RB5009UG+S+IN can be powered in 3 different ways:
a) PoE-in from Ethernet port #1
b) DC Jack
c) 2-pin terminal on the side
Can I use any combination of the available power options to power the RB5009UG+S+IN in case when a power source fails ?
Yes. From what I recall reading, the router will prefer the higher voltage power source and fall back to the next source should the first fail/ be disconnected.
Very good news, similarly the CRS125-24G-1S-IN
has 2 DC inputs (DC jack, PoE-IN), can I assume that the same holds for it?
It seems that all MT devices with multiple power inputs behave the same way as to powering of the device itself. It seems that all power inputs are connected to common “power bus” via diodes which make the “voltage selection” process very natural and independent of any software/firmware/… process running by any device component.
Devices able of PoE out mostly follow the same principle when providing voltage on PoE out port, only a few allow to explicitly select power source per port. Those models require different voltages on different power inputs so that they can provide low/high voltages on PoE out and in case one of power sources becomes unavailable, also PoE out ports become unpowered.
One more question, can I power the CRS125-24G-1S-IN from the POE port using directly mikrotiks 24V power supply by connecting it through the rj45 by building something like the following (short of)?
I just want to power it from dual DC sources from UPS in case of power failure and I don’t want to waste extra power to the POE injector and splitter (dc to dc conversions e.t.c), I will loose a port of course (technically I can use it for 100 Mbits).
I think it is OK since the CRS125-24G-1S-IN has passive POE and just waits for the 8-28V DC in the RJ45 POE port.
Not sure about DIY PoE injectors. You may want to check MT’s own RBGPOE, it’s pasive by its own. You can even use it as PoE “ejector” if the other end of UTP cable is compatible (e.g. when using a pair of RBGPOEs).
I just had mine die when It should have been powered by both the normal barrel jack and PoE.
I had a power outage/pulled the jack and the router failed and died due to power outage.
It DID reboot and restarted again with the power of PoE, but it was not instant and caused service interruptions.
How can this be prevented in the future?
- How does it choose it’s power source when several are provided and how can we make it to failover silently to the next available?
I just had the opportunity to test this and it is true. RB5009 does not utilize multiple power inputs as one would expect.
When 24V is applied via DC/barrel, then RB5009 will not request PoE (CRS354 saying “waiting for load”). I can force PoE, either high or low, which will then work fine, but that creates a risk that someone will plug non-poe device in the future and fry it.
If I connect PoE first and only then I add DC, then RB5009 keeps running from PoE indefinitely and will correctly failover to DC when PoE is lost. (but then it will not failover back when PoE is available again)
Hopefully this is just firmware/software update - so that RB5009 requests PoE no matter if DC is applied or not.
Yes,
If you use poe as an alternate power source, it must be passive/forced on poe or the Mikrotik may lose power when main/other power is lost. (Usually briefly but…)
@vecernik87
I think this is the “normal” way for 802.3af/at, i.e. it is “by design”, the “logic” is in the PSE.
There must be negotiation between the Power sourcing equipment (PSE) and the powered device (PD).
The PSE should normally (when the cable is disconnected) provide no power, then it will probe and check if there is some resistance using a very low voltage, only if it finds the right resistance (25 kΩ) it will provide the higher 48-57V voltage, and receive from the PD info about the amount of power (amperes) the device requires, then it will continuously monitor the current and if it is below 5-10 mA for more than 400 ms, it will remove power (for all the PSE knows it could be due to the cable being disconnected or cut and for safety reasons the power is then cut off).
From what I have read here and there, Mikrotiks will always use the higher voltage among the ones supplied.
When you power the device with the PoE first, and later add the 24 V jack power, the RB5009 will use the PoE.
If this power source fails the Mikrotik will use the DC in (which is already and always “on”).
When you instead power the device with jack first, the Mikrotik get its power from the 24 V and doesn’t “ask” for power on the PoE in port.
The RB5009 should “lie” to the PSE and ask for power on the PoE even if it is already powered and then switch to the PoE input (drawing some power from the PoE even when it is still powered by the 24 V).
Likely this cannot be done in “software” only, it would probably need some changes in the hardware/circuits. 
The logic of sending power is in PSE, but logic of requesting power is in PD. And you pointed it right that that RB5009 should lie and ask for power even if it is already powered by DC.
The problem is that user may incorrectly understand the failover mechanism. Lets say I power it with PoE first, then add DC. At that moment it is powered by two sources. If any of them fails, device continues to work. If PoE fails, router will remain working, but after power is restored, RB will not request it again and next time DC fails, RB5009 will reboot.
I think this is a bit counterintuitive and it should be at least clearly communicated, that PoE from active PSE (without forced output) can’t be used for redundancy. Keep in mind that promotional video lists power redundancy as one of main selling points of RB5009.
Yep 
, it is definitely counterintuitive, but you talk like all the rest of Mikrotik documentation and product characteristics and features were clear, simple and accurate. 
My doubt is that, even if the good guys at Mikrotik were willing to change the way the RB5009 behaves as PD, after the “initial lie” the RB5009 must draw some (little) power from the PoE to “keep alive” the power from the PSE and this might not be possible to change in software only, it is the “transition” phase that may need some changes in the hardware.
I don’t think that the risk of connecting by mistake a non-48V PoE compatible device (when setting the PoE PSE to “forced-on”) and frying it is that much relevant, putting a label on the ethernet cable near the RB5009 to the effect of “Live power at 48 V, DO NOT move this cable to other ports and DO NOT connect any device to it (unless you know where your towel is)” should be enough.
It is more the “generic” risk of what happens if the cable is cut (or eaten by rats, etc.) and shortcircuited, the continuos checking by the PSE has been evidently implemented in the 802.3xx standadf as a safety measures, if anything is sensed as “not right” the PSE cuts off power within 400 ms or so.
On the other hand the world is full of networks where passive PoE (either at 24 V or at 48 V) is used and it doesn’t seem like it is a common occurrence that these installation create issues.
The way most MT devices with multiple power inputs work is that they draw current from the source with highest voltage. If RB5009 does the same (and I don’t see why it wouldn’t), and if PoE is 48V (or there about, whichi is what I’d expect from 802.3 af/at PSE), then it’s pretty likely PoE would win in the voltage “battle” (e.g. supplied power adapter is a 24V one).
In case where RB5009 decides to draw power voia another source and there’s no power draw on PoE port, then even if PSE decides to cut the power and RB5009 is changed to request power (again), this would mean “power request loop” … which in absolute worst case would make PSD to quit providing power … in bad case power outage occurs during the “down time” and in best case piwer outage occurs during “up time” and in this case PSE continues to provide power due to provided power draw.
That is a good point. hmmm… I really really wish it was just a firmware stuff.
aaand that is the story of “how was vecernik allowed to buy a switch with active PoE instead of using cheap passive injectors?”. I had the label on, director of the company saw that, panicked and the rest is history… Not sure how to justify existence of the label after spending almost $1.5k
@mkx:
Yeah, it should but it does not (unless poe is forced)
Please check my sanity if I am misunderstanding something:
- DC 24V applied to RB5009UG+S+IN first, consumes 9W. PoE on CRS354 in PSE role set to “auto” (i.e. not forced) and “high” (aprox. 52V), applied as second.
- Both devices show “link ok”, negotiate 1GBit and data is flowing fine (so I know the cable is plugged right with all 8 wires). CRS in PoE tab shows “waiting for load”, RB5009 draws power from DC, yellow LED on RB5009’s Eth1 (which should signal PoE) blinks at a regular slow intervals about 2Hz which IMHO corresponds with af/at sending low power while testing for the load.
- Set CRS into “forced” mode → RB5009 accepts PoE, CRS shows 9W consumed. DC power goes almost to 0W (below measurable value)
- Set CRS back into “auto” mode → PoE is dropped, DC power again consumes 9W. Basically same state as step 2
I think the “voltage battle” in reality works only for passive/forced power inputs.
Thats really good thinking! I guess that might be the reason behind the logic “if DC is applied, do not request PoE at all”. Although the loop would happen only if DC is same or higher voltage than PoE, it would require RB5009 to actively measure and compare input voltages in order to decide, if it PoE can be requested or not and to prevent loops. And we know well that is not happening (and possibly can’t even happen becaues active PoE does not send full voltage until the discovery and classification ends)
Cool, that explains everything. I guess it would be good if mikrotik described it better (to prevent confusion and possibly reboots for other users) but there are worse problems than this one.
afk, printing new label
Bullshit. 
No director of a company will ever look at the cables of a router/switch. 
Anyway - in case - invoke safety, cite vaguely how in Sweden (and in Latvia) it has already become compulsory to place such a label on any cable carrying more than 36 V (here you can start telling how DC is safer than AC, how Edison once electrocuted an elephant in Coney Island with AC to show how dangerous it can be compared against DC, but that there are recorded cases of severe burns caused by low voltage DC and it is better to be safe than sorry…) 
Oh, and while you are at it, procure some silicone/plastic RJ45 dust covers and place them on all unused ports.
Not really, in this case the RB5009 is passive, it suffers the PoE voltage that is applied to it.
Well, as already mentioned, that’s up to PoE communication between PSE and RB5009 … if PSE doesn’t provide power (even if it’s just because RB5009 doesn’t request it), then for power consumption discussion PoE in is a missing power input.
I’d verify (with an oscilloscope) what happens with “ordinary” 802.3 af/at PD connected to that device (I guess it could even be RB5009 without any other power source connected) while PoE setting is toggled between “forced on” and “auto” … it might be that CRS (as PSE) first drops power output and then starts PoE negotiation (or waits for incentive from the other end) … which doesn’t happen due to RB5009 behaviour. And if that’s so, “voltage battle” inside RB5009 can’t start (due to missing opponent).