S+RJ10 in an RB4011 climbs to 95-99 C and thermally shuts down in an outdoor cabinet
Small tower site: an RB4011 in a ventilated roadside cabinet feeding a couple of radios, with one 10GBASE-T copper hop over to the switch because there was already cable in the duct. The port drops out on warm afternoons and comes back at night, which points fairly clearly at the module rather than at the link.
- MikroTik RB4011, no fan, mounted flat against the cabinet backplate
- MikroTik S+RJ10 in sfp-sfpplus1
- copper hop inside the cabinet plus the existing pair in the duct
- cabinet has passive vents and a filter, no active cooling
/interface ethernet monitor sfp-sfpplus1
sfp-temperature: 97It sits between 95 and 99 C whenever the sun is on that cabinet, then shuts itself down. What surprises me is that the router's own sensor reads 20-30 C below the module, so the box is not cooking the module - the module is cooking itself.
Tried so far: moved the routeer lower in the cabinet, cleaned the filter, propped the door open one afternoon as a test, which helped and is obviously not a solution.
Is 95-99 C simply what these copper modules do, and is there anything short of forced air that fixes it? I also hvae a Ubiquiti UF-RJ45-10G on the shelf which as far as I can tell reports no temperature at all - sensible swap, or just switching off the warning light?
Comments 4
95-99 C is high, but the module is not faulty. A 10GBASE-T PHY in an SFP+ shell burns somewhere around 5 W, several times what an optical module does, and it has nowhere to dump that heat except the cage and the board it is sitting in. The RB4011 is built to run warm without a fan, so the two of them together in a sun-heated cabinet land exactly where you are. MikroTik's own datasheet for the S+RJ10 asks for active cooling, which tells you this is the designed behaviour rather than a defect.
What actually works is forced air over the router's fins, not over the module - the heat leaves through the chassis, so cooling the chassis is what moves the module's number. The build that fixed it properly for one site: an aluminium enclosure with a 70 mm 12 V fan sealed against the router's fins, after which the module settled at about 70 C even on hot days. Cheaper things to try first are mounting the router vertically so the fins convect the way they were meant to, and a variable-speed fan sitting over the heatsink.
The UF-RJ45-10G will not save you. Same class of part, same power to dissipate, it just does not report a temperature - you would keep the heat and lose the early warning and the thermal shutdown that currently protects it.
And the honest answer for a tower site: if you can get fibre or a DAC into that hop, the whole problem goes away. Copper 10G in a passively cooled outdoor box is a permanent maintenance item.
Before the cooling discussion: what rate did that link actually negotiate, and how long is the copper run once you count the patch cords at both ends?
S+RJ10 is picky about cable. Five metres of CAT6 held a stable 10G for me, a 25 m run settled at 5G, and a real 20 m in-wall run plus a couple of patch cords negotiated 10G and then flapped every few seconds. If your afternoon outages are flaps rather than a clean thermal shutdown with a temperature reading behind it, the cable belongs on the suspect list too.
Also check whether yours is the r2 revision - that is the
/r2at the end of the serial numebr.Built the enclosure over a weekend: aluminium box, 70 mm 12 V fan blowing across the fins, sealed so the air goes through them instead of around them. On the hottest afternoon since, the module settled at about 70 C and the port has not dropped once. Call it 25-30 C for the price of a fan and some sheet metal.
One thing still nags at me though. The rated range for the S+RJ10 is 0-70 C and I am now sitting right on the top edge of it. Is that worth worrying about long term, or is this class of module just like that?
That is the class, not a warning sign. These copper modules live near the top of their range in normal service - 70 C on an S+RJ10 is what plenty of people read on DDM, and the same type of module runs at 80 C in Cisco gear and keeps working. You are at the edge of the printed range, but for a 10GBASE-T PHY that edge is the normal working point rather than a cliff.
One thing not to try as a heat fix: forcing a lower speed on the prot. MikroTik's own documentation says the copper modules only work with speed and duplex auto-negotiated and that a statically forced rate will not bring the link up at all. Practice is messier than the rule - some RB4011 and RB5009 owners only got the smaller S-RJ01 stable by forcing 1G full duplex - so if you do experiment, try it both ways and have console access to the box first.