So I purchased 5009 and I don't want to experiment as it is in prod but there is an important question

Unit has 3 possible way of powering - normal jack, dc green jack, poe on eth1 - what is the priority? so I would like to have it run on PoE and if PoE fails it would do go green jack, but when I did one try before disconnecting dc power powered down 5009 and it went back later on poe but it was not uninterrupted action.

Can someone deliberate on this?

What are the three voltages?

My bet: it goes highest to lowest.

But that’s just some EE horse sense, not a fact resulting from having tested it on actual hardware.

Higher voltage has greater priority...
Read here viewtopic.php?t=153113#p756002
So i guess the same applies to RB5009...

Hello,
Yes, nothing haven't changed, as @Zacharias mentioned.
If you want it to work as completely redundant, then poe-in must be passive/forced mode.

but when I did one try before disconnecting dc power powered down 5009 and it went back later on poe but it was not uninterrupted action.

As far as I understand, the different power options are for redundancy, not increased availability.
Meaning if one source will fail, the unit will restart (re-powered from another, available source) but service will be interrupted.
For HA, you'll need a unit with redundant, hot swap PSUs.

Incorrect, I’ve powered mine with the AC adapter and with passive PoE/PoE+ at the same time. When I have removed either input, the unit has not powered down but continued to run with no interruption whatsoever.

...cool thing.
Were all of the supplied voltage inputs of the same level, like all 24V or all 48v?

Power source with highest voltage will be taken as priority. It is the same way on all MikroTik products that have multiple ways to power it.
If any power source for any reason have their voltage changed, it will jump over to one with highest voltage automagically.

The trick with multiple power sources on MT devices is really trivial ... every power port is equipped with a simple diode which has two functions; 1) it protects from wrong polarity and 2) blocks power leakage through power-in port. After that all power sources are hard-wired together. The common power bus will have voltage equal to the highest voltage connected while diodes will block voltage to leak through other power ports. Then a DC-DC downconverter will drop voltage to levels needed for electronics to run.
If highest voltage power source fails, voltage on power bus will drop to voltage of second higgest and as long as it's within specified range, the only element sensing the drop will be DC-DC downconverter (and voltage sensor if device has any). If power source other than highest one fails, nothing happens.

The fun stuff begins if two power sources provide almost identical voltage. In this case some load-sharing can occur, exact current ratio depends on exact power adapter voltages and wiring resistance (translating to slight voltage drop) ...

So I made some testing and...

Power sources:
I. 2-pin terminal (power brick with output - 24V - 1.5A = 36W, exactly the same as a delivered power supply for DC jack with the router)
II. ZyXEL GS1920-24HP PoE (802.3at compliant [power modes: legacy, 802.3af, pre-802.3at, 802.3at]) switch (50.0–57.0V - 0.6A = 34.2W capped at 30W as per PoE spec)

Scenarios:
A) Both sources connected at startup
The 2-pin terminal takes charge and PoE is never delivered (confirmed with web-management switch portal).
Aa) when PoE cable is disconnected router works as before
Ab) when the 2-pin terminal is disconnected router turns off and after 5-10 seconds it is powering up using PoE. This creates interrupted service, not HA
B) 2-pin terminal connected at startup, PoE connected later
Results are the same as A)
C) PoE connected at startup, 2-pin terminal connected later
Ca) when the 2-pin terminal is disconnected router works as before with PoE delivering power
Cb) when PoE cable is disconnected router works as before with the 2-pin terminal taking over the power supply
Cba) following Cb scenario when the 2-pin terminal is disconnected router turns off and after 5-10 seconds it is powering up using PoE. This creates interrupted service, not HA
Cbb) following Cb scenario when PoE cable is disconnected router works as before

Analysis:
PoE implementation is not good
a) as per https://mikrotik.com/product/rb5009ug_s_in PoE is 802.3af/at but it is not working when DC input is present (802.3af/at requires negotiation with device)
b) the only way for PoE to create an HA solution is a passive PoE solution which in most cases will be idiotic as it requires either a dumb PoE switch or a PoE dumb brick power supply

Potential fix:
1. Put information about such behavior in user documentation so user will not expect HA power solution with PoE (using 802.3af/at standard) and DC (jack/2-pin terminal) in.
2. Fix behavior of the routerboard - after attaching PoE cable board negotiates power with switch. As almost always 802.3af/at switch provides higher voltage this will become the primary source of power for the board and when DC power is disconnectet router can still be up

...
b) the only way for PoE to create an HA solution is a passive PoE solution which in most cases will be idiotic as it requires either a dumb PoE switch or a PoE dumb brick power supply
...

Passive PoE is used in millions of devices serving hundreds of millions of end users around the world

I dont think is idiotic

the productive contribution will be the need to specify the possible need of passive PoE to achieve HA in power

Passive PoE is used in millions of devices serving hundreds of millions of end users around the world

I dont think is idiotic

the productive contribution will be the need to specify the possible need of passive PoE to achieve HA in power

You are absolutely right - but if the specification says you are supporting 802.3af/at but in fact, you are not in the most important use case, then...

CISCO WAPs, for example, you can connect to 802.3af/at and external power and it will behave as you expect (HA mode) not in an idiotic way. I've worked with thousands of PoE network equipment and either the manufacturer states - yes, we can do proper HA with active PoE or they are adding information about not supporting HA use case with active PoE.

And yes - many can support dumb PoE but from ICT equipment manufacturer I expect more.

Also passive PoE is dumb when you have more than 4 devices in your network - total power draw, possible power redundancy etc.

Passive PoE is used in millions of devices serving hundreds of millions of end users around the world

I dont think is idiotic

the productive contribution will be the need to specify the possible need of passive PoE to achieve HA in power

You are absolutely right - but if the specification says you are supporting 802.3af/at but in fact, you are not in the most important use case, then...

CISCO WAPs, for example, you can connect to 802.3af/at and external power and it will behave as you expect (HA mode) not in an idiotic way. I've worked with thousands of PoE network equipment and either the manufacturer states - yes, we can do proper HA with active PoE or they are adding information about not supporting HA use case with active PoE.

And yes - many can support dumb PoE but from ICT equipment manufacturer I expect more.

Also passive PoE is dumb when you have more than 4 devices in your network - total power draw, possible power redundancy etc.

i think the passive PoE dilemma is a matter of context, In corporate market like cisco aruba hpe etc Standard 802.3 PoE is mostly used
But in WISP market passive PoE is the most frequently used
MikroTik equipment is very popular in WISP market, not to say that this is its most important market
In this context the 802.3 PoE compatibility is a novelty and it's more of a curiosity
i think its important for MikroTik to declare if the use of 802.3 PoE source causes the redundancy problems that you discovered

i think the passive PoE dilemma is a matter of context, In corporate market like cisco aruba hpe etc Standard 802.3 PoE is mostly used
But in WISP market passive PoE is the most frequently used
MikroTik equipment is very popular in WISP market, not to say that this is its most important market
In this context the 802.3 PoE compatibility is a novelty and it's more of a curiosity
i think its important for MikroTik to declare if the use of 802.3 PoE source causes the redundancy problems that you discovered

I think you touched few points which nowadays makes sense but in the long run it is pointless. Especially with 802.3bz on the horizon.

BUT - can you point someone at MikroTik to my findings so they can respond to it?

I think the problem here is that RB5009 only performs 802.3 af/at handshake when not running. And without properly completed handshake any 802.3 af/at conforming PSE will not apply power to that port not to damage connected device.

BUT - can you point someone at MikroTik to my findings so they can respond to it?

You can do it yourself ... either through support portal or via e-mail sent to support@mikrotik.com .

Scenarios:
A) Both sources connected at startup
The 2-pin terminal takes charge and PoE is never delivered (confirmed with web-management switch portal).

I think your problem is here...
RB5009 should be able to negotiate and draw power through the POE in regardless that the 2pin terminal is connected...
So something is not working as it should here...
Did you try to POE power the RB5009 using another POE switch capable of af/at ?

Did you try to POE power the RB5009 using another POE switch capable of af/at ?

No, I am running on this PoE switch 6 WAPs (Cisco AP3700 series), 2 IP Cameras, 1 Raspberry Pi.

I was playing with WAPs (as they have an independent power supply) as well as with RaspberryPi. Only MT behaved weirdly.

It's not switch - it is 5009

It's not switch - it is 5009

I am reffering to the device that gives power to the RB5009 through POE...

I am reffering to the device that gives power to the RB5009 through POE...

In this case I don't have another one. I used to have 2 but when I changed jobs I had to send back all their toys

.
Hello,

There is no obligation for PD to be aware of all other power sources and not require PoE handshake negotiation on PoE-In ports if the device is powered by other power sources. In your case, if you want to use PoE-In ports as backup power you should use Passive-PoE with a forced-on feature.

But thank you for you suggestion.

Best regards,
Oļegs Š.

E-mail from support

Now I understand the difference between industry leaders (i.e. Cisco) and discount/niche manufacturers.

FYI