TL-SG3452P and PLANET IGS switches: SM5110-SR is not recognised and the module runs hot in the cage
Small office, two buildings, and all I want is a plaain 1000M fibre link between the racks. Both ends are configured for 1000M, both stay dark, and this has been dragging on for weeks now.
- TP-Link TL-SG3452P, hardware V3, firmware 3.20.4, link taken from one of its SFP cages
- PLANET IGS-6329-8UP2S4 in the second rack, plus an IGS-R4215-24P4X on another run
- SM5110-SR at one end, SM311LM at the other, multimode patch between them
- Ports on the PLANET side hard set to 1000M
The switch never acknowledges the module at all:
end A: SM5110-SR
end B: SM311LM
result: module not recognised, link never comes upOn top of that, after twenty minutes in the cage the SM5110-SR is noticeably warm, warmer than any of the 1G modules we run elsewhere, and that is the prt that actually worries me.
Already tried:
- firmware upgrades on both sides, no change
- moved the modules around all four cages across both switch pairs, same result everywhere
- replaced the patch cord and cleaned the connectors
Is this a module compatibility problem, or is there something about these cages I am missing? And should a module the switch does not even recognise be running that warm?
Also available in: English
العربية
Deutsch
Español
Français
हिन्दी
Bahasa Indonesia
Italiano
日本語
한국어
Nederlands
Polski
Português
Русский
Türkçe
Tiếng Việt
简体中文
Comments 4
There are two separate mistakes stacked on top of each other hre.
The first one you have just found yoourself: SM5110-SR is an SFP+ part, and the TL-SG3452P in hardware V3 has plain 1G SFP cages. It fits mechanically and it gets power, but the switch has no way to drive it, so it stays unrecognised no matter how many firmware versions you go through.
The second is the assumption that forcing the PLANET port to 1000M fixes the other end of the problem. Configuring a cage for a given rate tells you nothing about the module you put into it. Each module has fixed specifications and does not change rate because a port config asked it to, so both ends have to carry modules of the same rate, on cages that support that rate.
The practical fix is a matched pair of SM311LR modules, one at each end, instead of the mixed pair you have now. If the run between the two switches is short, under roughly 3 m, a DAC is cheaper and takes the module question off the table completely.
On the warm module I have no honest explanation. A part that is not being driven should not be working hard, and I have never seen that satisfactorily explained. I would not leave it powered in the cage while you sort the rest out.
One thing needs nailing down before anyoe can help you: what rate are the cages themselves on each switch? The TP-Link and the PLANET boxes are not the same animal there, and the port config screens will not tell you - picking 1000M in the UI says nothing about what the slot is wired for, or about what the optic sitting in it can do. Go and read the port specification for each model, not the speed you selected.
Checked, and that was the right question. The SFP cages on the TP-Link are 1G only - there is no 10G option anywhere in the port config for them. The PLANET side does have 10G cages, which is why I was forcing those down to 1000M by hand in the first place. So I have a 10G part sitting in a 1G-only cage on one side. That would explain why it is never recognised, but I still do not see why it heats up.
Worth adding that getting the cage rate right is not the end of the story with these switches. What I see here is the 10G link coming back at 1G every single time the attached server restarts, while the port config still insists on 10G. Two boxes do it: a TL-SG3428X on 1.30.0 (HW V1) and a TL-SX3008F on 1.20.0 (HW V1.0), real SFP+ cages on both.
I chased the host side for weeks. Intel X520-DA2, then a Mellanox card, then an HP one. Intel E10GSFPSR optics, then 10GTek. A pile of driver builds on Linux and on Windows. Profiles in Omada. Even a chassis swap. None of it moved anything. Bouncing the switch, or dropping the port speed off 10G and setting it back, buys you 10G again until the next host reset.
What stuck was putting TP-Link SM5110-SR modules in place of the third-party ones, and the same result came back from people on an SG3428XMPP and a TL-SX3016F. So the optics bite twice: once for cage rate, once for how the switch renegotiates after a host-side link reset. Try one branded module on a single port before you order a boxful.