One RB5009 powering another

RouterOS General
1

I have two RB5009 PoE routers that are connected using an UTP cable. Each of them have their included 48V power supply.

As one of them is on a UPS and the other is not, I thought it may be a good plan to power the second one from the first, so when the local power fails it is still powered via the UTP cable (from the UPS on the first).

However, the result is a little confusing. When I select “auto on”, the port on the first router remains “waiting for load”. Apparently it does not recognize the second router (ether1) as a PoE load.

When I select “forced on”, the power is applied and a very low current flows (7mA, 0.2W) which does not surprise me, as the local supply still supplies the power and its voltage is likely a bit higher than the circuit via the UTP cable.

However, in the second router there is also the output of /system health print:

#  NAME                 VALUE  TYPE
0  poe-out-consumption  6.9    W   
1  cpu-temperature      41     C   
2  jack-voltage         48.9   V   
3  2pin-voltage         0      V   
4  poe-in-voltage       0      V

I don’t understand why poe-in-voltage is still 0V in this case, while on the other router the poe-out-voltage is 48.8V on the port (ether2).

How is this supposed to work?

2

I believe that until you disconnect on the second router the power in jack, it won't get power from PoE.

In PoE parlance the first RB5009 acts as PSE (Power Sourcing Equipment) and the second as PD (Powered Device), according to 802.3af/at, there is a "conversation" happening that more or less goes along the lines of (just invented):
PSE: Hallo there, are you a Poe device?
PD: Yes I am.
PSE: Are you ok with 48v?
PD: Yes I am.
PSE: Are you ok with power being delivered as Mode B?
PD: Yes I am.
PSE: Good, so, how much Amperes (or Watts) do you want?
PD: Thanks I need not any, I am powered from my local power source.
PSE: Ah, ok, then I won't deliver any current.

So, the readings you have on the second RB5009 (PD) should mean that there is no voltage on PoE-in as it has not been requested.

And the ones on the first RB5009 (PSE), once you set it as forced on should mean that it is delivering 48V BUT no (or very little) load is connected.

Since the RB5009's can also be powered at 24V, you could try using a 24V on the second RB5009 to see if the (now higher) almost 48V over PoE are preferred.

3

Ok well then I think when the local power on the second RB5009 fails, it should switch over to PoE powering, right?

Hopefully that happens quickly enough that it will not reboot. If it does, that would not be fatal either, just not preferred.

4

I seem to remember a discussion about some other MT device type, where owner tried same setup: PoE-in and local powering. And the behaviour was as follows: if both power supplies were fine when device booted and then local power failed, device rebooted. After local power was restored ... and PoE power failed, device kept running. Restoring PoE power meant device behaved as from initial state (rebooting if local power failed).
I don't think that device offered health monitoring of power sources so there were no measurement data mentioned in discussion.

So conclusion was suspicion that PD doesn't perform PoE negotiation if already powered (from local source). If powering on from PoE, then it performed PoE negotiation and kept PoE going also when local power was added.

@pe1chl you may want to experiment in this direction.

5

Well yeah I do not intend to experiment much, it works as it is and when we are lucky it will survive a power drop, when unlucky with a reboot.

I only came to this because today there was a power glitch at another location where we do not have a UPS. The glitch was so short that the RB5009 rebooted but an Aruba switch and a Raspberry Pi survived it without rebooting…

6

I don't think it will be fast enough, maybe with "forced-on", but I wouldn't count too much on it.

7

Well I don’t count on it… it is just to provide better service to the people there, but there is no guarantee and no need for it. They could always work to the part of the building served by the main router. (both power a Unifi access point)

I only thought of it as an interesting solution. It would have been great when the second router just negoitiated being powered via PoE and then fed that power via the usual diode wired-or together with the local power supply into the router, so it would be fast enough.

That is how it works with older devices that have passive PoE. Of course in that case the direction of powering would be determined by how you connected PoE-out on one device to PoE-in on another, and connecting two RB5009 in default configuration (PoE auto on) would result in an unknown powering direction. But I would be happy to configure PoE to “input” when that was required.