Making a Linux home server sleep on idle and wake on demand (2023)

(dgross.ca)

294 points AgaoAnar | 2 comments | 02 Sep 25 19:42 UTC | HN request time: 0.598s | source

Show context

privatelypublic ◴[02 Sep 25 21:01 UTC] No.45108992[source]▶

Well done. But note: you can grab an ATX control board, or configure the RPi as a USB Gadget to wake the machine via power button or keyboard.

replies(3): >>45110663 #>>45111253 #>>45112378 #

bjackman ◴[02 Sep 25 23:56 UTC] No.45110663[source]▶

>>45108992 #

I did this recently as I was struggling to get WoL to work with my consumer PC. It seems like this ultra low-level stuff is a total crapshoot so if you can dodge it by just wiring up the power button, that's a good option.

In in the end I just went the whole hog and set up a PiKVM, so now if I mess up the machine's networking (or even completely break the OS) I can still recover it remotely even though it doesn't have a proper BMC or anything like that.

In general this approach seems ugly in principle but I really like it in practice. It lets you retrofit solid remote capabilities onto consumer hardware. That way you have such a broader market to buy from.

replies(1): >>45117538 #

1. privatelypublic ◴[03 Sep 25 16:17 UTC] No.45117538[source]▶

>>45110663 #

I'm absolutely excited for the nanoKVM-PCIe. They were out of stock last I looked, but they've released the firmware source as promised.

replies(1): >>45126381 #

2. bjackman ◴[04 Sep 25 12:15 UTC] No.45126381[source]▶

>>45117538 (TP) #

I hadn't seen that before, seems neat as the physical inconvenience factor of the KVM approach is definitely real.

It's pretty funny that you have both the input and output physically installed in and powered by the chassis, but then you run cables to connect them!

It would seem reasonable to at least have the option to have it directly enumerate as a USB hub/display device on the PCIe bus it gets power from! But maybe that would add a lot of bulk/cost?

↑