Hard Drive Longevity

[rpmurray]

I've heard that the accepted wisdom is that you should replace hard drives every three years because the older they get the more likely they are to fail. I've never actually followed that advice because I keep daily backups so I figure they should run until they die.

What's the longest you've ever had a hard drive running continuously? I have 4 1TB Western Digital HDs (WDC WD1001FALS-00J781) running in a Mac Pro 2008 nearly continuously (24/7/365) since 2009, other than for updates that require a shutdown or a reboot and the rare power outage. The Mac Pro has been hooked to various APC Smart-UPS over the years, currently a 1500VA. So far they all show no errors when checked with SMART Utility (using both the short and long tests) and have over 78000 hours on them.

Back in the day I tested them with SpinRite before putting them into production to make sure they didn't have any bad sectors. Nowadays I use the disk certification feature in SoftRAID to do the same for drives I use with my other Macs.

 

[bsbeamer]

I've had drives in some machines run 10+ years without issues. Had one transported from internal G5 Mac tower to a 1,1 and then to a 3,1 at a client's office (in external case) and believe it's still "going" today, but not connected to that 3,1 regularly. Have had others in a SCSI 12+ bay server RAIDs fail almost every 3-4 years like clockwork. Some others in SATA 8-12 bay RAIDs would go 7+ years without even a hiccup. Almost all those drives were either 7200rpm or 10,000rpm.

DriveDx to monitor health MAY help give you a head's up about longevity. Been using it for some time on personal machines.

If the data is important, have a clone to another drive in a safe place. Carbon Copy Cloner can clone the entire drive contents easily. Depending on what the data is and how frequently you're changing or adding to it, a simple weekly or bi-weekly schedule may be enough, especially if you're using in conjunction with Time Machine or another backup.

 

[Zeke D]

I had a WD apple-branded SATA 160GB drive that came with my G4 Digital Audio on an add-in card that was transferred with the card into my Mirrored Drive Door, then into my single-CPU G5, then dual CPU G5, then my mac pro 1,1, then 3,1, then 5,1. When I prepped the 1,1 for sale, I moved it back into it. I'd say 12 years?

 

[Macschrauber]

I just pulled two Apple branded WD Drives from Mac Pro 5.1, both with too much pending sectors and just a few 1000 hours.

I'd replace the drives and keep them for a 2nd backup

 

[AidenShaw]

I just pulled two Apple branded WD Drives from Mac Pro 5.1, both with too much pending sectors and just a few 1000 hours.

I'd replace the drives and keep them for a 2nd backup

A few "pending sectors" doesn't mean much. A few dozen, or a few more each week, is cause for serious worry.

 

[Macschrauber]

That drives were dead, beachballing and the amount of dead sectors were counting up...

 

[ixxx69]

I've heard that the accepted wisdom is that you should replace hard drives every three years because the older they get the more likely they are to fail.

That is not accepted wisdom... that's some random internet dude who doesn't know what they're talking about.

 

[flyingmanatee]

Really, you should just back up any important data and use a drive until it starts showing warning signs of dying or dies. I have several HDDs that are well over 8 years old. All three internal Seagate drives I owned died in a 1 year span of each other (a 1.5 and two 2 TB) after 4 years, whereas I have an old 2.6 160 GB Western Digital HD that's an external that's lasted 12 years. Make/Model has a lot to do with it, backblaze publishes quarterly drive failure rates if you're curious.

Notably, Western Digital Black drives come with a 5 year warranty, which is entirely defies said "every three years" quip.

 

[TC_GoldRush]

I have an old ibook g3 Clamshell with all original parts, including the ancient 3gb HDD, which is nearly 19 years old at this point. It's a combination of upkeep (dusting, cleaning, etc) and luck (not dropping it, and praying it doesn't go).

 

[kschendel]

One of the 4 original drives in my (4,1) died after almost exactly two years. The other three ran without issue until I replaced them a couple years ago, when they were about 7 years old. I have some 18 year old drives in the Sun Ultra 60 in the back basement, but that machine only runs a few dozen hours a year any more; I'd guess that the cMP drives had more hours when I removed them.

Drives don't need to be replaced on a schedule. Back up properly, test the backups occasionally to make sure they are working and readable, and have a replacement plan. For old drives (more than a few years old), proactive replacement might be a good idea if losing a drive at the wrong time will cost you income; or, if you're upgrading for performance (which was my case). Otherwise I'd agree with running them until they fail or show signs of distress.

 

[HDFan]

See Backblaze numbers (may be a bit dated):

https://www.backblaze.com/hard-drive.html

78% last > 4 years
median 6 years
Seagate, Western Digital variable reliability

 

[AidenShaw]

See Backblaze numbers (may be a bit dated):

https://www.backblaze.com/hard-drive.html

78% last > 4 years
median 6 years
Seagate, Western Digital variable reliability

Take Backblaze with a heaping teaspoon of salt.

They admittedly buy the cheapest disks available - disks designed for light duty desktop usage.

And they run them in a heavily loaded 24x7 enterprise server environment. (The automotive analogy is that they buy Honda Fit subcompacts and tow 24 ft trailers in the mountains.)

Look at their statistics in that context - but if you're going to be running 24x7 loads buy enterprise disks, not the cheap stuff that Backblaze buys. (Not to say that Backblaze isn't doing the right thing for their business model - but you're probably not looking at ROI over tens of thousands of drives in an infrastructure that is very resilient to drive failures.)

And also note that some drives are born bad. A few years ago one model of 3TB Seagate drives were crap. Not "Seagate drives are crap", but that "one model of Seagate 3TB drives are crap". That bad drive still skews Backblaze's statistics.

I have several hundred drives spinning - all high end enterprise drives (and mostly Seagate). I get one or two drive failures per year (and at this point the average age is approaching 3 years). My typical disk "drive" purchase is for a 96TB or larger shelf for a SAS extension.

I never lose data though - with RAID-60 and hot spares set for predictive failure I get an email alert that S.M.A.R.T. data showed that a drive was predicted to fail, and has been already replaced by a hot spare. (I do have daily backups - but those are usually needed only when a user types the "rm" command when they shouldn't - not because of a storage system failure.)

 

[HDFan]

Take Backblaze with a heaping teaspoon of salt.

If someone has a disk reliability dataset that is better than Backblaze I'd be interested in hearing about it. In the 2017 failure report comparing the failure rates of the Seagate consumer and enterprise 8 TB drives (ST8000DM002 and ST8000NM055) with about a million drive days each the consumer drive has a lower failure rate than the enterprise drive (1.13% vs 1.22%). This is in a production environment.

https://www.backblaze.com/blog/hard-drive-stats-for-2017/

I have several hundred drives spinning - all high end enterprise drives

And they run them in a heavily loaded 24x7 enterprise server environment. (The automotive analogy is that they buy Honda Fit subcompacts and tow 24 ft trailers in the mountains.)

Admittedly I run maybe 24, not hundreds of drives. But 200 drives still gives just ~220000 drive days in 3 years vs Backblaze's ~1,000,000 for ~1 year. And the one time I did spring for one very expensive HGST helium drive it failed.

One can argue about Backblaze statistics, but looking at Amazon you pay about $100 more for the enterprise drive. Why throw away away $100 when you get about the same cost/failure ratio with the consumer drive? If my $16K Honda fit can pull that 24 foot trailer safely without issues, and lasts for 200,000 miles I personally don't care that I don't look super cool driving a $52K Ford F-450. I could do a lot of things with $36K.

 

[Mikael H]

@rpmurray
There's usually no point in preemptively replacing drives that haven't gone bad yet. A good backup strategy means you'll live even if a drive dies; that's the most relevant factor to take into account when discussing data storage.

If you can't live with your system being down while waiting for new hardware, invest in a RAID solution (software or hardware based), but read up on pros and cons of the various RAID schemas available with regards to your use case. If possible do some benchmarking with a production-like load so you can make an informed decision on security vs performance vs investment.

 

[AlexMaximus]

I've heard that the accepted wisdom is that you should replace hard drives every three years because the older they get the more likely they are to fail. I've never actually followed that advice because I keep daily backups so I figure they should run until they die.

What's the longest you've ever had a hard drive running continuously? I have 4 1TB Western Digital HDs (WDC WD1001FALS-00J781) running in a Mac Pro 2008 nearly continuously (24/7/365) since 2009, other than for updates that require a shutdown or a reboot and the rare power outage. The Mac Pro has been hooked to various APC Smart-UPS over the years, currently a 1500VA. So far they all show no errors when checked with SMART Utility (using both the short and long tests) and have over 78000 hours on them.

Back in the day I tested them with SpinRite before putting them into production to make sure they didn't have any bad sectors. Nowadays I use the disk certification feature in SoftRAID to do the same for drives I use with my other Macs.

The WD Black drives are very good drives in my book of experience. I think you have nothing to worry because the there golden points are met:

- No excessive heat
- No power fluctuations or peaks through UPS
- Most likely a low vibration environment -> low risk

There have been models that kind of sucked big time or had a higher failure rate in the past. Older WD Raptor 10K drives are an example for that or the ones Aiden Shaw mentioned above.
But since you have your Black HDDs in a MP 3.1 you most likely don't have cooling problems. A stuffy, cramped and hot server environment would be a whole different story. Technically your 800 MHz Ram modules are at higher risk failing then your HDDs. The 3.1 is known for very hot running RAM.

 

[HDFan]

There's usually no point in preemptively replacing drives that haven't gone bad yet. A good backup strategy means you'll live even if a drive dies; that's the most relevant factor to take into account when discussing data storage.

I agree. Replacing a drive that hasn't failed means you are throwing away money if it actually would run for x more years. Personally I find it's a psychological issue. If your system is designed to handle failures then you aren't constantly worrying about whether a drive is going to die, or if you are throwing away money by replacing them prematurely. But some people like stress ...

 

[InuNacho]

Here's my unscientific anecdote on hard drive lifespan. I work in computer surplus/recycling and we take drives out of just about everything that has a hard drive and test to see if they work. Out of the tens of thousands of drives that pass through our facility, only a small fraction no longer work and those that don't typically come from the SCSI and IDE eras or are laptop drives that have seen regular physical movement.

 

[kschendel]

I agree. Replacing a drive that hasn't failed means you are throwing away money if it actually would run for x more years.

Yes, but unfortunately the only way to know if it will run for x more years is to try it. I don't generally approve of replacing hard drives on a time basis, but sometimes the cost of downtime exceeds the cost of replacement, and in those cases you aren't throwing money away, you're saving it. It all depends on the situation.

 

[bsbeamer]

Having clones of your data as part of the backup plan would be another way to minimize risk, downtime, and overall costs.

Carbon Copy Cloner can be setup to automatically clone and/or just update clones upon the clone drive's connection to the system. I use it to basically have version clones of the system drive, attempting to keep one version or OS update back (rotating between two drives). Also have separate clones for all media drives.

One point some people do not always 100% get with system drive backups/clones - if your system drive is an SSD, your clone does not NEED to be an SSD. You can clone to an HDD or USB external hard drive. Recommend formatting the clone drive as HFS+ (not APFS) and update at least weekly or bi-weekly depending on usage.

 

[CapnDavey]

hell I still have drives running in older power mac's that are almost 20 years old

 

[Vinegar Tom]

I'm writing this on an 2008 iMac and have no problems with the HD. Admittedly, the Mac is only used for internet/email/word processing so the HD isn't put under too much stress, but so far, everything fine...