Remote Homelab in an Unheated Building

[lincolna-gaundo]

Upgrading my remote homelab shelf into a racked remote homelab. ITX system, switch, and fiber network has worked fine for two winters. Components have generated enough heat to melt snow on the roof from convection. ITX storage was composed of internal hard drives. They survived -25C but were heated by losses from an internal 750W power supply.

New remote homelab components include KVM, 1500VA / 1000W UPS, USB3 portable storage, 2U ITX case, and a wall mounted enclosed 9U rack. The ITX had been housed in a mid-tower desktop case but moved into a 2U with a smaller 500W PS.

UPS and external USB storage make me paranoid wrt environmental control. Believe that keeping temperatures above freezing should allow these components, especially batteries, to operate without excessive wear, degraded performance, or failure. This paranoia may not be required. Don’t know. Not certtain if the planned solution with environmental controls will meet expectations.

Anticipated cost and effort for environmental controls is a tiny fraction of the overall project. Development effort is limited to remote monitoring software running on ESP32_ETH01.

Curios if anyone has run a remote homelab in an unheated building.

Will keep a build log.

Components being evaluated:

• Bestol Insulated PTC Ceramic Air Heater with PTC Heating Elements (110V 100W 983226MM)
• STC-1000 All-Purpose Digital Temperature Controller, Cooling and Heating Thermostat with NTC Sensor
• CyberPower CP1500PFCRM2U PFC Sinewave UPS System, 1500VA/1000W or,
• APC Smart-UPS 1500VA 980W 120V Output: (6)5-15R (Battery Backup) DLA1500RM2U
• Sipeed NanoKVM Full Mini Remote Control
• PrimeCables 9U Wall Mount Server Rack Cabinet

[lincolna-gaundo]

Got the APC UPS. It has a depth of 18" and the ITX case 16". Cancelled the order for PrimeCables 9U Wall Mount Server Rack Cabinet. A user reports, with a tape measure and picture evidence, that it has an effective depth of 13" - 14". Product listing is misleading.

Survey of available “Server” Cabinets suggest commercial wall mount is a red herring because existing network and computing gear is not compatible.

Insulated cabinet built to spec is being evaluated. This is doable.

[mattlach]

I’m not 100% sure, but if there is some insulation and the room they are in is on the small side, the heat from the systems ought to keep it reasonable in there even during the winter lows. Maybe not ideal for human comfort, but it ought to be warm enough that it does not cause problems for the systems.

I’d honestly be more concerned about the summer highs. 28C plus whatever trapped heat you have from the system running is going to be on the high side of the systems, and will not be great for the UPS batteries. They generally do not like running hot. Hard drives also don’t really care for running at really high temps.

If the outside air is 28C, I can easily see it being about 10C higher inside, and then the hard drives will likely run at about 10C-15C above that. They won’t be happy at all at an operating temperature above 60C. Ideally they should stay below 40C.

Generally when it comes to this stuff, I tend to think of heat as the far greater enemy than cold.

[lincolna-gaundo]

@mattlach

Generally when it comes to this stuff, I tend to think of heat as the far greater enemy than cold.

In general you are likely right. A heater and PID control may be an unnecessary investment.

The concrete floor is a large thermal mass that affects the building overall. Generally acts against temperature changes. Staying close to that and not the roof should be my best course of action.

The problem is a tough nut to crack. First iteration likely to be four vertical rails, four post shelf x2 for structure, and a 1" rigid insulation shell. Baby steps.

[MichellinFlan]

Are the temps you listed outdoor temperatures, or for the room itself?

Could you share some extra details?

• room dimensions
• is the room adjacent to the outdoors? (Aka is there a roof or exterior walls?)
• any windows?
• do you or other people use the room often?
• Any other heat generating equipment not listed?
  -nameplate watts or actual power draw
  -construction of the walls and roof that enclose the room? (Gyp with studs? Concrete? Any insulation?)

I could do a quick back-of-the-hand calc for your temperature rise, but there’s a lot of variety here, depending on the room itself and what’s in it.

[zez77]

@lincolna-gaundo

UPS and external USB storage make me paranoid wrt environmental control. Believe that keeping temperatures above freezing should allow these components, especially batteries, to operate without excessive wear, degraded performance, or failure. This paranoia may not be required. Don’t know. Not certtain if the planned solution with environmental controls will meet expectations.

I’m surprised the lead acid battery ones list 0 as min operating temp. Lead acid doesn’t have a problem with cold afaik unless you go like -15. Same chemistry as cars. Do you have those special winterized batteries where you live? That’s the only reason you would even need to consider damaging batteries, even then unlikely given the waste heat from the UPS.

I believe around the time cell balancing became a common thing, most lithium batteries on phones/laptops/cars started being able to produce more heat internally to optimize temp of the cells based on load (to at least some degree). Not sure if that translated to UPS’s. Would the boards and stuff not work below freezing? what exactly is your concern? It’s highly unlikely to affect the batteries. They might have lower efficiency, but that would be them working to try to stay at an ideal temp or just being slightly less efficient at a lower temp. Only way to hurt them is to freeze them, but I suppose your mitigations wouldn’t help if power was cut out for extended periods of time anyway. That is probably the higher risk you have if you go lithium because the storage temp min is not as low.

Have you looked at how RV’s and those winterized larger vehicles have batteries? I did a quick lookup for winterized UPS’s and they do have some like these for wild markups for the cold certifications (but I suppose so do winterized batteries in general) https://tripplite.eaton.com/extreme-temperature-ups-system-for-industrial-traffic-control-120vac-24vdc-1600w-3u~SMART1524ET

[lincolna-gaundo]

Outside historical records. Worst case over 5 years.

The structure:

• fully detached and free standing;
• 24’ square;
• concrete slab;
• no foundation;
• 2 x 4 16" OC framing anchored to slab;
• clad with 1/2" black joe and about 3/4" concrete;
• hip roof a few feet high … planks not OSB;
• openings include 2 insulated garage doors 8’x9’ with outside gasket,
• 34" pre-hung insulated side door with insulated door stop;
• one window 2’x18" single pain with storm window.

It is a storage building with a parked vehicle and a ring of storage cabinets along the perimeter.

In the winter the doors may be opened once per month and the side door once per week.

The Remote HomeLab is the primary source of heat.

[lincolna-gaundo]

@zez77

Would the boards and stuff not work below freezing?

Ya. Batteries and electrolytic caps, maybe other components too, require care when used outside their vendor recommended operating environment. I’ve seen overclockers use liquid nitrogen on a CPU and the project looked like fun. Not what I want right now.

Won’t challenge what the vendor recommends. Too much liability if I act with undue care.

[ozlay]

I wonder if you can use GEL batteries in a UPS. I haven’t seen anyone do that yet. But GEL batteries work better in cold weather. And you can get them in a 12v 9ah size.

[MichellinFlan]

Calcs are attached! Or they would be, but I’m still a new user. Can’t share links or images

It was a 1D assessment using the construction if you’ve provided. Losses/gains due to ground slab, roof, ext walls, windows, doorways. It doesn’t take into account important considerations like 3D heat transfer, thermal mass, radiation heat transfer, etc. Your biggest loss is the assumed infiltration, which comes in at about 88% of what you’re expected to lose at steady state conditions. Walls and roof could be improved, but the impact of adding more insulation would be minimal in comparison.

I’d agree, an insulated cabinet with a heater is a must-have. Without the supplemental heat or the insulation, your delta T is a measly 0.1 C rise against outdoor conditions. Will 100 W be enough? Depends on the insulation you’ve spec’d for the cabinet.

[lincolna-gaundo]

Thanks!

Found a TECMOJO 12U Open Frame Network Rack, four posts, with three shelves. Should fit the larger components or spill out the back if necessary.

Intend to prototype with 1" rigid pink insulation. There’s a power budget of one tail circuit on a single breaker for the entire structure.

A proper experiment is needed but it will take a while for nature to accommodate. It would be hilarious if the answer is a raspberry pi in a 10" rack with a cell phone battery.

[lincolna-gaundo]

Government of Canada reports:

• Temperature: 6.3°C
• Pressure: 103.6 kPa
• Humidity: 58%

Default calibration BME 280 inside the structure reports (Remote HomeLab to be installed next to the sensor platform):

• Temperature: 11°C
• Pressure: 102.9 kPa
• Humidity: 48% rH

TECMOJO 12U Open Frame Network Rack to be delivered today. Remote HomeLab will initially leach network support from main rack until moved into the remote building but otherwise as initially imagined.

[lincolna-gaundo]

Used Brave AI to answer questions wrt this project. Got to say it was amazingly helpful and detailed.

how to keep a computer case above freezing

• ensure it is located in a room with a stable ambient temperature

must use unheated building

• ensure it remains powered on continuously to generate internal heat
• use a thermostatically controlled heating source
• controller can effectively regulate temperature

need to shut down periodically

• insulate the case or place it in a sealed, ventilated cabinet

is there a pid that will use pwm fan control and a heat source to adjust temperature

• AI provides DIY description of a PID controller and DIY instructions. highlights follow.
• test the system with a step change in setpoint and adjust PID gains to minimize oscillation and settling time.
• tools and materials
• step-by-step instructions
• program logic

provide reference for commercial product

• The Adaptive Junior V2 (ADJ-48-450-UR-V2)
• UK$247.24

Am ready to assemble Remote HomeLab. Second 120mm fan to be delivered today. ETH_01 should be here next week. Hopefully the assembly goes well.

• TECMOJO 12U Open Frame Network Rack
• Iwillink 24 Port Blank Keystone Patch Panel
• VEVOR 1U PDU, 10 Outlets, Rack Mount Power Strip
• CyberPower CP1500PFCRM2U PFC Sinewave UPS System
• Sipeed NanoKVM Full Mini Remote
• STC-1000 All-Purpose Digital Temperature Controller, Cooling and Heating Thermostat with NTC Sensor
• Bestol Insulated PTC Ceramic Air Heater with PTC Heating Elements (110V 100W 983226MM)
• 2x Wathai 120mm x 25mm 110V 220V AC Powered Fan with Speed Controller 3V to 12V

[zez77]

@lincolna-gaundo

Ya. Batteries and electrolytic caps, maybe other components too

I would put a thermocouple on one place inside your rack air in and on some components to see if this is an issue (not sure what would be highest risk here). I think you underestimate how much heat is produced by these components. In an insulated cabinet as long as you have low air flow you should be fine just from heat from your rack. I assumed at least 100W at all times given you have a rack, system, and switch. I didn’t realize until now you have only a 100W heater spec’d. What is the idle power draw of your entire rack? 100W in a small enclosed rack even with poor insulation is gonna heat up quite a bit if you turn the fans off.

I still see your big problem being what happens if power goes off for a prolonged period of time. Do you have a generator/access to the property in the case of an outage? I’ve known several people have had in house battery backups just not work after being in an unfinished basement for 3 days after a big snow storm. I believe it was ~-15 to -20F for 3 days. I assume that’s why they’re not rated for low storage temps. Only way to really protect from that is a generator which would actually be doable if it didn’t violate codes to run one indoors.

After I saw that winterized UPS for cell towers, I saw they have an entire ecosystem of stuff for industrial PCs and equipment for places like you have that can experience days long power outages. You might wanna look to see what they do to operate at low temps if that’s a concern.

[lincolna-gaundo]

@zez77

I still see your big problem being what happens if power goes off for a prolonged period of time.

It will shut off automatically. Existing use-case only calls for it to be available once-per-week as an rsync server for backups. This will change but waiting for new project to launch. Probably March. Would something like the MINISFORUM MS-S1 MAX fit on the shelf?

Test fit of components. Things always take more space than expected.

The temperature controller has a great design but execution is poor. Will fix the relay traces and put in steel box, fused, and grounded to earth. Feel like BigClive.Com.

[lincolna-gaundo]

Added a heat chamber to bottom of rack.

Cooling has manually adjusted intake and exhaust fans. Can choose positive, balanced, or negative pressure and airflow. Need to remove excess heat, say above 20C.

Heating is a 12V 100W heater with fan. This option requires a power supply. Using 12V 30A likely drawing 10A based on spec. Will see how these perform when tested. My assumption is that it comes nowhere near 100W however the power supply itself is also a heat source. Intend to keep enclosure above 5C.

Second option is 110VAC 100W heater using convection only. No fan.

Will be interesting to discover if heat circuit is ever powered by the thermostat.

[lincolna-gaundo]

Built remote homelab work area. Pegboard shelves will hold tools and supplies. TecMojo Rack will sit under the desk when prototype is ready to be deployed.

Working on control system for case temperature monitoring and AC power.

Prefer one of ESP32-S3, ESP32-C6-Zero, ESP32-C3 but these require using an untried W5500 Ethernet module, or WT32-S1 with built-in Ethernet. Normally just prototype on ESP32 using Micro Python but WT32-S1 is not cooperating. Need Tasmota or Arduino development environment. Have used in the past but need to install and re-learn.

Current prototype, as pictured, using RPi5 with a pair of BME280 sensors and USB TTY, or Ethernet TTY bridge. Would like to run entire control system on Ethernet TTY bridge (WT32-S1). An optimization for later.

[lincolna-gaundo]

AI Query

8 feet3 volume surface area 24 ft2 R6 insulation how many watts of heat required to raise temp 1C

NOTE: This does not include intake and exhaust openings which may require louvers. Don’t know.

To determine the number of watts required to raise the temperature by 1°C in an insulated enclosure with a volume of 8 ft³, a surface area of 24 ft², and R6 insulation, the primary factor is the rate of heat loss through the insulation, which must be overcome to maintain or raise the temperature.

The heat loss through the insulation can be calculated using the formula for heat transfer:

q(dot) = A⋅ΔT / R

where,

• $\dot{q}$ is the heat loss in watts (W),
• $A$ is the surface area in square feet (ft²),
• $\Delta T$ is the temperature difference in degrees Fahrenheit (°F),
• $R$ is the R-value of the insulation in °F·ft²·hr/BTU.

Since the temperature rise required is 1°C, this is equivalent to 1.8°F. The surface area is 24 ft², and the R-value is 6. Substituting these values:

q(dot) = 24⋅1.8 / 6 = 7.2 BTU/hr

To convert BTU/hr to watts, use the conversion factor 1 watt = 3.412 BTU/hr:

7.2 BTU/hr÷3.412≈2.11 watts

Therefore, approximately 2.11 watts of heating power are required to counteract the heat loss and raise the temperature by 1°C in the enclosure, assuming steady-state conditions and no additional heat sources or sinks.

NOTE

Same question but maintain temperature requires 21.1W.

Outside lowest low, past 5 years, is -26C and target low of enclosure is 5C. Want enclosure above freezing. 31C delta * 2.11W = 65.41W.

This calculation assumes the temperature difference is maintained across the insulation and does not account for transient effects or internal heat generation.

Dry Fit

Inner Hull Complete

[electric_anteater]

Cool project. I’ve been thinking about setting up a backup server in my shed. Interesting to see your approach to managing temperatures.

@lincolna-gaundo

is there a pid that will use pwm fan control and a heat source to adjust temperature

Lm-sensors plus mprime?

[lincolna-gaundo]

Should I consider this a recommendation for the next version?

Am using an STC1000 Temperature Controller. For value you can’t beat an STC1000 and is recommended and well documented by www.nationalhomebrewclub.ie.