| View previous topic :: View next topic |
| Author |
Message |
frostschutz
Advocate
Joined: 22 Feb 2005
Posts: 2977
Location: Germany
|
 Posted: Sun Sep 13, 2015 11:48 pm Post subject: |
 |
|
--examine is not on /dev/md?? but /dev/sd??
But 0.90 metadata does not have an offset, either way. Metadata is at the end of the device. |
|
| Back to top |
|
 |
maiku
Guru
Joined: 24 Mar 2004
Posts: 581
Location: Escaping from NY
|
| Posted: Mon Sep 14, 2015 12:23 am Post subject: |
|
|
That makes sense. I can do all of that math. What do you think of the stiped RAID because of the RAID10? The bad sectors also belonged to two striped drives.
_________________
Michael |
|
| Back to top |
|
|
NeddySeagoon
Administrator
Joined: 05 Jul 2003
Posts: 54214
Location: 56N 3W
|
| Posted: Mon Sep 14, 2015 10:30 pm Post subject: |
|
|
maiku,
If you have correct images of sda and sdd, which is all your raid was running on before, you can restore them to the new drives and they will work like nothing has happened.
You can then try to rebuild the raid. I suspect that the other two drives are in a bad way too but you can try.
Have a look at their smart data. We aready know that sd[ad] had a lot of pending sectors.
You might even find that ddrescue has forced sector relocation, so you can rebuld the array from the old drives but there is little point as you really don't want to be using drives that can't read their own writing.
_________________
Regards,
NeddySeagoon
Computer users fall into two groups:-
those that do backups
those that have never had a hard drive fail. |
|
| Back to top |
|
|
maiku
Guru
Joined: 24 Mar 2004
Posts: 581
Location: Escaping from NY
|
| Posted: Thu Sep 17, 2015 8:18 pm Post subject: |
|
|
Thank you, everybody, for your help. The saga ends finally. I was able to get most of the data off of the arrays then I rebuild it on new drives. It is back up | Quote: | md2 : active raid10 sdd1[3] sdc1[1] sdb1[0] sda1[2]
3907023872 blocks 64K chunks 2 near-copies [4/4] [UUUU] |
And it's healthy!
I decided to give up on finding out exactly what files were corrupt. I tried to do some math to see if I could find out, but in the blocks I was looking at I didn't care much about that data because it could always be regenerated. I mostly have big drives to store junk. A very small portion of it is actually valuable (or invaluable).
Thanks again!
_________________
Michael |
|
| Back to top |
|
|
NeddySeagoon
Administrator
Joined: 05 Jul 2003
Posts: 54214
Location: 56N 3W
|
| Posted: Thu Sep 17, 2015 10:38 pm Post subject: |
|
|
maiku,
Time to run | Code: | | echo repair > /sys/block/md2/md/sync_action |
Check the smart data before and after on all the drives in md2.
I'll let you read up on the details. Essentially, it reads all the blocks on all partitions in the raid set and checks the that the redundant data is correct.
If it gets a read error it recalculates the redundant data and writes it to the failed sector, which the drive is expected to relocate.
You will see this in the smart data and in dmesg.
Do all this monthly in a cron job on all your raid sets to keep them healthy.
It picks op on failed sectors you didn't even know you had.
_________________
Regards,
NeddySeagoon
Computer users fall into two groups:-
those that do backups
those that have never had a hard drive fail. |
|
| Back to top |
|
|
|
Kernel & Hardware
