Testing degraded RAID 1 device without downtime

[odioworks_com]

Two of my server's partitions (md1 & md3) have degraded arrays. Since both of the degraded arrays are on the same drive (/dev/sdb), I'd like to test the drive for an actual physical failure before I add it back into the array.

Does anyone know of any good drive-level testing software that can be run without taking the whole machine offline? Preferably, I'd just like to take that device entirely out of the array using "mdadm /dev/mdX -r /dev/sdbX," and test it while the rest of the array is running.

Alternatively, I could take the server down for 5 mins to remove the defective drive, and then test it in another machine. I'm assuming the number next to the drive in /proc/mdstat is the physical drive connection? For example, [0] is SATA0 & [1] is SATA1? I'm paranoid about mistakenly removing the good drive and making things worse.

Code: Select all

md1 : active raid1 sda1[0]
      40064 blocks [2/1] [U_]
      
md2 : active raid1 sdb2[1] sda2[0]
      2008000 blocks [2/2] [UU]
      
md3 : active raid1 sda3[0]
      154240000 blocks [2/1] [U_]

Oh, and is there anything else I need to know before shutting down the machine and pulling out the drive? I would just die if I took this machine down to troubleshoot the drive and couldn't get it back up...

TIA!

[odioworks_com]

Well, I pulled out the drive on SATA 1 - and I guess I picked correctly!

So it sounds like the [0] and [1] refer to the drive connection letters in /proc/mdstat. They also seem to correspond to the RaidDevice number in mdadm --detail /dev/mdX

[HeissFuss]

hdparm -I <device> can tell you the disk's serial number so that you can physically check that you pull the correct drive. Usually badblocks is sufficient to tell if a drive is defective. You'll probably get a bunch of io errors and bad block numbers if the drive is bad.