Mirror recovery?

Mirror recovery?

[Ross Vandegrift]

Hello all,

	Lost a drive in my box last night - woke up to it locked
completely solid.  So I reset the machine, it started to come back up
and remirror the drives.  I set about trying to figure out why it locked
(the terminal had blanked overnight of course, and wouldn't unblank so
I missed the probable standard failure messages).  Then the box locked
hard again, right as I heard one of my IDE drives make some terrible
noises.

	So I rebooted, and ran the manufacturer's disk test, sure enough
it's headed south.  Once I've got a replacement drive, will I be able to
get the data on my raid partition back?  I don't think I've ever had a
failure happen on a RAID1 array during resync.  Once on RAID5, and that
was a disaster....

Ross Vandegrift
ross@willow.seitz.com

Re: Mirror recovery?

[Neil Brown]

On Thursday June 20, ross@willow.seitz.com wrote:
> Hello all,
> 
> 	Lost a drive in my box last night - woke up to it locked
> completely solid.  So I reset the machine, it started to come back up
> and remirror the drives.  I set about trying to figure out why it locked
> (the terminal had blanked overnight of course, and wouldn't unblank so
> I missed the probable standard failure messages).  Then the box locked
> hard again, right as I heard one of my IDE drives make some terrible
> noises.

Don't you hate that noise...

> 
> 	So I rebooted, and ran the manufacturer's disk test, sure enough
> it's headed south.  Once I've got a replacement drive, will I be able to
> get the data on my raid partition back?  I don't think I've ever had a
> failure happen on a RAID1 array during resync.  Once on RAID5, and that
> was a disaster....

You shouldn't have any problems getting your data back.  Indeed you
should be able to get it even before the new drives arrive, that is
what RAID is about.

My (biased) guess is that the machine locked up because the IDE driver
was not coping elegantly with the drive failure. 
I think it would be nice if the raid system could tell the device "if
you get an error, don't retry too hard, I'm happy to cope".

NeilBrown

Re: Mirror recovery?

[Ross Vandegrift]

> > 	So I rebooted, and ran the manufacturer's disk test, sure enough
> > it's headed south.  Once I've got a replacement drive, will I be able to
> > get the data on my raid partition back?  I don't think I've ever had a
> > failure happen on a RAID1 array during resync.  Once on RAID5, and that
> > was a disaster....
> 
> You shouldn't have any problems getting your data back.  Indeed you
> should be able to get it even before the new drives arrive, that is
> what RAID is about.

Ok, I wasn't sure if the second failure during reconstruction was
disastrous for RAID1.  Looks like it's not - my boss lent me a disk to
use until my RMA'ed Deathstar comes in, and as soon as I hot added a new
partition, reconstruction started with all my data.

Somedays when the simple things go right it's enough to make your day
::-)

> My (biased) guess is that the machine locked up because the IDE driver
> was not coping elegantly with the drive failure. 
> I think it would be nice if the raid system could tell the device "if
> you get an error, don't retry too hard, I'm happy to cope".

Hardly a biased opinion.  I've seen more than my share of IDE disk
failures at work - it's always the IDE driver that falls over first.  

Ross Vandegrift
ross@willow.seitz.com