Sun Network 10GbE Switch 72p: x2129a 1G module in an SFP+ port stays down while 10G neighbours are fine
I inherited a Sun Network 10GbE Switch 72p with a couple of legacy boxes hanging off it that will never do more than a gigabit. The plan was to drop a 1G capable module into one of the SFP+ cages and be done with it.
- Sun Network 10GbE Switch 72p, SEFOS CLI over the serial console
- x2129a dual-rate 10G/1G multimode module
- OM3 duplex LC to a 1G port on an older access switch
- 10G links on the same chassis are up and passing traffic normally
The port with that module never comes up. No alarms, no errors in the log, just a port sitting down while its 10G neighbours are fine, and the far end sees nothing either.
Tried so far:
- reseated the module and swapped the patch cord
- moved it into a second cage
- checked the far end port, which is up and waiting for a partner
Is a 1G link supported in these ports at all, and if it is, what am I missing on the switch side?
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 5
Both halves have to be right and neither one happens on its own.
Switch side: only ports 65-72 are dual speed on this platform, so port 71 is fine. The x2129a is the qualified dual-rate 10G/1G multimode part for exactly this job, and x2123a is the RJ45 one if you ever need 1G copper out of those cages. But the port does not fall back by itself, you set the speed manually in SEFOS:
Shut the port before changing the speed and bring it back up afterwards. Skipping that is the usual reason people report the setting did not take. And
write startup-config, otherwise it is gone at the next reload.Far end: disable autonegotiation and force that port to 1000 full duplex. In 1Gbps mode these ports do not autonegotiate, so a partner left on autoneg is the single most common reason the link never forms, which fits your symptoms exactly, 10G neighbours happy and this one silent in both directons.
Do both, then look at the port again. If it still stays down with the far end forced, that is a different fault and worth a proper look at the fibre.
Which port number is it in? That matters here, because not every SFP+ cage on this chassis is dual speed, only a block of eight at the end will do anything other than 10G. If your module is sitting in one of the others you can stop looking for a config mistake, it is never going to work there.
Second question: whta is on the far end and is autonegotiation still enabled on that port? That is the other half of the answer, and "never touched it" almost always means it is on.
Port 71, so it is inside the last block. The far end is a 1G multimode port on an older access switch, autonegotiation enabled, factory default, I never touched it.
On the Sun side I have not forced anything either, the port is at whatever it comes up with by default. If both ends need changing I can take that link down for a few minutes without asking anyone, it carries nothing important yet.
1G in a 10G cage is never plug and play, and the rules differ per platform even inside one vendor. On Extreme VOSS the VSP 4450 wants autonegotiation left enabled when the far end is a third-party switch, while the VSP 7254XSQ does not support autonegotiation at all, so there the remote device has to have it disabled before a 1G link forms. Opposite advice, same vendor, two boxes. For their 1000BASE-T SFP, AA1419043-E6, also listed as 10070H, the recommended practice is custom autoneg on the remote copper port.
MikroTik has its own flavour of this with multi-rate optics. An XS+31LC10D in a plain SFP+ port has to be forced rather than autonegotiated, with the rate-select pin pulled low, otherwise data can get corrupted in transmission:
Different speeds, same class of problem: the cage can carry more than one rate and something has to tell it which one you meant.
Worth appreciating that you have a speed knob at all. On QNAP's managed QSW switches, QSW-M5216-1T, QSW-M2116P-2T2S and the rest of that line, the data mode of the SFP+ ports cannot be adjusted in QSS, only the RJ45 ports can be limited. So a 1G SFP either gets autodetected by the port firmware or it does not link, and there is no configuration to fall back on. People end up hunting for a module the port already likes, or dropping in a DAC instead.
That is also why you see reports where 10G to 10G comes up fine in the same cage while a 1G module stays dark. On your box it is a manual setting, annoying once and then solved for good.