Bug with RAID1 hot spares?

Bug with RAID1 hot spares?

[Chase Venters]

Greetings,
	I was just testing a server I was about to send into production on kernel 
2.6.18.1. The server has three SCSI disks with "md1" set to a RAID1 with 2 
mirrors and 1 spare. The mirrors are sda3 and sdb3, spare is sdc3. I manually 
failed sdb3, and as expected, sdc3 was activated. Strangely 
enough, /proc/mdstat did not indicate that sdc3 was being synced. I thought 
these spares weren't kept mirrored until needed?
	In order to further test my theory, I manually failed sda3, leaving only sdc3 
(the original spare) active. I ran "find /" for a bit to see if any errors 
cropped up and none did; however, when I added sda3 and sdb3 back to the 
array and a resync started, I was soon faced with what appeared to be a 
_very_ corrupted reiserfs.
	Strangely enough, after booting on a livecd and assembling md1 with just 
sda3, I was able to add sdb3 and sdc3, after which the array resynced and 
left sdb3 a mirror and sdc3 a spare.
	So there's definitely something odd happening here... why did no resync to 
the sdc3 spare start when I failed sdb3?

Thanks,
Chase

Re: Bug with RAID1 hot spares?

[Neil Brown]

On Friday October 20, chase.venters@clientec.com wrote:
> Greetings,
> 	I was just testing a server I was about to send into production on kernel 
> 2.6.18.1. The server has three SCSI disks with "md1" set to a RAID1 with 2 
> mirrors and 1 spare. The mirrors are sda3 and sdb3, spare is sdc3. I manually 
> failed sdb3, and as expected, sdc3 was activated. Strangely 
> enough, /proc/mdstat did not indicate that sdc3 was being synced. I thought 
> these spares weren't kept mirrored until needed?

Correct.  They are not kept mirrored.

> 	So there's definitely something odd happening here... why did no resync to 
> the sdc3 spare start when I failed sdb3?

yes... can you check that this fixes it please?
Thanks,
NeilBrown

Signed-off-by: Neil Brown <neilb@suse.de>

### Diffstat output
 ./drivers/md/md.c |    1 +
 1 file changed, 1 insertion(+)

diff .prev/drivers/md/md.c ./drivers/md/md.c
--- .prev/drivers/md/md.c	2006-10-23 12:09:08.000000000 +1000
+++ ./drivers/md/md.c	2006-10-23 14:10:46.000000000 +1000
@@ -2003,6 +2003,7 @@ static mdk_rdev_t *md_import_device(dev_
 	kobject_init(&rdev->kobj);
 
 	rdev->desc_nr = -1;
+	rdev->saved_raid_disk = -1;
 	rdev->flags = 0;
 	rdev->data_offset = 0;
 	rdev->sb_events = 0;

Re: Bug with RAID1 hot spares?

[Bill Davidsen]

Chase Venters wrote:

>Greetings,
>	I was just testing a server I was about to send into production on kernel 
>2.6.18.1. The server has three SCSI disks with "md1" set to a RAID1 with 2 
>mirrors and 1 spare.
>
I have to ask, why? If the array is mostly written you might save a bit 
of bus time, but for reads having another copy of the data to read 
(usually) helps the performance by reducing wait for read occurences.

-- 
bill davidsen <davidsen@tmr.com>
  CTO TMR Associates, Inc
  Doing interesting things with small computers since 1979

Re: Bug with RAID1 hot spares?

[Chase Venters]

On Tuesday 24 October 2006 20:41, Bill Davidsen wrote:
> Chase Venters wrote:
> >Greetings,
> >	I was just testing a server I was about to send into production on kernel
> >2.6.18.1. The server has three SCSI disks with "md1" set to a RAID1 with 2
> >mirrors and 1 spare.
>
> I have to ask, why? If the array is mostly written you might save a bit
> of bus time, but for reads having another copy of the data to read
> (usually) helps the performance by reducing wait for read occurences.

The main idea is to not exercise the spare as much as the other disks. All 
three disks are from the same lot. Having three disks fail at once is 
admittedly unlikely, but keeping one disk as a spare rather than full mirror 
should probably reduce the wear on that disk so if there is some 
manufacturing defect the third drive wouldn't be as close to failing and 
could hopefully keep the box online until someone makes it to the datacenter 
to do a swap.

Thanks,
Chase

Re: Bug with RAID1 hot spares?

[Mario 'BitKoenig' Holbe]

Chase Venters <chase.venters@clientec.com> wrote:
> The main idea is to not exercise the spare as much as the other disks. All 

Btw. you can also keep the spare-disk spinned down most of the time.
You should probably just make sure to spin it up from time to time to
see if it's still okay - I spin up my spares one hour per night when
smartd issues short selftests and a few more hours when smartd issues
long selftests.


regards
   Mario
-- 
<jv> Oh well, config
<jv> one actually wonders what force in the universe is holding it
<jv> and makes it working
<Beeth> chances and accidents :)