I have an old machine that has a number of hard drives in it, all 500 GB or less. The SMART status shows they’re all terribly old and are “pre-failure”. Is there any way to stress test the drives to weed out the ones that are truly bad? They’re in a zpool right now which went degraded a few months ago, presumably because of hardware issues. I’d love prune out the worst of them. Any suggestions?
Well…running a resilver on that pool will stress them quite a bit. Other than that…you can scrub the pool 24/7 (hourly schedule or more depending on how long it takes for a scrub job) or run IO benchmarks like fio
. Or both. scrub is sequential reads, pair that with a recurring fio job with random writes and those drives will have a hell of a time.
If you got a case where you can put the drives >40-45°C at all times that helps too.
But HDDs don’t play by the rules in my experience. They fail whenever they feel like to fail and not when the user is expecting/demanding it. So brute-forcing horrible conditions might lead you to safe assumptions only to get a failure a month after you considered everything running well again.
In the end these are <=500GB drives and considered e-waste by today’s standards. But even old/bad drives make good boot devices.
I use badblocks to test drives. You can use -sw to write and read the drive to find bad sectors. Just keep in mind it WILL erase the drive.
This topic was automatically closed 273 days after the last reply. New replies are no longer allowed.