D-Link DGS-1250-28X and DGS-1250-52X: BiDi SFP OSNOVO detected, but link on one fiber won't come up
Servicing a small office network, two buildings linked by a single single-mode fiber that contractors pulled. The job was supposed to take half an hour: bring up a gigabit link between the switches in the two buildings. It's been two weeks and it still won't come up.
What's installed:
- main building: D-Link DGS-1250-28X
- second building: D-Link DGS-1250-52X
- both ends: SFP OSNOVO S1LC13-G-1550-1310, single fiber, SC
- single-mode inter-building run, splices were not done by us
What I see:
DGS-1250-28X, SFP port: Link Down, LED is off
DGS-1250-52X, SFP port: Link Down, LED is off
both modules show up in the web UI, no transceiver error entries in the logWhat I've already tried:
- checked continuity with a flashlight, light is clearly visible from the far end;
- put in a second pair of identical modules, exact same behavior;
- moved the modules to neighboring SFP ports and rebooted both switches.
Where should I dig next: the fiber, the switches, or the modules themselves?
Show original (Русский)
Also available in: Русский
العربية
Deutsch
English
Español
Français
हिन्दी
Bahasa Indonesia
Italiano
日本語
한국어
Nederlands
Polski
Português
Türkçe
Tiếng Việt
简体中文
Comments 6
This is exactly the symptom you get with two identical BiDi modules. A single-fiber module transmits on one wavelength and listens on the other, so the two ends have to be mirror images: one TX 1310 / RX 1550, the other TX 1550 / RX 1310. The marking usually splits the halves with a letter U and D (1000BASE-BX-U and 1000BASE-BX-D, for SFP+ it's 10G BXU and 10G BXD), and physically they're distinguished by housing or latch color.
If what you're holding is two modules of the same color with the identical label S1LC13-G-1550-1310, then both are transmitting in the same direction and both are listening for something nobody is emitting. The transmitter shines straight into the transmitter, the receiver sees nothing, and the link will never come up in that arrangement no matter how many pairs you swap. Judging by your description, the second pair was the same, hence the same result.
Get a kit that's specifically a matched pair, U + D, and make sure both halves are from the same batch: different brands can differ in center wavelength and receiver bandwidth, and two modules that are "correct on paper" can still fail to match. After swapping, reseat the module or do a shutdown / no shutdown on the port so the switch re-reads the EEPROM and doesn't hang on to old state.
Are you sure the modules are actually a matched pair? Put both units side by side and look at the housings themselves, not the box: frame and latch color, what's printed on each label. Does the text match word for word or not.
And two more questions while we're at it. Is the fiber definitely single-mode 9/125 and not multimode, which lights up just as well with a flashlight? And did you read the log only through the web UI, or through the console too? The web interface shows a trimmed-down summary, and some transceiver messages never make it there at all. Hook up to the console port, pull and reseat the module, and show what gets dumped to the terminal.
Let me add something on diagnostics, because this is where a lot of people go the wrong way. Two identical BiDi modules are detected by the host just fine, DDM shows sane numbers for the transmitter - and it creates a full impression that the modules are alive and the fiber must be to blame. What you should be looking at isn't whether it's detected, but the receive level: if the receiver sees nothing at all while there is light on the line, the question isn't about the run.
The manufacturer prints the wavelength assignment on the label. You need to read it on both ends, not just the one that's easier to get to.
Had a similar story, just with different hardware: also turned out to be a single strand between buildings, HP 1910-8G-PoE (JG350A) on one side, D-Link DGS-1210-10 on the other. All the HP modules on hand were dual-fiber, so we had to pick a WDM pair from scratch.
The main gotcha there wasn't the optics but the host. HP is noticeably pickier about a module's coding matching the exact model, D-Link accepts almost anything, so it makes sense to compromise on the D-Link side and pick something for the HP side that's known to work in it. Specifics that came up were DEM-330 for the D-Link side and SNR modules as a way to make the two vendors get along, but we never settled on a final kit at the time - we were picking based on the spec sheet.
Another reason a single-fiber link might not come up is a hard-set speed with autonegotiation disabled. Ran into this on a MOXA and ELTEX pairing: the optics were completely fine, but a port fixed at 1000/full simply wouldn't negotiate with the other end, and from the outside it looked exactly like a dead fiber.
Before ordering new modules, I'd set both ports back to auto - the check costs nothing. Though if your modules really are identical, this won't change anything.
The autoneg point above is off the mark. With the wrong wavelength assignment the receiver sees no light at all, and no port mode is going to fix that; that's exactly what the symptom shows, both ends are deaf at the same time.
What's more useful is to keep a normal troubleshooting routine in mind in case the pair turns out to be correct and the link still won't come up: confirm the fiber really is single-mode and matches the modules, check the switch console for transceiver warnings, swap one component at a time against a known-good one - patch cord, module, port - and finally measure attenuation with an OTDR. A flashlight only confirms the strand is continuous; it says nothing about loss at splices or dirt in the connectors.