Zyxel XS1930-10 will not link a 5 m 10Gtek DAC that still works on an XGS1250-12
Small home rack, one server and one switch. I replaced an XGS1250-12 with an XS1930-10 and the only thing that did not survive the swap is the server uplink.
- Zyxel XS1930-10, SFP+ port with media type set to DAC10G
- Asus XG-C100F in the server
- 10Gtek CAB-10GSFP-P5M, 5 m passive twinax, AWG26
- a second 10Gtek DAC, 1.5 m, kept as a spare
The switch port stays down while the NIC is convinced it has a link:
XS1930-10, SFP+ port: media type DAC10G link status Down
server, XG-C100F:
Speed: 10000Mb/s
Duplex: Full
Link detected: yesWhat I have tried:
- the 1.5 m 10Gtek DAC in the same port, same NIC - links immediately
- media type set explicitly to DAC10G instead of left on auto - no change
- the 5 m cable back on the XGS1250-12 - still works there, so the cable is not dead
Is 5 m of passive AWG26 simply more than this switch will carry, or is there a port setting on the XS1930-10 that I am still missing?
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
Expect that loopback to come up and the link to the NIC to stay down anyway - that is the usual shape of this one.
The media type is a prerequisite, not a fix. Left on auto an SFP+ port is prone to not linking with a DAC at all, so DAC10G has to be set. Once it is set and the port is still down you are out of settings and into the physical budget of the port.
Five metres of passive AWG26 is a lot of copper loss for this class of switch. Zyxel's closing position on the XS1930-10 is that a passive run that long loses too much signal for its SerDes and that anything past roughly 3 m belongs on fibre; the guarantee never stretched past their own branded modules and DAC cables anyway. That also covers your other observation - the same cable working on the XGS1250-12 proves nothing about the XS1930-10, because a length of twinax that behaves in one model's cage tells you very little about the next model, even inside one vendor's own range.
Practical route: keep the 1.5 m cable for that link, and if you actually need the 5 m distance put SR optics and a duplex LC patch in instead of twinax. If you want it on record, set the media type, confirm it still fails and open a support ticket - on the XGS1930-52HP that is exactly the route their own FAQ gives for a DAC still down after the media type is set, and it goes through the Nebula Control Center, so I would expect the same handling here. I would not sit waiting for a firmware release to change the loss on a cable.
Before you decide anything about the cable, loop it: both ends of the 5 m DAC into the two SFP+ ports of the XS1930-10 itself, media type DAC10G on both. If it comes up in loopback then the cable and the cages are fine and what you have is a pair problem, not a cable problem.
Two things you have not sadi. Is the port forced to 10G full duplex, or left on auto after you set the media type? And what was the XGS1250-12 port on when the same cable worked there - auto or DAC? If it was on auto, switch that one to DAC before you compare the two switches: the media type is a per-port setting on both boxes, so a test where one of them is left on auto is not really the same test.
Different vendor, same category of surprise. I had an SFP+ DAC that would not link between a TL-SG2210MP, where the cage is a plain 1G SFP, and a TL-SX3008F on firmware 1.2. Forcing both ports to 1000 Mbps full duplex changed nothing, and the same cable was fine between two 10G-capable boxes.
The explanation there was coding rather than length: plenty of passive SFP+ assemblies advertise 10G and nothing else, with no 1000BASE-X fallback in them at all, so they sit dark in a 1G cage while a dual-rate cable links. Someone in the same discussion had a Unifi SFP+ DAC running between an ER8411 and an SG2428P with no drama at all.
Not your case, since both your ends are 10G. The point is that 'it works over there' carries much less weight with twinax than it does with fibre.
Worth knowing how the host learns about a twinax cable at all: a passive DAC carries its length and its attenuatin in EEPROM, per SFF-8472, and the port drives the cage from what the assembly claims. Nothing is measured on the wire. That is a large part of why one 5 m cable behaves differently on two switches of the same brand - different receivers, different tolerance for what that claim actually costs in loss.
Also do not count on twinax when you need a link slower than its rate. On a Netgate 6100 a DAC gives you a 10G link and nothing slower, with no way to drag it down to 1 Gbps; the interface simply comes up as 10Gbase-Twinax full duplex, so if whatever sits on the other end is pinned to a gigabit, twinax will not get you there and it is back to 1G optics. Short cables in that brand are unremarkable in general - a CAB-10GSFP-P1M and a Ubiquiti UACC-DAC-SFP10 both run at 10 GbE into a USW-Enterprise-8-PoE here. Length is the variable to test first, not the brand.