[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Mirroring a failing HDD



Douglas Tutty wrote:
On Fri, Nov 10, 2006 at 12:38:43PM +0000, Shri Shrikumar wrote:
There is a server(sarge) that I maintain that used to be mirrored and
all was well. However, the mirror was recently broken and when trying to
rebuild, I run into an interesting problem.

The array rebuilds to about 80% and then restarts. dmesg has the following:

Hi Doug,

Thanks for you response.

You had a raid consisting of two drives, one of which is /dev/sdb3.  One
died, leaving sdb3.  However, during rebuild /dev/sdb is having read
errors.

Not quite. The other drive did not die. I upgraded the kernel and on re-boot, the other drive was mis-read and not re-linked as part of the raid. It was not a hard disk failure that caused
the mirror to be broken.

Was the origional failed drive on the same controller?  Could that
drive failure also have killed the controller?  Could it have been a
controller failure and not a drive failure?  Do you have another
controller in the box to which you could connect the drive that is now
hdb?  Are there other drives on this controller that are working OK?

The problem is that its a live server, so can't fiddle around with it much. There is another disk on it which seems to be running just fine. I am fairly confident that it is the disk that is failing and not
the controller.

Can you read other partitions on hdb? (and therefore prove that both the
controller and drive are OK)

Yes, there is another partition that is mirrored just fine with sda.

I hope you have good backups.

Thankfully, I do. It continues to run fine as well due to the corruption being outside of the disk being
used.

How does the system work if you disconnect the new drive (sda) so that the
raid runs in degradded mode?  Do you still get read errors?  If you

The raid is currently in degraded mode with just sdb. There are no read errors. I am using LVM on raid to allocate disk space and there is about 60gb unallocated. I am guessing that the failed part of the disk is
in this unallocated area.

I suppose the worst case would be to build a brand new mirror using just the new disk and pvmove the data across and then de-commission the dying hdd. I am hoping for an easier (and quicker solution though)

Best Wishes,


Shri

--
Shri Shrikumar
Technologist Extraordinaire
Kraya

t: 0845 644 4745
d: 0131 247 8021
f: 0131 478 7377
w: www.kraya.co.uk



Reply to: