←back to thread

1071 points mtlynch | 6 comments | | HN request time: 0.74s | source | bottom
1. JonChesterfield ◴[] No.39401208[source]
Their tinypilot kvm works. It's in the category of boring infrastructure - I leave it plugged into whatever machine is currently in need of attention and `https://tinypilot/` connects to it. Little box hanging off the PoE switch, much less bother than plugging a keyboard and monitor in. Especially so when I'm away from the office.

Recommended. Saved me a lot of hassle. Thanks for the product!

I'd like to be able to power cycle the attached machine but that seems inherently messy to implement.

replies(2): >>39401343 #>>39401735 #
2. tgtweak ◴[] No.39401343[source]
It's a good idea for a feature - many other out-of-band solutions have this. I think the issue comes mostly in that you would need to unplug the server to add (and remove...) some kind of relay-controlled power interrupt.

I would assume that most users of this are in a rack/datacenter/server environment and have switched PDUs they can remotely trigger power with, and it might make more sense to build in PDU management as a software feature over the network than try to reproduce the hardware functions.

Would be kind of cool to have a C14 input plug on the device and a C13 out beside it that goes to the server/device, then you can put an relayed interrupt inside it and also an ammeter clamp to measure power draw and detect drops or spikes in device power - for those systems that don't have built in IPMI already.

replies(2): >>39401628 #>>39402268 #
3. justsomehnguy ◴[] No.39401628[source]
And noone[0] whould allow to use that in a DC because this is now a high-voltage device what can be source of fire, shock and death. You need a certification for this kind of tech and it's not cheap.

[0] of course most of time it would just go under the radar, but you can find yourself in trouble, especially if the device malfunctions

4. mtlynch ◴[] No.39401735[source]
Thanks for the kind words!

Yeah, power cycling is one of our top requests and a feature I'd like to have personally.

If we just plugged into the ATX pins on the motherboard, that would be technically pretty straightforward, but it increases the work and expertise that end-users have to have. I've looked into something like a smart plug, but then users need to configure their BIOS to always power on when power is available.

Definitely on my list of considerations for Voyager 3.

replies(1): >>39402349 #
5. JonChesterfield ◴[] No.39402268[source]
This thing https://www.audiophonics.fr/en/diy-kits-and-boards/audiophon... is a C13/C14 block with a 3.5mm control. AC power on if 12V minimal current applied via the 3.5mm, otherwise off. I haven't got around to it yet but my sketchy DIY idea is power-over-ethernet into a splitter into that. That'll power off the server when the switch port is turned off.

Strangely I haven't found a relay in the same packaging anywhere else.

6. JonChesterfield ◴[] No.39402349[source]
The two ATX pins you short to turn the box on connected to the general purpose IO on the pi is probably the obvious play. The pikvm people suggest running all the ATX pins over another ethernet connection which is probably as neat as it'll get, though I'd inevitably connect the ATX-over-RJ45 thing to the switch PoE at some point and that's probably bad with all 8 wires connected.

Plus you'd want a ATX header to RJ45-or-whatever port per machine. Workable but not very elegant.