Checking drives before use

L

LordCalin

Gawd
Joined
Oct 5, 2009
Messages
901
I'm working towards building a storage server with 6x WD20EARS drives.

I just got 4 of them because the price was insanely cheap, and the last 2 I won't be able to get before the 15 day direct exchange runs out with the store I bought from.

What's the best way to verify the drives I did get? low level format, or just running wddiag in extended test mode.
 
If you are familiar with linux do a 4 pass badblocks read/write test and look at the SMART. If you are not familiar with linux do a full format on all drives and look at the SMART data with a program like CrystalDiskInfo.
 
Whenever I get drives to add to my raid I run a full pass or two of the manufacturer's extended diagnostic test
 
I do a 4 pass destructive testing with badblocks. And yes it takes over 30 hours for a 4 pass test on a 2TB drive. You can do less passes if you want..
 
OldSchool said:
Whenever I get drives to add to my raid I run a full pass or two of the manufacturer's extended diagnostic test
Click to expand...

I do not recommend this. I have seen several cases here at work where the manufacturer tests passed for very bad drives. 1 case was a Seagate drive that had 1/3 of the drive was totally unreadable but the drive passed all of the Seatools tests.. WDC testing software has also failed as well here at a work.
 
yeah i've seen the same, in fact I have a customers system on the bench beside me that wddiag keeps saying is fine, but within 6 hours of use windows stops working with file system corruption and chkdsk shows bad blocks.
 
For each new drive I receive, before I trust it in an array expansion/build, I do the following:

1. Check SMART values to ensure drive has not been used (low power on hours count)
2. Use HD Tune to do a full read scan
3. Use HD Tune to do a full write scan
4. If a WD drive, perform the full destructive test using their WinDLG tool
5. Check SMART tables again to ensure there have been no reallocated or pending bad sectors

+1 on the manufacturer's drive tool to determine drive health. SMART values after some read + write load testing are the best indications IMO.
 
BadBlocks just seems to take an incredibly long time. Are there certain parameters I should be running it with to speed it up? A 40GB drive will take over a day to scan for me.
 
Code: 
badblocks -wsv -p 1 /dev/sdX

where X= your device

You can also speed it up by increasing the block size.

-b 16384
 
drescherjm said:
Code: 
badblocks -wsv -p 1 /dev/sdX

where X= your device

You can also speed it up by increasing the block size.

-b 16384
Click to expand...

I assume that this is a destructive type of test that should only be used on new drives, or blank drives, or where you don't care about the data any longer?
 
I assume that this is a destructive type of test that should only be used on new drives, or blank drives, or where you don't care about the data any longer?
Click to expand...

Exactly. If you care about the data you can do a non destructive test.

Note: A non destructive test will be slower since badblocks has to put the data back that it overwrites.
 
So I'm on windows and have been doing the 'full format' when I get new hard drives.

Should I use linux to do the badblocks instead? Would it be ok to run it off a live CD or I could install to a USB and unplug my Windows OS drive
 
Full format in windows is fine and quicker but I do not believe as thorough however it may not matter.

Would it be ok to run it off a live CD or I could install to a USB and unplug my Windows OS drive
Click to expand...

Yes running badblocks from a livecd/ usbstick is fine as well.


In my case I have lots of linux machines at work and at home so its not that big of a deal to hotplug a drive (or several) into a system and let it run badblocks over a weekend..

BTW, this is how I detected that there was a problem with my samsung f4s back in Q4 2010. Although I admit I was not sure what the cause was at that time..
 
Last edited:
You must log in or register to reply here.
Back
Top