Hacker News Reader

Comment by not_your_vase | Hacker News Reader

A few years ago we had capacitor plague. Are we living now the storage plague? It's getting ridiculous that all storage is getting worse and worse. WD is making HDDs crappy with SMR, manufacturers says that 3 years operating time is already too much for SSDs and HDDs, and they don't joke. I just had a Kingston SSD (okay, that was like 8 years old) and a portable WD HDD (~2 years old) die just this year.

The internet is full of problems lately about data loss and longevity issues.

I remember 20 years ago HDDs were not meant for eternity either, but they definitely outlived the usefulness of the computer that they were bought with...

View on HN

I've had a lot more luck with hard disks these days than I did 20 years ago. Remember 20 years ago was the era of the infamous IBM Deathstar drives where the magnetic coating would literally start sprinkling off the platters. Also the era of terrible terrible Maxtor drives that died in 1-2 years, which Seagate then bought and made their drives also unstable for a while. I ran a server with around 8 drives and had to keep replacing disks at the rate of about one per year.

Meanwhile today I'm helping admin a ZFS server with 20+ drives and drives have about a 4-5 year lifespan.

> but they definitely outlived the usefulness of the computer that they were bought with

Computers were also much more quickly obsoleted back then. When today a 6 year old computer is totally useable, back then you really felt it if your machine was just 3 years old.

I remember those IBM Deathstar drives...

Was at the Aussie Tribes 2 launch LAN and there was a guy who had one die on him.

At that time in the LAN scene there would always be someone who had a Deskstar die ... You could hear the clicking over the noise of the LAN.

I realised back then, I can only trust Seagate.

Hah, I had a Deathstar die on me back in the early 00s too. Surprisingly, about a decade later I hammered it with ddrescue and was able to get almost all the data off it!

The memory of the Deathstar drives that stands out most in my mind was a coworker managing to destroy hardware with SQL. We were at a really ... frugal ... interactive advertising firm and our dev server had been slapped together with a RAID-1 array of cheap IBM drives. One day said coworker was testing conversion of a large database table in MySQL from MyISAM to InnoDB format (to see how long it'd take, what query perf afterwards was like, etc.) and all of a sudden the server went hard down. We went over to the server closet and discovered that the IO had been enough for both drives to grenade themselves at the same time. Good times. I'm just glad we had semi-decent backups and it wasn't a production machine.

> Remember 20 years ago was the era of the infamous IBM Deathstar drives where the magnetic coating would literally start sprinkling off the platters.

I have even older memories of problematic IBM drives. During the early 90s the shop I briefly worked with, found a supplier for IBM SCSI drives at a very convenient price, so they ordered a good lot of them. They worked great on PCs, but some of us also had Amiga machines and of course would love to benefit from the offer. So we tried one, but it didn't work; then another, and another; nothing, they were normal SCSI drives but refused to work on any Amiga with a SCSI controller, although any other drive would work in there. In the end we abandoned all hopes and took the drives for a reformat to be used on PCs, but... they were all dead. Completely, not even detectable by any controller; the mere connection to an Amiga SCSI controller destroyed them instantly. We never discovered where the problem was; those drives worked perfectly on all PCs, while we could install any other drive on every Amiga and expect it to work, but no way to put those in an Amiga and expect it to survive. Good old times indeed:)

4-5 years? My old backup server has almost been running for 15 years now, 6x 500GB hard drives, one is even running on PATA. 8GB ECC ram, Athlon II. I could save some money replacing those 6 drives with 2x 14TB hard drives today. But as long it is working fine I ain't gonna do something before I run out of hard drive space.

Seagate had 1.5TB drives I think about 15 years ago with really high failure rates. Somewhere I saw close to 33%. Anecdotally, mine failed after about a year and the refurb warranty replacement also failed after about a year.

I think it was about 10 years ago some Seagate drives had higher than industry failure rates. Iirc one of their factories was producing drives that failed much more frequently than others (there might have also been something to do with the platter counts/model)

Don't worry, your chips will start glitching after a few soon too.

We're hitting scaling limits. Exponential growth is slowing.

that does not make any sense if you are talking about the controllers.

https://support.google.com/cloud/answer/10759085?hl=en

> I remember 20 years ago HDDs were not meant for eternity either, but they definitely outlived the usefulness of the computer that they were bought with...

Anecdotally I remember HDDs failing sometimes for me and my friends/relatives back in the day. Now it barely happens for SSDs. Hell, even supposedly problematic old Intel from late 2000s still works fine in the same old MacBook I gave to my mother after using it for years.

I wonder what it the actual data regarding this.

All WD drives I bought in the last decade work as expected (0), including the NAS ones bought after the introduction of that SMR thing; I just made sure they are either Plus or Pro, not the plain Red ones, which are SMR-plagued. I was also lucky with SSDs, but especially on desktops I use small ones as I still prefer to keep /home dirs and RAID arrays on old rusty drives.

0: A couple exceptions: Two WD Red (before SMR) which I took out from my old NAS to put bigger drives in place, and put in a drawer while they were still perfectly healthy. After like 2.5 years in their anti static bags and normal conditions, no excessive heat, no moisture, no magnetic fields etc, I took them out because I needed a spare disk and checked them: both were not working, one completely dead and the other barely recognizable but unreadable. The first didn't even show up once connected; I tried to clean all contacts, including the pins on the controller pcb to no avail, and eventually had to ditch it; the 2nd one could be reused only after a full reformat; no way to recover old data, not even using testdisk. I never experienced nor expected anything like that, and frankly it worries me quite a lot.

~20 years ago was the HGST "deathstar," another drive so bad there was a class action lawsuit filed. Disks have always randomly died, that's part of the reason Sun made zfs.

Reply on news.ycombinator.com