15

For a hard drive with SMART, how often should the quick test be run, and how often should the full test be run? Should the tests be run more frequently as the drive gets older? Does the manufacturer and drive-type matter?

Note that I do make regular backups, so I'm not depending on SMART to avoid backups until the last possible moment. It's just that I'd like some advance notice if anything goes wrong.

2
  • Is this a desktop computer that you activly view, that you know what is going on with? or a computer your in charge of but are never at, a server (even media server) or you connect remotly and would want to know why something does not work?
    – Psycogeek
    Commented Feb 13, 2014 at 4:17
  • If it is a computer you normally use, there are utilities that load at boot and run in the background that will monitor the SMART values and alert you to a problem. I rarely use Windows so I'm not up on current offerings, but do a Google search to see what's available now.
    – fixer1234
    Commented Dec 26, 2014 at 20:55

3 Answers 3

18

I agree with @user1433123 answer about the bad advice from other answers suggesting you ignore or delay running any SMART test until you have a disk problem. The purpose of the SMART attributes and self-tests is to anticipate disk problems when possible.

The minimum you should do is to enable the Automatic Offline data collection and check the SMART attribute values once a month or so, even if you don't run manually the SMART self-tests. (Note that some SMART attributes are not updated unless the offline data collection is enabled).

That being said, running the short/long self-tests (manually) too frequently in a desktop computer may be a waste of time. In a server the best approach is to use something like "smartd" (from the smartmontools software, Linux and Windows) to monitor and execute the tests on a regular basis, so that you will be notified when a possible problem arises.

Now, regarding your question about how often to run them: For a server I suggest to enable the offline data collection always and run (automatically, using smartd or similar software) the short self-test daily or at least weekly, then the long self-test monthly. Note that the self-tests should not cause any performance problem.

For a desktop, if you have to do it manually, I would run the short self-test monthly and just enable the offline data collection, maybe run the long self-test every couple of months or so. If you can use automated tools like smartd, then you can do it more frequently.

For portable disks (like USB removable disks) I suggest to run it whenever possible, as hard disk in movement are somewhat more prone to problems and the SMART values may give you early warning. For example, we use some removable hard disk for offsite backups and the backup script we use dump the SMART report of the program smartclt to a local log and run a SMART short test every time we make a backup (weekly); while the short test is running (it takes 3 to 5 minutes) the backup people read the SMART report looking for any anomaly.

So in summary, I would say that:

  • Always enable the SMART offline data collection.
  • The more critical the data, the more frequent you should run the SMART self-tests.
  • Always try to use automated software that run the tests for you.

Check the man page of smartctl (from the smartmontools) for some information about each kind of test (http://smartmontools.sourceforge.net/man/smartctl.8.html), even if you use a GUI program like GSmartControl to run the tests.

Regards, MV.

6
  • 1
    To achieve this I ran: smartctl --offlineauto=on /dev/sda But, it sounds like this could also be achieved by editing /etc/smartd.conf to have: /dev/sda -o on -S on -a -m [email protected] -s (S/../.././06|L/../01/./07) where "-o on" enables offline tests, the "S/../.././06" runs short tests once a day at 6 am, and the "L/../01/./07" runs a long test on the first day of each month at 7 am. I have not confirmed this information is correct, only this is what I tried. Commented Nov 30, 2019 at 19:02
  • Does running the short test contribute to head wear at all? I have it scheduled through smartd for hourly. It makes quite a bit of noise when it runs, which made me wonder... Commented Jan 31, 2020 at 0:25
  • @IanSchoonover : It should not cause more stress than normal usage. However, running the short test hourly may be too frequent and may not offer much benefit. Anyway, do the following questions/answers help you?: superuser.com/questions/1022628 and superuser.com/questions/378067
    – MV.
    Commented Jan 31, 2020 at 15:17
  • What if I have a personal NAS server in 3 disks and the one I use more there can go 2-3 weeks minimum without me touching on it? Should I put the fast test once in a month and the long one once in 3 months on all disks, for example? 1 month fast, 6 months long? (1 year long!?) Or that might be too late sometimes? Note: all disks spin down in 4 hours without use.
    – Edw590
    Commented Mar 2, 2021 at 15:38
  • 1
    @DADi590 for cases like that when the disks are unpowered or sleeping most of the time (like the backup example I mentioned above) running the SMART tests manually is, in my opinion, fine. Just be sure to enable the Automatic Offline data collection. Or you can automate the testing with time intervals similar to your usage (weeks).
    – MV.
    Commented Mar 3, 2021 at 8:04
6

"You really shouldn't need to check the HDD until it is near the end of it's warranty period. In which case I recommend you throw away the drive. Rather then to chance the fact that it is near sudden unpredictable death."

This is bad advice. If you are worried about your drive failing, then you should run smart tests on it so you can identify the failing drive quickly and take action if needed.

Smart is designed to warn you of a drive failure before it happens, whereas good hardware like a RAID controller will let you know once it has failed.

Warranty length is not an indication of drive health.

My advice to you would be

  1. Backup
  2. Run a daily SMART short test on critical drives like in home server's RAID array or NAS devices if they support it.
  3. If the drive(s) is especially critical run a weekly long test in off hours to identify a failing drive as soon as possible before the data is unreadable.

After you identify an issue, RMA the drive if it is under warranty, or purchase a replacement if it is not.

If uptime is required then invest in a simple RAID solution like RAID 1/5/6 if the data needs to be available 24/7.

I don't know of a single person who is proactively replacing their laptop hard drives every 1-3 years depending on the warranty.

3
  • 1
    Running tests on RAID drives is especially important. Consider the following sequence of events on a simple RAID 1 (mirroring) system containing 2 drives : 1) Drive 1 silently becomes degraded. By luck, nobody accesses bad sectors on drive 1. 2) Drive 2 fails catastrophically. Raid software reports failure. 3) Administrator replaces Drive 2. 4) Raid software rebuilds drive 2. While drive 2 is being rebuilt, drive 1 fails catastrophically. 5) Data loss occurs. Data loss could have been prevented if drive 1 had been replaced between step 1 and step 2.
    – Brian
    Commented Oct 11, 2017 at 17:26
  • 1
    Please note that the sequence of events I described in the above comment is based on personal experience. A failing drive is more likely to die during a rebuild (because the drive is experiencing more reads than normal and those reads are touching more of the drive). Here, the goal of testing is to reduce the likelihood of overlapping drive failures by narrowing the length of time a failing drive is in use (by detecting the failure sooner, allowing the drive to be replaced).
    – Brian
    Commented Oct 11, 2017 at 17:32
  • I use ZFS which will detect bit rot situations and is capable of fixing them as well, on extended SMART tests, I think there is such a thing as too often. On my NAS I currently do a scrub once a month (bitrot check on ZFS) and a short SMART test once a month as well, and have automated checks monitoring key SMART parameters. I have never had a spindle fail without warning signs on SMART first in 2 decades and a bit of using computers and managing servers. Best thing is keep drives cool and keep them spinning 24/7. Then they will last probably at least double warranty period.
    – Chris C
    Commented Jul 24, 2023 at 21:55
0

I wouldn't run it unless I suspected that something was wrong - weird sounds, or something seemed off. That said, if you absolutely had to, I'd run a quick test once a month - my laptop comes with software that does that automatically, or install something that does it automatically, like acronis drive monitor.

I'd say that drives failing smart tests are a sign something definitely is wrong, but you can have issues with drives that seem fine. Sudden death happens - life's too short to keep checking your disks, but not too short to do regular backups.

1
  • 4
    By the point you notice something is wrong, it is often too late, as a lot of low-level errors get handled transparently to the user. The point of running SMART tests is that it can give you a heads-up that the drive is getting less reliable. Commented Jan 15, 2020 at 23:20

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .