mdadm Drive Failed... maybe?

[LiQuId StEeL]

So I finally finished up my NAS build, and after a good deal of headache caused by a terrible Intel-made motherboard, it was all running smooth.

My array was created by mdadm and then expanded into a 6x1.5tb Raid-6 and all was well. Today, however, I receieved this lovely status/error email:

This is an automatically generated mail message from mdadm
running on BROTHEL

A Fail event had been detected on md device /dev/md0.

It could be related to component device /dev/sde1.

Faithfully yours, etc.

P.S. The /proc/mdstat file currently contains the following:

Personalities : [linear] [multipath] [raid0] [raid1] [raid6] [raid5] [raid4] [raid10]
md0 : active raid6 sde1[6](F) sdb1[0] sdf1[4] sdc1[1] sdd1[2] sdg1[5]
5860543744 blocks level 6, 64k chunk, algorithm 2 [6/5] [UUU_UU]

unused devices: <none>

I checked with mdadm --detail /dev/md0 and it initially listed /dev/sde1 as "faulty spare", so I rebooted the server and the drive did not show up at all. No reference in fdisk -l

I've had terrible luck with these Seagate drives (3 RMA'd ones in the last year, but no loss of data), but I'm financially stuck with them for now.

I pulled the "dead" drive out of my NAS and tossed it in my Windows box and ran Seagate's Seatools. It passed SMART checks as well as the Short DST. I ran the Long DST from SeaTools, and that passed as well.


What I'm wondering, though, is what should I do? Seagate doesn't seem keen on RMA'ing a drive that doesn't fail its checks, so... what? Format on the Windows side, pop back in the NAS, format back to ext4 and re-add to mdadm?

Should I just strongarm Seagate into taking it back? Advice?

Thanks <3

 

[drescherjm]

Seagate doesn't seem keen on RMA'ing a drive that doesn't fail its checks, so...

They do for me. 4 of the 5 seagate 7200.11 drives I have RMA'd this year for work passed their smart tests. On the web RMA there is a choice like I do not have a code but I know the drive is bad use that and they will take your RMA.

 

[drescherjm]

I would however look at the smart raw data on the drive before sending it out. You should have access to smartctl. Check the Reallocated_Sector_Ct and Current_Pending_Sector. Having a few in Reallocated_Sector_Ct is not bad if it is not growing but having any in Current_Pending_Sector is bad.

 

[drescherjm]

format back to ext4 and re-add to mdadm?

No. You do not format drives before you add them to mdadm. The filesystem sits on top of mdadm. You may partition it for mdadm but never format ext4 then add to mdadm.

 

[LiQuId StEeL]

No. You do not format drives before you add them to mdadm. The filesystem sits on top of mdadm. You may partition it for mdadm but never format ext4 then add to mdadm.

I was under the impression that you need to set it to ext4, then set the RAID flag before mdadm -a /dev/sde1 ['ing].


I'll toss it back into the windows machine and pull the full SMART data tomorrow.


/bad luck

 

[drescherjm]

It sounds like you are using a GUI partition creator that may have this silly requirement..

Normally I create partitions with fdisk. With that you create a partition by specifying its start, size then setting its type to fd (raid auto detect). And that is all. No need to format any filesystem.

Edit: In addition to that a more advanced / smarter way to add a disk to a raid (when all partitions are the same on all disks) is

sfdisk -d /dev/sda | sfdisk /dev/sdb

That assumes /dev/sda is the disk with the correct disk layout and /dev/sdb is the new disk. Basically this creates the partition table on /dev/sdb to match the partition table on /dev/sda.

 

[LiQuId StEeL]

100 in Current Pending Sector


Looks like I'll be shooting this back to Seagate. Thanks for the advice <3

 

[drescherjm]

No that is 0 in the current pending. Look at the Data column.

As for the drive I would be a little concerned with 61 failed sectors. Its not really a problem if that remains constant but if it grows daily I would send the drive back. One way to test this is to run badblocks (or some other program that writes to and reads every sector) on the drive for a day then see if the number stays the same or greatly increases. The last drive I sent back to seagate added 20 to 200 bad sectors per day when in raid. In a period of two weeks it went from 14 reallocated sectors to over 2000. I pulled it out of the raid then and ran badlocks over a 4 day weekend and the drive disappeared from the bios sometime during the first day. After putting the drive in a windows box the drive was back but it was easy to get a failure code for the drive at that point. So that was the one we sent back with a code.