Really strange problem here. I have a customer with four remote sites, each using a pair of RB5009UG+S+IN units as their edge routers. They are powered via 802.3at from two independent UniFi US-48-500W PoE switches and everything worked well... until 3:00 this morning.
At that point, both routers at one of the sites powered off for no apparent reason. There had been no configuration changes on the routers or switches. The software version on all devices was consistent with the other sites and had been running for months. The RB5009 units were completely dead, but the PoE phones that were connected to the switches were working properly. We swapped cables, upgraded software on both the UniFi switches and the RB5009 units, but nothing changed the behaviour.
We powered one of the routers on with the DC brick and noticed that the amber light on ether1 was slowly and dimly blinking, but everything else was working. As a shot in the dark, I switched the PoE mode on the second router's switch port to Passive 24V. (This was a bit risky as passive 24V isn't listed as an input method for the RB5009.) The amber light on the second router's ether1 went solid, the router booted correctly, and everything came back. We disconnected the DC brick on the first router, switched its switch port to Passive 24V and went on our merry way.
The other sites are still powering their RB5009 units via 802.3at and are working correctly. I'm going to switch them to Passive 24V just to be sure, but am wondering if anyone else has seen this sort of thing and what you did about it. Thoughts?
If I can make a guess, the RB5009 (and most Mikrotik devices) are "native" 24V.
As a matter of fact in Mikrotikish (the strange language where It seems you pay a fee for each word you type in documentation) 802.3af/at means "802.3af/at AND PASSiVE voltage in the same range as the DC jack accepts".
So, I Imagine that there Is an "added circuit" that manages the 802.3af/at negotiation and - possibly - converts the 48-57 Volts, but if you feed directly and forcibly the 24V It Is bypassed.
This would explain why those are working now, but It doesn't explain how two devices failed in the same way at the same time.
That would definitely explain the symptoms. Something environmental might have knocked out both converters at the same time. Might take a spare RB5009 over there and see if it wakes up when plugged in.
The switch supports both 802.3af/at and 24V passive. I have already configured the switch port to 24V and it's powering the RB5009 without any need for an adapter.
The RB5009 units are supposed to support 802.3af/at natively and I shouldn't need to use 24V passive in the first place.
You're right but that might be a simple and cheap solution to try to fix the issue. I use many of these devices to power 24V passive POE ubiquiti antenna from Mikrotik Powerbox pro 802.3af outdoor switch, and they work like a charm.
I've used these for passive 24V devices on switches that can't do passive 24V natively. If I run into the problem on any of our switches that are only 802.3af/at, I'll definitely be going this route.
I suspect the RB5009 PoE-in is not fully 802.3af/at compliant - no isolated DC/DC converter, negative ground connected to the case. A few other MT devices are like this, for example the CubePro - I have a PoE switch (Netgear GS110TP) powered from -48V DC (positive ground) and when connected to that switch, the CubePro has the negative voltage on the metal case. The real 802.3af/at standard calls for 1500V DC isolation (same as provided by Ethernet transformers), which is a good thing especially with long cables, harsh EMI environments, outdoor devices.
Definitely a possibility. It's just really strange that two RB5009 units powered by two separate switches that had been working on 802.3af/at for months suddenly decided they weren't going to work at the same time. Tomorrow, I'm going to have the customer plug in a new RB5009 on one of the switches' 802.3af/at interfaces to see if it powers up. That should at least narrow my options a bit.
Well, I never got to a definitive answer, but it looks like something environmental occurred that affected both the UniFi switches and the RB5009 units. The ports that the RB5009 units were connected to will no longer provide 802.3af/at power to any device. They have to be set to 24V passive to provide power at all. The RB5009 units will no longer accept 802.3af/at power from any device but will power up using 24V passive. A third RB5009 unit plugged into the same switch on a different port will power up using 802.3af/at without any trouble.