Shut down True Nas Scale System

[alva]

Hey there, I know similar questions has been asked, but for ecological and economical reasons I would like to hear your experience regarding this scenario.

My hardware is as follows:

X570 Gigabyte board
3900X AM 4 CPU
32GB DDR-Ram (no ECC but this will change)
2070 Super GPU

I am living between two countries and I sometimes need the power of my GPU for machine learning tasks as well as have to access to files from my Archive which is 4TB large + backups of my Macbook. Cloud is a good alternative but not all the time.

Sometimes means: I have to work on a project for intensive three weeks and in this time I need the GPU power as well as data. After this, the machine will run idle for 2 or 3 months. I would like to shut it down during this time.

I am aware that TrueNAS purpose is to serve files 24/7 in a local network but I don’t see the point of causing more carbon emissions than necessary.

Do you have any experience with similar use cases of utilizing your hardware and did the drives fail at one point from spinning up and down in a similar use case?
New drives also cause emissions...
From a hardware perspective I just have the experience of owning a 6 year old WD red, just to store data and it still seems fine.

Running ZFS for the Archive is to avoid bitrot - but I am not sure if this means running 24/7 is required?
I totally get the concept, but what can rot if it is not running? Scrubs and backups will be part of the process.

Thank you in advance

 

[Ericloewe]

I totally get the concept, but what can rot if it is not running?

By definition of bit rot, as far as anyone can agree on one, these would be processes that affect data at rest. Things like memory cells in an SSD losing their charge, HDD sectors changing their polarization, or perhaps even more esoteric stuff like a disk controller (SSD or HDD alike) "growing" a subtle failure that corrupts data when it does get read later.

Bit rot and running 24/7 aren't very related, beyond the tangential: If you're running 24/7, it's more likely that bit rot or other failures can be corrected earlier, but you may also be inducing additional wear on your system if it's poorly-designed (e.g. if it overheats).
General wisdom is that frequent power cycles are bad, for a number of very valid engineering reasons. The devil is in the details, of course, but for most people 24/7 is easier to manage because it's a known quantity, provides maximum uptime, allows for stable maintenance schedules, and generally matches the design intent of server hardware.

The absolute worst-case might be a power controller on a server motherboard that has a bug that causes a regulator to wear out when the system is booted (IBM/Lenovo X3550 m5 says hi), making power cycles much more damaging than months of continuous operation.

 

[Arwen]

If you do end up shutting the NAS server down for a bit, make absolutely sure that when it's up, it completes both SMART short tests, (which should be quite quick) and ZFS scrub(s). Those ZFS scrubs will detect any bit-rot and attempt to correct. I say attempt, because you don't specify your pool layout. Reduced redundancy, (like using many large disks with RAID-Z1), can affect ZFS' correction capability.

 

[alva]

Thanks you two, this was helpful input.
My pool is nothing fancy, a 1x12 TB mirror dev with ~60% empty space.
A NVME for System and VMs, but thats not so important.
I will keep your input in mind - maybe I will have time to tweak the idle power consumption of GPU and CPU, but when I got this stuff I was not aware of utilizing a computer that way. (VMs, remote access, ML). Now I am down the rabbit hole and even use VMs on my Mac because it keeps evrything so tidy!