On of my HDDs in my server has a power-on time of 5.7 years already.

[ithoughtisawa2@infosec.exchange]

@jakesmolka @homelab just because it's old doesn't mean it needs to be replaced. It is showing any issues or SMART warnings? Your data protection strategy should be able to survive a single drive failure. When/if it fails you can replace it then

[jakesmolka@hachyderm.io]

@homelab I'll also try to implement the UPS strategy I got up my sleeve ASAP. Can't risk no head crashes at this rate.

[jakesmolka@hachyderm.io]

@ithoughtisawa2 That's a good hint. Thanks.
No problems so far, but I wasn't aware it was running so long already. SMART looks okay. Maybe this HDD will still last some time. But at those prices and how they just get higher and higher, it'll probably be wise to buy something and anticipate the failure.

Re data protection: my server runs Unraid and this HDD is the parity drive. So no problem. Same, if it would be one of the data drives.
With this setup I could even use the new drive already as second parity drive and just increase the level of fault tolerance.

[jan@fedi.kcore.org]

@jakesmolka

I have one that now has 8 years of power-on time, and is showing no issues. It's part of a ZFS RAID10, so I can survive if it goes down.

I would only replace when needed - but you do need to have redundancy and backups in place.

@ithoughtisawa2 @homelab

[woe2you@beige.party]

@jakesmolka @homelab Beats me. The disks I used to get for £65-70 are now running to £150-180.

I'm starting to think the play is to disappear into the woods and never touch anything newer than a vacuum tube again.

[inkomtech@infosec.exchange]

@jakesmolka @homelab I’m looking *hard* at pivoting a bunch of my info into ‘cold’ storage: usb3 drives that are off unless needed.

[willglynn@hachyderm.io]

@woe2you @jakesmolka @rachel Yeah… you know how the DRAM and NAND markets have been (violently) reshaped by datacenters? HDDs have too.

QLC flash had been price competitive with HDDs on a total cost of ownership basis for a lot of applications for a couple years prior to *gestures around*. 6-10 TB HDDs haven't really made sense in volume – what's the market? – but hey.

Now, add a supply crunch. As QLC goes up, it shifts the bulk purchasing calculations and thus the HDD manufacturing plans. HDDs are sellable even at higher prices, especially if you can pack 28 TB into the same power/heat/size envelope.

My solution is tiering. I run a mix of TLC, QLC, HDD, and LTO-8, letting me add capacity over a wide a range of IOPS and $/TB. This is getting increasingly impractical at small scale, and IMO there just isn't a great answer for homelab capacities in the 10-40 TB range.

[isaackuo@spacey.space]

@ithoughtisawa2 @jakesmolka @homelab I agree with this suggestion that you can save money by simply not buying a replacement now (wait when/if you have a failure).

Personally, I just get the cheapest drives I can get. But the amount of critical data I have is low enough that it's all at least triplicated. So I'm okay with two simultaneous drive failures.

[b00g13@mastodon.com.pl]

@InkomTech @jakesmolka @homelab doesnt data on unpowered usb drives decay?

[inkomtech@infosec.exchange]

@b00g13 @jakesmolka @homelab good q, but I’d planned magnetic, redundant and scheduled full validation reads periodically. SMART can warn me if I’m being foolish.

[chris@mastodon.chrispelli.fun]

@b00g13 @InkomTech @jakesmolka @homelab Exactly. Unpowered drives decay. If you want cold storage look into tape or M-Disk.

[woof@meow.social]

@b00g13 @InkomTech @jakesmolka @homelab Only if they are flash/SSD storage. Unpowered HDDs are fine.

[inkomtech@infosec.exchange]

@b00g13 @jakesmolka @homelab imma need to see evidence (even a preponderance of anecdotes) to back up the ‘magnetic decays’ statements. Show me that a few days on a year is riskier than on-perpetually. Because we all see the lifespans of these drives when on. And I *have* old sub-terabyte drives going back decades that still read.