DigitalDilemma

Nice. It’ll be good for most of the world to not have to rely on an increasingly unstable US export market for cpus.

Never heard of it, but isn’t it the case that if it is shut down, that group of people will just move elsewhere as they have before?

That’s fun, but it’s just a third party summary of your posting history, not what Reddit is using.

so something like RAID 4, 5, 6, or 10 is a great start.

Sorry - whilst most of your advice is great, this is a bit misleading.

• RAID 4 is very rarely used. It’s not a particularly safe or efficient use of striping, and was replaced by 5 shortly after it was invented.

• RAID 5 itself is now strongly discouraged for large arrays. (Google, “don’t use raid 5 for large arrays” for literally millions of pages explaining this, but it basically boils down to; “If a drive fails, the chance of a second drive failing whilst rebuilding is very high”)

But 6 is good if you’ve got enough drives and 10 (1+0) is also a fairly well regarded method for arrays of equal-numbered arrays.

I’m confused that you’re talking of buying 20tb SSDs - you must be very rich. Spinny drives are more usually used in homelab archive RAIDs since they are more cost effective at large size and RAID offsets some of the slowness associated with them. I’m going to assume you meant HDDs not SSDs, but the advice applies to both if I’m wrong about that.

Yes, you will want to RAID them. That gives some protection against individual drive failure, and yes, absolutely that is a concern. Whilst the chance of drives failing these days is less than it was, they still do fail without warning, even when relatively new, and because of the bigger sizes, the consequences are greater.

The alternative to RAID is JBOD (Just a Bunch Of Drives) which means lots of individual drives being presented, each with their true size, in multiple shares. Most folk don’t want that.

What RAID level you choose depends on:

1. How many drives you fit. 4+ is good, and “more smaller” is better than “fewer larger” for safety, although the compromise is an extra 10watts or so of power per drive.
2. Current best practice; Don’t use RAIDs 0 or 5 on large arrays. (0 means exponential increase of data loss. 5 is strongly discouraged due to rebuild times of large disks) 6 is good if you have enough disks. 1+0 (mirrored and striped) is reasonable, and the choice I made for mine.
3. The hardware you’re using. Whether a linux PC or a bespoke NAS tool. Whilst the RAID levels are similar, the tools used vary a lot.

Notes:

• Also, be realistic about the space you need. Don’t over-size. Plan for 3-5 years growth, by then you’ll be wanting to change because of speed changes or drive failure.
• Some raid types slow down writing of data, some speed it up. Most are much faster at READing data.
• https://en.wikipedia.org/wiki/Standard_RAID_levels gives some explanation of the types.
• Google for “RAID CALCULATOR” for lots of free websites that allow you to see what space different sized drives give you with different RAID levels.
• Do not omit a strong backup strategy. RAID only protects against some types of hardware failure. A lightning strike, fire, rogue bios or software update, the host dying with an incompatible raid system. Buy disks for backups that aren’t in your RAID. (Good branded USB 3 disk and caddies are sensible). Automate backups if you can. Backup only what’s not easily replaceable.
• I wrote some thoughts on backups here.

If there’s no errors on the host, that points more strongly at a problem with the drives, their power, or the usb cable, imo.

What you have, @basketugly, is a keenness to learn. Hold onto that, it’ll pay you dividends throughout your life.

I’m guessing here because you didn’t give the exact error message you’re seeing - but two external SSD’s - should be fine for almost anything - PROVIDED they have sufficient power. Check the power needs of the enclosure and drive, and then what your S12/13 is supplying to the USB ports. You might need an additional power supply or powered hub that gives them enough beef.

Or the USB lead itself is pants - that’s definitely a possibility.

I don’t know what prosody is, but my go-to for hosting anything behind a cgnat that needs incoming connections is to use cloudflare tunnel. Free, reliable, lightweight.

You have no local network access to it at all? (WOL will work across the subnet, you don’t need to connect directly to it)

If genuinely not (and I’d be surprised if that is the case) add another nic of any sort (cheap) for WOL, or use something like Switchbot?

You would need a third device monitoring both for this edge case. Once the server has been told to shut down, it’s going to shut down.

The third device (also on the UPS, like an Rpi or ESP) can then check for power availability through the UPS and whatever logic you want to apply, can then use wake on lan to the server to power it up once it shuts down.

You could do it for free. Take the guts out of your old PC, leave the HDD’s in there and the existing PSU. Extend the sata cables through to your MiniPC.

If the PSU won’t fire up, then there’s a couple of pins in the main block you can jumper - or fit a momentary switch to - to act as a switch.

The old PSU will still be reasonable efficient, since power is not wasted except as heat, and it shouldn’t get hot running just the hdd’s. 3.5 hdd’s use around 8-20watt each, depending on spindle speed, so at most it’s 100w at startup, but probably settle down to ~40 for the drives.

Or - yes, those things you linked will work too, but they’re basically doing the same job as the above.

I hope this sack of shit burns to death in his own crappy creation some day.

Whilst I share your sentiment, Elon Musk did not create Tesla Motors.

Martin Eberhard and Marc Tarpenning did. Musk only got involved, and later inserted himself on the board and ultimately took it over, after they sought him out for capital investment. I often wonder what they think about that decision today.

A reasonably unbiased write up

Ok, that’s fair, thank you.

But distributed geographical sites? Useful for SME’s and above, but aside from a few edge cases where friends might want to share hosting resources, is that a homelab thing?

Aimed at self hosting, but S3-compatible and designed to run at different physical locations?

Surely the venn diagram for that has not such a big overlap?

I don’t think I can agree with that, and I’m a pretty agreeable chap.

In the days when people actually cared about the html layout and readability, FP spammed everything hugely, and inserted a lot of terrible cruft. Inventing zillions of new <style> tags for everything, even when the user just wanted to italicise a word. Use a <i> tag? No! We’ll invent a whole new style class and embed it in the headers.

A few years ago I rather stupidly agreed to take over hosting of a website for someone that was dying. It had been written with FP and it took me months to de-cruft it using a lot of regexp and scrifting. (Some 8,000 images and around 2000 .html files).

So you need the self control required to add this extension for those sites you don’t have the self control not to visit too often?

If it prevents us having another crappy week thanks to the like of Crowdstrike, good.

Exactly. Yet another truly awful something is about to happen that’ll get buried under his new patio.

Other have answered the runtime and load question very well already.

I have three other points.

1. Batteries degrade over time. Over-speccing your UPS means more likelyhood that things will hold up in three years time as the capacity given is for new ones. Plus, not running your UPS at 100% capacity reduces its stress. Again, more reliable.

2. You can get a much better quality UPS by buying a second hand one without batteries off ebay and replacing them yourself, typically for a fraction of the cost of buying new. Plus you know you have new batteries. UPS is something where quality genuinely matters. I’ve had to carry a cheap and badly made UPS out of an office whilst it was on fire, so now I spec more carefully. (And ensure they’re metal bodied!)

3. Consider what you NEED to power. What sort of power cuts are you expecting? Does it matter if something goes down?

I UPS my servers and my main desktop, but not my routers, nor my wifi or IOT things. My internet provider also goes out when there’s a cut (I’m on a mesh system so rely on neighbours, who will typically also be down) and I can’t do much without power anyway, but it keeps the disks spinning. We typically get very short automated outages here of less than 10s (yesterday was a bad day, we had 9 within 2 hours)

Because Musk has turned it into somewhere that hate speech is not only tolerated, but encouraged.

Lemmy is literally the antithesis of X, no wonder you’re being downvoted.