Batch testing HDDs?

[TeeJayHoward]

I've got about 50x 2TB SAS drives that are used, of an unknown quality. I'd like to toss them all into 2x SC846 chassis and find out which ones are good, which ones are bad, and which ones are "meh". Does anyone have a utility they use for things like this?

 

[Zarathustra[H]]

My thought on HD testing if throwing them into a RAID is, why waste the time?

If they are bad, you'll find out enough, and if you are smart enough to use RAID6/RAIDz2 you won't have to worry about losing your data.

Just fire up an array, and if any drive causes problems, replace it

 

[TeeJayHoward]

Just fire up an array, and if any drive causes problems, replace it

I currently have two drives sitting at home that, if plugged in to a running Solaris 11.1 box, will crash it. Just by being plugged in. I have another drive that will only transfer data at 15MB/s - 10% of the speed it was transferring at a few months ago. I have another drive that will light up the "faulty disk" light, despite having nothing wrong with it. I have another drive that keeps the "disk activity" light permanently lit, even when it's unconfigured. When dealing with this many used disks, I like to know the "character" of each disk as much as possible.

But that's not really related to the question, just something I thought was interesting. I can probably run a SMART short test across each drive. If they pass, I call 'em good, plug 'em in, and start using them. I was hoping for something a little more mature, though. The magic of the internet returned "badblocks". I might make a new VM, pass through the controller, and give that a shot. Any other ideas?

 

[Zarathustra[H]]

I currently have two drives sitting at home that, if plugged in to a running Solaris 11.1 box, will crash it. Just by being plugged in. I have another drive that will only transfer data at 15MB/s - 10% of the speed it was transferring at a few months ago. I have another drive that will light up the "faulty disk" light, despite having nothing wrong with it. I have another drive that keeps the "disk activity" light permanently lit, even when it's unconfigured. When dealing with this many used disks, I like to know the "character" of each disk as much as possible.

But that's not really related to the question, just something I thought was interesting. I can probably run a SMART short test across each drive. If they pass, I call 'em good, plug 'em in, and start using them. I was hoping for something a little more mature, though. The magic of the internet returned "badblocks". I might make a new VM, pass through the controller, and give that a shot. Any other ideas?

Wow.

Interesting. I have never had a drive crash a system simply by being plugged in! Granted, I also have never used used drives, and I just replace them at the first sign of anything gone awry in the slightest.

 

[TeeJayHoward]

It looks like there were more disks than I thought! Score!

SMART RESULTS:

50 drives tested
3 failed immediately (1x not powered on, 1x not detected, 1x failed test). Those were set aside.
47 passed a SMART short test.
3 of the drives that passed had issues of some sort (noise, failure prediction threshold, etc). Those were set aside.
4 of the drives were of a different model than the rest. Those were set aside.
The remaining 39 drives were tested for a SMART long test overnight.
34 drives passed the SMART long test. 3 failed, 1 dropped, and 1 pretended that the test never ran.

Now I need to speed test these 34 drives to ensure I don't have a repeat of the 15MB/s disk. Anyone have any ideas on how? I'd like to test inner track r/w speeds and outer track r/w speeds, but I don't want to waste time filling the whole disk.

I've got another 9 (I think) disks to test, and then I'd like to long-test the 4 drives with the different model number. I had originally hoped that 48 of the disks would be good so I could just swap out my 18x3TB drives and fill the SC846s. It doesn't look like that's going to be the case. Instead, now I'm thinking of putting all 18x3TB drives in one SC846, and just having 24x2TB drives in the other... With a hell of a lot of spares. Seems so wasteful, though.

Blah, blah, blah. Picture time!

Disks:

Removed from packaging. Note left disks have notes on them listing defects:

Testing in progress:

When I stopped for the night:

The testbed:

As it sits this morning: