Post #1974043
2026-05-04 19:57 UTC
Skipping on documenting deployments is the worst type of technical debt
I spent half a day just documenting what hardware my new lab has and how it set up. It would be an hour or so, if vendors published a bit more than zero shit on their websites.
Like this motherboard has ITE IT8613E Super I/O controller, that is not supported by mainline kernel and has no datasheet available at all. I don’t know why PWM and thermal control is so cursed on motherboards.
Kudos to a person who maintains (and tries to upstream) IT87 driver, at least I can manually control fans. This is second time I had to write “middleware” to adjust PWM. On previous lab server fan controller was exposed only to BMC own OS, so I had to adjust them using terrible Gigabyte implementation of Redfish API.
Replies (1)
-
@alex@feed.yopp.me 2026-05-04 20:04
All started with “hm, why I haven’t seen any reboot alerts after blackout”. Reason was simple: host booted into initramfs and was silently waiting for zfs unlock key. And you can’t automate zfs unlock, because debians zfsunlockcan’t get key from STDIN. And of course I did not document anything and completely forgot about that. Looks like last time I tried to set automatic ssh unlock I got tired and here we are.