Ancestors

Toot

Written by Paco Hope #resist on 2025-01-21 at 16:31

I have a server in my #homelab that I #selfhost in my literal garage. It has a funny problem. When it gets too cold (and right now it is 11ºF, -11ºC in Virginia), it crashes. 🤦‍♂️

I bought a replacement a few weeks ago and I moved the last VM off of it last night. This morning I woke up and found that it was crashed....again. So I hit the power button turning it off, and said in my best Darth Vader voice:

"you have failed me for the last time."

[#]selfhosting

=> View attached media

=> More informations about this toot | More toots from paco@infosec.exchange

Descendants

Written by Graham Sutherland / Polynomial on 2025-01-21 at 16:33

@paco I can only imagine it's a condensation issue at that temperature. -11°C isn't very low for any of the components in that machine, especially when you consider the self-heating will be bringing it higher. so you might have to watch out for that on the new machine too.

=> More informations about this toot | More toots from gsuberland@chaos.social

Written by Graham Sutherland / Polynomial on 2025-01-21 at 16:34

@paco if you have temperature and humidity monitoring in there you can calculate the dew point, which should give you more of an idea whether condensation is a likely factor.

=> More informations about this toot | More toots from gsuberland@chaos.social

Written by Paco Hope #resist on 2025-01-21 at 16:34

@gsuberland Ambient air temp is about 66ºF / 19ºC.

Good call on checking humidity/dew point.

=> More informations about this toot | More toots from paco@infosec.exchange

Written by John Regehr on 2025-01-21 at 16:44

@paco @gsuberland but also, sometimes there's a contact that will break when temperature changes expand/shrink some part past a threshold. I've only seen this firsthand in the other direction though.

=> More informations about this toot | More toots from regehr@mastodon.social

Written by Graham Sutherland / Polynomial on 2025-01-21 at 16:46

@regehr @paco the contact pressure should be more than enough to handle that unless it's a very sudden temperature change. but even then I've seen folks splash LN2 on running boards during extreme overclocking benchmarks and they continue working just fine.

=> More informations about this toot | More toots from gsuberland@chaos.social

Written by John Regehr on 2025-01-21 at 16:50

@gsuberland @paco it was a PCI board where I saw this happen. could have been a bit bent, have a damaged contact,. or whatever

=> More informations about this toot | More toots from regehr@mastodon.social

Written by yopp on 2025-01-21 at 18:52

@paco @gsuberland could it also be caused by drop in mains voltage?

=> More informations about this toot | More toots from alex@yopp.me

Written by Qlaras on 2025-01-21 at 17:36

@gsuberland @paco

Hmmm I may be doing something similar for an ADS-B box in the future. (Uninsulated garage attic, steel siding so have to get around that.)

Would a 5-sided insulating shell (open on the bottom) maybe work?

Would let humidity still escape for any humidity/condensation buildup but keep 'some' heat trapped.

=> More informations about this toot | More toots from qlaras@mastodon.social

Written by Graham Sutherland / Polynomial on 2025-01-21 at 17:44

@qlaras @paco are we talking about an SBC or a server here?

=> More informations about this toot | More toots from gsuberland@chaos.social

Written by Qlaras on 2025-01-21 at 17:53

@gsuberland @paco

I would be doing a Lenovo ThinkCentre Tiny (1 liter volume) or other SFF x86_64 box.

The lab rack mounts get to live in the basement, but haven't put those to use yet - waiting on additional power circuits installation in a few weeks and fiber coming this spring.

=> More informations about this toot | More toots from qlaras@mastodon.social

Written by Graham Sutherland / Polynomial on 2025-01-21 at 19:12

@qlaras @paco hmm, tricky. in a space like that with fairly low heat rejection I'm betting it'll have a very wide temperature range, along the lines of -10°C to 50°C?

for the cold it's easy: stick it in a box, the system's own thermal output will heat it up beyond the dew point. dessicant bag in the box in case the power goes out or something, just to tide it over.

for the heat you're going to be limited to the ambient temperature so it's probably mostly a case of ventilating the space itself.

=> More informations about this toot | More toots from gsuberland@chaos.social

Written by Qlaras on 2025-01-21 at 19:17

@gsuberland @paco

Exactly. Here in MN -10C is putting it mildly... Went below -25C yesterday, -40 is not uncommon.

Probably keep it just below the trusses so it doesn't get too much heat in the summer, and I can make covering/uncovering it for winter part of the garage roof vent plugging in the fall and unplugging in the spring process. Then the antenna can stick above the trusses (and thus the steel siding) although I'll have to stay clear of the end.

Or poke/follow a hole to the exterior.

=> More informations about this toot | More toots from qlaras@mastodon.social

Written by Graham Sutherland / Polynomial on 2025-01-21 at 19:26

@qlaras @paco one thing you can do if the heat from the server itself isn't sufficient to keep the enclosure above the dew point is to get one of those little outdoor tube heaters with the inbuilt thermostat and put it inside the enclosure. for a small volume it probably only needs 30W or so. the fans in the server will help circulate the air inside. this is how most outdoor comms cabinets prevent condensation inside them.

=> More informations about this toot | More toots from gsuberland@chaos.social