Hacker News Reader

Comment by masfuerte | Hacker News Reader

I've had several hard drives fail over the years. They all signalled their impending doom by making interesting noises so I never lost any data. When my SSD fails I think I'm going to miss the noises.

View on HN

I've worked with lots of (enterprisy) hard drives and SSDs. The hard drive failure rate was easily 10x the rate of the SSDs, but the SSDs basically never gave symptoms of impending doom. They just disappeared from the data channel. We did have one group of drives that gave symptoms... 50% of the batch dropped dead within 2 weeks of entering service, a good number failed over the time it took to get the rest of them replaced. And one drive that had very poor throughput for several minutes did show SMART metrics that it was reallocating sectors, but we took it as a sign of death and replaced it; I think it might have worked fine once it calmed down, but we'll never know.

This, SSD failure modes are scary! Even enterprise SSDs will just go from "all ok" to "no device connected" at any random time.

So far, most of the HDDs I've seen fail, gave _some_ indication prior to it, via smart or "strange" behaviour.

Which is weird because the only microSD card I've ever had fail on me locked itself into write-protect mode, allowing me to get my data. I wonder why SSDs don't do the same.

They're supposed to, and I've heard anecdotes that they sometimes do, but it seems like many of the failures I've experienced were firmware failures (in some cases, upgrading firmware on remaining drives seemed to result in lower failure rates anyway). Firmware failures have no rules; spinner drives sometimes have them too, and they might disappear from the bus as well, although usually those will "just" disappear after a reboot, not in the middle of a request.

Anecdata is interesting. Here's mine, I've recently had a spate of a few Dell system 3-5 years old getting progressively more sluggish. Quick tests don't turn up issues. When I'm given the time required I use the hard drive manufacturer's testing utility and do thorough testing at which point damaged sectors start to turn up, not too many but enough to cause seek issues or something. I use chkdsk to attempt to recover data and mark the problematic sectors and at least for a while the system are a bit better. Of course, as soon as possible replacements are needed.

One day I rebooted my desktop and my C drive SSD was just gone. No warnings of any kind, great performance until it died.

I've always distrusted SSDs, and manufacturers continue to use lower quality controllers and flash, including sometimes changing a currently great drive to use cheaper components and sell it as the exact same device.

Meanwhile games load from my spinning rust within a minute, and SSDs never seem to actually improve that that much, because the code responsible for parsing and loading RAM seems to matter more.

Reply on news.ycombinator.com