SmallNetBuilder

RAID 1 Fail Test

Some commenters on the 323's review complained about problems with RAID 1 recovery. So this time, I put a RAID 1 fail test on the review to-do list and I'm glad that I did because I found problems.

I first enabled and successfully tested the email alert feature (Figure 6) so that I could see if the 321 sent me any love notes. I then started a large folder copy to the 321, waited for it to get underway, then pulled the left-hand drive.

Figure 6: Email alert screen

It took about 30 seconds for the 321 to change the drive LED from off to red to flag the failed drive. I also got this email around the same time:

Hello Administrator,
Left Hard Drive Has Failed
Sincerely,
Your DNS-321

The Status screen also changed to show degraded status (Figure 7)

Figure 7: Status with failed drive

So I shut down the 321 (since hot-swap isn't supported), plugged the "failed" drive back in and rebooted. During boot, I noticed the "failed" drive light initially blinking a lot and then turning a solid purple. It also took much longer than usual—2-3 minutes—for the login screen to come back.

When I logged in, there was no warning of drive failure and I found that the Status screen still showed as degraded. So I navigated to the Tools > RAID screen to start the rebuild, but found only the option shown in Figure 8.

Figure 8: No rebuild option

I refreshed the RAID and Status screens a few times and even tried Internet Explorer 6 (I was using Firefox 3). But I could not get the rebuild option that the User Manual told me that I should see. I even rebooted the NAS a few times, but got the same result each time. Not a good sign.

After I found no clues in the User Manual, I hit the D-Link support site, searched the FAQ and found a clue in this FAQ, i.e.

Note: The new drive Should be blank with no partition on the drive The drive may still be detected and prompt to format if formatted in NTFS

Fortunately, D-Link had sent four drives for the upcoming review of the DNS-343. So I powered down the 321, popped in a fresh drive, restarted and was rewarded with the screen in Figure 9 upon login.

Figure 9: Replacement drive rebuild

I clicked the Skip button and navigated to the Tools > RAID screen to be sure that the rebuild option was there and found the screen in Figure 10. I should note that a complete resync took only about 15 minutes for a 78 GB RAID 1 array. Also note that I received no emails telling me that resync had either started or was completed.

Figure 10: Replacement drive rebuild, RAID screen

After the rebuild was complete, I decided to fail the right-hand drive. After I pulled the drive, the Status screen wasn't updated and showed no sign of the failure. There was also no indication of failure after I logged out and back in. It wasn't until I rebooted the 321 that the Sync Time Remaining line in the Status page changed to "Degraded".

I then shut down the 321, reinserted the "failed" drive and powered back up. As the 321 booted, both drive lights blinked in tandem and it again took 2-3 minutes to get the login screen. But this time when I logged in and checked the status page, I found the 321 in the process of rebuilding the RAID 1 array on both the Status page and when I clicked over to the Tools > RAID page (Figure 11).

Figure 11: RAID screen during rebuild

Note that I could use the NAS while the resync was running, but I found the resync time started growing significantly as I started a large folder transfer.

While this test doesn't simulate what happens in an actual failure (since you wouldn't be reinserting the same drive), we run the same test on all RAID products. And the 321 is the only product we've encountered that requires a fresh drive in order to correctly resync the RAID 1 array (at least for one of the drives!). I'll leave it up to you to decide whether this is a reasonable requirement.

My final two notes from this exercise are that there is no format option and no checkdisk function. Drives are formatted using the EXT2 filesystem.