oxidecomputer / omicron

Omicron: Oxide control plane
Mozilla Public License 2.0
252 stars 40 forks source link

Wicket could show basic transceiver status during rack setup #6710

Open bnaecker opened 1 month ago

bnaecker commented 1 month ago

Most customer installs have spent time handling the low-level physical details around transceivers. Common issues include swapped transmit / receive fiber pairs, dirty fiber, or not-quite-seated transceiver modules. Many of these can be diagnosed by looking at the basic transceiver monitoring information, most notably transmit / receive power. This data is available via Dendrite, but only available through the public API after rack setup has completed and Nexus is running. Before that, it's only accessible from the switch zone with swadm or xcvradm, both of which must be run in the switch zone by an Oxide employee.

It would be extremely valuable to show this information during rack setup, in the wicket UI. Wicketd can ask Dendrite (over localhost) for all transceivers that are plugged in, or we could optionally restrict that to the transceivers in the switch ports that the RSS TOML configuration expects to use. Either of these would be really useful when getting the basic networking connectivity off the ground at new customer sites.

sunshowers commented 1 month ago

Chatted with Ben -- it would be good to get this in before the next customer install, but it's not a blocker for r11.