Yeah it’s the drives and the controller for all the drives that are making the power usage what it is. I could replace some of the older drives with a newer one and be able to ditch the smaller drives and controllers, but it seems a waste to do that until they die.

Also, I wouldn’t mind ditching for a Sufficient™ amount of nvme storage, but SSDs aren’t actually getting cheaper and are probably going to do the opposite, so I’ll likely end up doing uh, nothing,

Anecdata, but SSDs will last longer than you want to use them in terms of write endurance.

My NAS OS SSDs are 500gb hynix drives from about 8 years ago, and they’re pushing 150 TBW.

150TB is a LOT of write cycles on a small drive, and they’re still reporting 94% endurance remaining.

The controller will die or I’ll upgrade well before that breaks at the rate it’s going.

Also keep in mind that you can read flash all you want and that doesn’t wear anything (unlike a HDD, amusingly), so for most consumer use cases, they’ll load the drive up with their data, and then only slowly modify or add to it, but have lots and lots of read access.

HDDs will draw around 4W idle each, 8W in total

Whether your drives are idle is also a very use-case specific thing and I wouldn’t spend any time trying to generalize based on that math as a “oh this is how it works for everyone”.

In my case, I’ve got 5 drives all spun up at all times because of torrrent clients, Jellyfin users, and just general media acquisition and public content serving.

This thing would dramatically reduce my power footprint and save me giant buckets of money over it’s lifespan while being smaller/faster IO performance/lower noise.

(My current nas sucks down about 120-140w 24/7, so…)

Yeah, for sure. SCSI died when SAS emerged, and that’s been basically 20 years now.

Any SCSI stuff left laying around is going to be literally a decade+ old and yeah, unless you have a VERY specific need that requires it (which really is just trying to get another few years out of already installed gear), it’s effectively dead and shouldn’t be bought for anything other than paperweights or for a coffee table.

MSA30

Unless my memory fails, that’s billion year old SCSI drives.

Do not buy billion year old SCSI drives, enclosures for SCSI drives, or uh, well, anything like that.

It’s going to use an enormous amount of power, perform slower than a single modern drive, and be prone to failure because well, it’s a billion years old.

That’s not something you want.

For bandwidth intensive stuff I like wholesale internet’s stuff.

The hardware is very uh, old, but the network quality is great since they run an ix. And it’s unmetered too so it’s probably sufficient.

Universiality, basically: almost everyone, everywhere has an email account, or can find one for free. As well as every OS and every device has a giant pile of mail clients for you to chose from.

And I mean, email is a simple tech stack and well understood and reliable: I host an internal mail server for notifications and updates and shit, and it’s rapid, fast, and works perfectly.

It’s only when you suddenly need to email someone OTHER than your local shit that it turns to complete shit.

10940X

“They say”, but they’re right. Ryzen chips do have worse idle power usage, but you’re talking about 10w or so, at most.

And uh, if you were looking at an X-series CPU, I can’t see how that 10w is a dealbreaker, because you were already looking at a shockingly inefficient chip.

Debian stable is great: it’s, well, stable. It’s well supported, has an extremely long support window, and the distro has a pretty stellar track record of not doing anything stupid.

It’s very much in the install-once-and-forget-it category, just gotta do updates.

I run everything in containers for management (but I’m also running something like 90 containers, so a little more complex than your setup) and am firmly of the opinion that, unless you have a compelling reason to NOT run something in a container, just use the containerized version.

I’m the same way. If it’s split license, then it’s a matter of when and not if it’s going to have some MBA come along and enshittify it.

There’s just way, way too much prior experience where that’s what eventually will happen for me to be willing to trust any project that’s doing that, since the split means they’re going to monetize it, and then have all the incentive in the world to shit all over the “free” userbase to try to get them to convert.

Fedora was always a bleeding-edge distro and never all that stable or reliable.

The problem is RedHat/IBM have been fucking with everything, and Fedora has suffered along with everything else and it’s just kinda decayed a bit over the past few years.

…Ubuntu went to shit at least a decade ago, if not longer.

Snapraid parity is offline, so that’s not strictly accurate.

You build the parity, and then until you do a sync/scrub/rebuild they don’t do shit, so there’s no reason to keep them spun up.

If the drive are slow spinning up, then this is probably not a fatal concern, but there’s zero details here.

See, IBM (with OS/2) and Microsoft (with Windows 2.x and 3.x) were cooperating initially.

Right-ish, but I’d say there was actually a simpler problem than the one you laid out.

The immediate and obvious thing that killed OS/2 wasn’t the compatibility layer, it was driven by IBM not having any drivers for any hardware that was not sold by IBM, and Windows having (relatively) broad support for everything anyone was likely to actually have.

Worse, IBM pushed for support for features that IBM hardware support didn’t support to be killed, so you ended up with a Windows that supported your hardware, the features you wanted, and ran on cheaper hardware fighting it out with an OS/2 that did none of that.

IBM essentially decided to, well, be IBM and committed suicide in the market, and didn’t really address a lot of the stupid crap until Warp 3, at which point it didn’t matter and was years too late, and Windows 95 came swooping in shortly thereafter and that was the end of any real competition on the desktop OS scene for quite a while.

That could probably work.

Were it me, I’d build a script that would re-hash and compare all the data to the previous hash as the first step of adding more files, and if the data comes out consistent, I’d copy the files over, hash everything again, save the hash results elsewhere and then repeat as needed.

The format is the tape in the drive, or the disk or whatever.

Tape existed 50 years ago: nothing modern and in production can read those tapes.

The problem is, given a big enough time window, the literal drives to read it will simply no longer exist, and you won’t be able to access even non-rotted media because of that.

As for data integrity, there’s a lot of options: you can make a md5 sum of each file, and then do it again and see if anything is different.

The only caveat here is you have to make sure whatever you’re using to make the checksums gets stored somewhere that’s not JUST on the drive because if the drive DOES corrupt itself, and your only record of the “good” hashes is on the drive, well, you can’t necessarily trust those hashes either.

So, 50 years isn’t a reasonable goal unless you have a pretty big budget for this. Essentially no media is likely to survive that long and be readable unless they’re stored in a vault, under perfect climate controlled conditions. And even if the media is fine, finding an ancient drive to read a format that no longer exists is not a guaranteed proposition.

You frankly should be expecting to have to replace everything every couple of years, and maybe more often if your routine tests of the media show it’s started rotting.

Long term archival storage really isn’t just a dump it to some media and lock it up and never look at ever again.

Alternately, you could just make someone else pay for all of this, and shove all of this to something like Glacier and make the media Amazon’s problem. (Assuming Amazon is around that long and that nothing catches fire.)

I’m using blu-ray disks for the 3rd copy, but I’m not backing up nearly as much data as you are.

The only problem with optical media is that you should only expect it to be readable for a couple of years, best case, at this point and probably not even that as the tier 1 guys all stop making it and you’re left with the dregs.

You almost certainly want some sort of tape option, assuming you want long retention periods and are only likely to add incremental changes to a large dataset.

Edit: I know there’s longer-life archival optical media, but for what that costs, uh, you want tape if at all possible.

Change your port definitions so that they’re only binding to localhost, like so:

• “127.0.0.1:8001:8001”

That’ll stop access from anywhere but the local host. You’ll have to redo your reverse proxy configuration to use 127.0.0.1 instead of whatever you’re using now, though.

The old ‘this is not an airport, you don’t have to announce your departure’ line seems like it might apply here, since that’s the least interesting why-im-leaving post I think I’ve read in a while.

Oh I wasn’t saying to not, I was just saying make sure you’re aware of what recovery entails since a lot of raid controllers don’t just write bytes to the disk and can, if you don’t have spares, make recovery a pain in the ass.

I’m using MD raid for my boot SSDs and yeah, the install was a complete pain in the ass since the debian installer will let you, but it’s very much in the linux sense of ‘let you’: you can do it, but you’re figuring it out on your own.