What sort of "burn-in" testing is recommended for newly acquired hard drives?
- Thread starter NeverDie
- Start date
Notice: Page may contain affiliate links for which we may earn a small commission through services like Amazon Affiliates or Skimlinks.
I generally do a re-initialize test with Hard Disk Sentinel. I use badblocks on linux, but I can't remember what command specifically. It takes several days on linux.
I found some links in post #16 here: https://forums.servethehome.com/index.php?threads/hdd-test-recertify-options.3942/#post-33644
By the way, is there any statistically validated approach to this? At some point diminishing returns kicks in. If Google or Facebook burns in all their drives a certain way, then there's probably a rationale for why they do it that way, instead of simply throwing the kitchen sink of long-running tests at every drive and hoping it's the right thing to do based on just a theory but little evidence. With large disks, these tests can take a long time to run. In particular, the number of iterations to do obviously has a profound effect on required time. Hypothetically, if 1 iteration catches 99.99% of problems and 5 iterations catches 99.991%, then I'll take the 1 iteration, thank you.
By the way, is there any statistically validated approach to this? At some point diminishing returns kicks in. If Google or Facebook burns in all their drives a certain way, then there's probably a rationale for why they do it that way, instead of simply throwing the kitchen sink of long-running tests at every drive and hoping it's the right thing to do based on just a theory but little evidence. With large disks, these tests can take a long time to run. In particular, the number of iterations to do obviously has a profound effect on required time. Hypothetically, if 1 iteration catches 99.99% of problems and 5 iterations catches 99.991%, then I'll take the 1 iteration, thank you.
Last edited:
I just typically transfer large 3-4 gb files for a day or two. So queue a bunch of files to it. There is probably a better way to do it. Other times I just copy and paste large folders, like a windows install.
The idea is that if the harddrive is going to fail, you do it early before you put it in production. You will read dumb things on sites about how people thought this would make their drives faster...it was never about speed. It was about reliability.
The idea is that if the harddrive is going to fail, you do it early before you put it in production. You will read dumb things on sites about how people thought this would make their drives faster...it was never about speed. It was about reliability.
I guess fairly soon I'll be trying FreeNAS's recommended burn-in: [How To] Hard Drive Burn-In Testing | FreeNAS Community
I've started the FreeNAS burn-in procedure (above). What I'm finding I like about it is that all of the drives can be tested in parallel. So, if you have a lot of drives, it doesn't really take any meaningful amount longer than if you were testing just a single drive.
Update for anyone curious: I'm 30 hours into running the badblocks test. The first pattern is 0xaa. The second is 0x55. The third is 0xff. So I'm guessing the fourth and final pass will be 0x00 when that pattern comes up.
All told, for 3TB WD Green drives it will probably take about 72+ hours to run all the tests as prescribed in the link above. Testing nine drives in parallel the load on my E3-1230v3 CPU is just 0.11, so I'm guessing the number of drives isn't slowing down the testing much, if any. Thus far, no errors have been reported on either badblocks or the SMART self-tests that preceded it.
I'm making this post for anyone who might want a yardstick for gauging how long a burn-in will take.
As drives continue to increase in capacity, I hope someone comes up with a faster way!
All told, for 3TB WD Green drives it will probably take about 72+ hours to run all the tests as prescribed in the link above. Testing nine drives in parallel the load on my E3-1230v3 CPU is just 0.11, so I'm guessing the number of drives isn't slowing down the testing much, if any. Thus far, no errors have been reported on either badblocks or the SMART self-tests that preceded it.
I'm making this post for anyone who might want a yardstick for gauging how long a burn-in will take.
As drives continue to increase in capacity, I hope someone comes up with a faster way!
Last edited:
