Planning around older hard drives
September 17, 2018 5:30 AM   Subscribe

I have a number of hard drives, both internal and external, attached to my old desktop. Some are getting old, and are somewhat small by today’s standards. What’s the lifespan of platter hard drives? What about drives that are off 99% of the time?

Desktop has 4 internal platter drives and one SSD. SSD is the youngest, and is over four years old—I don’t recall the ages of the others, but easily 5-7 years.

I’m in the process of moving two of the internal drives to a new 4tb external, but it got me thinking about the others and how long they might have left. How do manage the life spans of your drives to avoid a catastrophic failure?

What about backup drives that may be old, but are powered up only for so long as it takes to dupe the main drive (to be clear, not just dismounted but disconnected from power).

Bonus Q: I had been using Crashplan as a cloud backup, but the home version is going away. As I transition some of these drives, I’d like to make a physical drive backup I can store offsite. What would be the most reliable way to do some cold storage of say, significantly less than 1tb? Possibly less than 128gb. A small spinning drive? SSD? USB stick? DVDs?
posted by Admiral Haddock to Computers & Internet (7 answers total) 7 users marked this as a favorite
Really, I wouldn't go for anything less than a NAS with RAID storage, alongside a cloud option. Periodically shifting data between hard disks, SSDs or whatever isn't a safe way to back things up. You can't rely on storage lifespans, because sometimes drives/USB sticks/DVDs just randomly fail, or you do something dumb and wipe them. Having said that. Backblaze are a good source of data.

Anecdata from personal experience is that hard disks are more likely to fail on systems that are powered on and off a lot. Maybe it's a thermal thing. So I'm always dubious about relying on hard disks in a cupboard being usable when connected in the future.

RAID 1, also called Disk Mirroring, is your basic way to reduce the danger of a failed drive. The data is held on two drives, each containing a copy. If one fails, the NAS (or other device) tells you, you swap out the drive, and it copies the data over to the new one. This won't save your data from a fire, theft, or human error, though. For this reason you want an off-site backup such as Crashplan, Backblaze, etc.

A simple off-site backup is great, but again won't save you from human error, unless your plan includes being able to recover changed and deleted files as well.

As I get older, I seem to get more conservative about my backups, to the point where I'd now have a really tough job if I wanted to delete something irretrievably.
posted by pipeski at 5:50 AM on September 17, 2018 [1 favorite]

A Synology NAS is in my future, but I’m working on an interim solution—just bought a MacBook Pro, and I’m trying to get my data in order so I can actually migrate to the laptop.

Getting the data in one place will help me focus on what I want to keep with the most secure back ups (and on what service). In the interim, I want to throw some of it on a drive or a disk offsite.
posted by Admiral Haddock at 6:56 AM on September 17, 2018

I'd look at the manufacturer's documentation and warranty for the specific drives you have. The big question is how your usage compares to the target duty cycle for that drive. A "consumer" drive (e.g. Western Digital Blue) has a different duty cycle than an "enterprise" drive (e.g. Western Digital Black). Consumer drives may have a three year warranty at a pretty light duty cycle, while other drives might assume a heavier duty cycle and still have a longer warranty. If your usage corresponds to a light duty cycle but you've got higher end drives with five year warranties, it's not unreasonable to expect 4+ years out of them. I'm not sure I'd be comfortable with seven years of even sporadic operation, though. As far as Backblaze statistics go, you're somewhere off to the right side of the failure bathtub curve. Even assuming you've got really reliable drives in extremely light duty, I'd be uncomfortable relying on those drives long term.

Your 99% idle estimate is probably too high, though. Every time the computer is powered on it will spin up each connected drive. It will also spin them up whenever you use the system search tools, and perhaps whenever it determines it needs to update its search indexes. If it's running Windows it may even spin them all up whenever you open the Windows Explorer file manager (or whatever it's called now). If it's a Mac it will spin them up for every Time Machine backup. If you're running a Linux or BSD you maybe have a better expectation that the drives aren't being spun up for as many file system traversals, but even so you've got to account for every startup, restart, or shutdown.

I'd look at rolling the data onto new platters and pulling those old drives out of use as soon as possible. From personal experience with questionable drives, I would maybe even go as far as to make a full copy in one operation (or in a few operations starting with the most important directories), then do any deduplication and syncing from the copy. If your drive is 5-7 years old, it might not do as well with arbitrary read/write operations as it would with one big, mostly sequential read.

My biggest fear with an old drive is running the risk of accumulating errors if there are multiple bad sectors. Drive manufacturers build error handling into their drive controllers, and they will recover from errors up to a predefined limit. At some point, however, if the cumulative error count is higher than the predefined limit, the controller will refuse to access the drive. There are drive recovery tools that might get around that, but what you don't want is for errors to accumulate up to the limit before you get the most important files off.
posted by fedward at 8:57 AM on September 17, 2018

From what I've been told you want to fire up those drives at least once every six months to do error correction and give the controller a shot at reassigning any bad sectors. Leaving an unattached drive for years on a shelf seems like it would be fine, but I guess you need to actually fire them up once in a while. Good luck.
posted by Sphinx at 9:40 AM on September 17, 2018

Thanks—narrowing in on the original question I had posed.

To be clear—the 99% estimate is not IDLE, but OFF. For the 99% drives, I mean a (for example) drive first put into service in 2012 that has been powered up for a cumulative 48-96 hours—only long enough to make the cumulative backups in the intervening six years.

Definitely sounds like the drives that are in the machine should be retired post haste.

Which still gets to the core of the original question—is the best practice to install a drive and label the housing or the machine “replace drive 3 by 9/17/21”? Do you use the warranty as a proxy for the “replace by” date?
posted by Admiral Haddock at 9:44 AM on September 17, 2018

I'm unaware of any industry best practice for that sort of very light duty. If you're only using these drives for manual backups at long intervals I'd be concerned about failure modes for which there's not a lot of data. That's not even a month worth of uptime. Based on operational hours you could still experience the "early" sort of failure from the left side of the bathtub curve even though your drives are old in calendar years.

Hard drives generally aren't supposed to be great for long term storage (my own habits notwithstanding). Are you verifying the backups on these drives when you add new data, or are you doing more of a write-once-and-forget-it kind of thing? If you're periodically rewriting the data (as documented in that article) then I'd be a little less concerned, and as long as you have two or more backups I'd just keep doing what you're doing, especially if your backups are automatic. If your backups are all manual, if you're not verifying and/or rewriting the data, or if you only have one known good backup right now, I'd figure out a better strategy for automation, rotation, and refreshing, and probably start those new habits with some new drives.
posted by fedward at 12:08 PM on September 17, 2018

On belated review, I didn't even address the Do you use the warranty as a proxy for the “replace by” date? question.

Personally? I use my drives until they die or get too full to be practical. Usually "too full to be practical" happens first, and I pull the drive, label it, and throw it on a shelf with the other decommissioned drives. I have multiple active backups. If a system drive dies, I replace it and restore its data from whichever active backup is most recent, but usually they're just a couple hours apart at most. If a backup drive dies, I replace it and generate a fresh starting point from every machine that's backing up to it. Luckily I haven't yet had two backup drives die simultaneously.

If you're rewriting regularly and you have multiple backups I'd think using your drives until they die is a reasonable strategy. I wouldn't personally pull a drive just because it's old or out of warranty unless I so badly needed to avoid unscheduled downtime that it was cost-and-time-effective for me to schedule downtime just to roll everything over. For personal use? I don't need that kind of availability.

FWIW I dig into old pulls once or twice a year, if I'm missing a file that was on an old machine or if I need to revert to a really old version of a file that's too old for my current backups. Since my current backups are regularly refreshed and it's really rare that I'd be missing a file, I don't really have a rewrite policy for my old pulls. I probably should, but in general the way I'm rolling data over means I've generally got at least two freshly written backup copies of everything I've touched on any computer I'm still using. I might not have freshly written backup copies of stuff on older computers if it wasn't rolled over when I decommissioned those computers, but that's rare.
posted by fedward at 1:14 PM on September 17, 2018

« Older I can't stop... [dubstep noises]   |   Overcoming Anxiety for Terrible Tasks? Newer »

You are not logged in, either login or create an account to post comments