ExtremeSwitching X690 refuses 100G on port 57 and shows the QSFP28 DAC as unsupported
Adding a 100G link between two racks on an X690 that has been quietly doing 10G access work for a couple of years. The uplink cage will not take the speed at all.
- ExtremeSwitching X690-48x-2q-4c
- 1 m QSFP28 direct attach cable, Q28+CR4_1m
- port 57, one of the QSFP28 uplink cages
The cable comes up as unsupported in the port listing, and setting the speed by hand is rejected outright:
* X690.1 # configure port 57 auto off speed 100000 duplex full
Speed specified is not supported on port 57What I tried:
- reseated the cable at both ends and swapped it for a second one of the same part;
- moved it to another QSFP28 cage, same message with that port number in it;
- left auto on and simply enabled the port, which leaves it down with no link.
The cages are supposed to be 100G capable, and the switch is refusing the speed rather tahn failing to link, so I suspect the port is not configured the way I think it is. What decides the speed of these uplinks, and does changing it need a rebooot?
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 3
The uplink cages on the X690 are not independent ports with their own speed setting. Ports 49, 53, 57, 61, 65 and 69 share a single 400 Gbit/s budget, and how that budget is cut up comes from a partition template, not from a per-port speed command. That is why the CLI refuses the speed instead of accepting it and leaving the port dark: at that moment port 57 is not a 100G interface at all. The default templatte gives you two 100G ports plus four 40G ports, and 57 is not one of the two.
Pick a template that provides the number of 100G interfaces you need, then assign the port itself:
Then reboot the switch. The new layout only takes effect after the restart, and both commands are accepted without complaint before it, which is what makes this confusing the first time.
Besides 1x100G the per-port options are 1x40G, 2x50G, 4x10G and 4x25G, so work out the whole uplink plan first and change the template once rather than per link. Same applies to the 48t model, the uplink block there is identical.
Confirmed. Set the template to 4x100G, assigned
configure ports 57 partition 1x100G, rebooted, and the port came back as a real 100G interface with the Q28+CR4_1m linking straight away, no more unsupported cable.The reboot is the part I would have missed. Before it the commands sit in the configuration looking applied while the port still rejects the speed, which is indistinguishable from the original problem. Anyone doing this on a production box should book the outage window up front rather than discovering it halfway through.
Same family, opposite trap, for anyone who lands here with an X695. There the QSFP28 ports default to 100G and only two of the eight, 51 and 55, can be partitioned at all. Try it anywhere else and you get:
On the partitionable pair you use
configure port 51 partition 1x40G. On the rest you force the speed instead, with the very command that fails on the X690:I did that for ports 49, 50, 55, 60, 61 and 62 on EXOS 30.6.1.11-patch1-4 to get 40G optics up. Ports 52-54 and 57-59 show Link State NP in the default layout, which is the layout talking and not a fault.
So: X690 uplinks are template driven, X695 QSFP28 ports are mostly speed driven. Check which box is in front of you before copying commands between them.