chip, and the port goes down during the test. Do not run it on the port you are sitting on, or on the uplink of a live site without telling anyone.

What the results usually mean:

• open on one pair: a conductor that does not connect through. On a self-crimped plug it is nearly always the plug.
• short: two conductors touching, often from stripping too far back.
• all pairs normal while the link still misbehaves: not a break but quality. A cable that is too long, one running alongside mains power, or one pinched in a door frame.

Link speed and duplex

/interface ethernet monitor ether1 once

This shows the link status, the speed it settled on and whether it is full duplex. The fastest way to see that a gigabit port runs at 100 Mbit.

A gigabit link uses all four pairs, 100 Mbit only two. A cable with one broken pair therefore gives you no fault, just a slow connection, and nobody notices until someone copies a large file. See also Measuring.

The configurator does not fix speed or duplex, and there is no field for it. Auto-negotiation works, and pinning one side while the other negotiates is a classic way to end up at half duplex. If equipment genuinely needs it, set it by hand on both ends.

What the tool does know is the speed of every port on every model in the catalogue. That lets it say on the network board that your gigabit line is in a 100 Mbit port, naming the port it belongs in. See Checks about cabling.

What a bad cable looks like in the log

A cable that fails rarely fails all at once. What you see is a pattern:

/log print where topics~"interface"

• link up and link down, repeated. Once is someone pulling a plug. Twenty times in an hour is a cable, a plug or a port.
• link up at a lower speed than expected. The line names the speed, so a link up (speed 100M, full duplex) on what should be a gigabit link is there in black and white.
• flapping that coincides with load. That points at power rather than data.

Counters tell the rest of the story:

/interface ethernet print stats

Look at rx-fcs-error and the other error counters. A handful of errors since boot is normal; a counter climbing visibly while you watch is not. They count since the last reboot, so put them next to the uptime. See Logs for keeping messages like these.

PoE-out

PoE-out means the port delivers power over the same network cable. You set it per port in Bridge and ports, with the field PoE-out per port; it only appears when your model has ports that can do it, and lists exactly those ports. See PoE-out per port for the models and the modes.

On the device you look at it like this:

/interface ethernet poe print
/interface ethernet poe monitor ether2 once

The monitor gives status, voltage, current and the power going through right now. That last number is what matters here.

The budget

Every device with PoE-out has a total budget, and it is less than the number of ports times the maximum per port. A switch with eight ports that can each deliver thirty watts does not carry two hundred and forty watts.

The configurator does not know that budget. It knows which ports have PoE-out, but not how many watts the device can deliver in total and not what your cameras and access points draw. So there is no check telling you that you are over it. Add it up yourself from the specifications of your device and of what you hang on it, and leave some room.

What happens when you go over depends on the device and is never tidy. The friendly version is that the port you plugged in last stops delivering power. The unfriendly one is that the device reboots under load, which shows up as an uptime that keeps starting at zero. That looks like a failing power supply and is often taken for one.

Two things quietly eat the budget: long cables, because voltage drop over fifty metres of copper is real, and a power supply smaller than what the device can handle.

The modes

The tool offers the three modes RouterOS has. auto-on is the default and the right choice: the port negotiates and only delivers power when it recognises a device at the other end asking for it. off delivers nothing. forced-on puts the voltage on without asking, and that is the mode to be careful with: the other end might be a laptop.

If a device does not come up on auto-on, that is usually passive PoE that does not announce itself properly. Check the voltage the device wants first, because forced-on plus the wrong voltage is an expensive combination.

The round you make

1. /interface ethernet monitor: is every port at the speed you expect?
2. /interface ethernet print stats: are errors climbing?
3. /log print where topics~"interface": is anything flapping?
4. /interface ethernet poe monitor: does the draw match what you hung on it?
5. A suspect port? cable-test, and only then a new patch cable.

Read on: PoE-out per port, Checks about cabling and Logs.