advanced partition format, mdadm, and raw disks

Everyone,

I've created a RAID-10 array with mdadm. I created the array using
partitions /dev/sd[abcd]1, where the partition starts at sector 2048 and
uses the rest of the disk, as specified here:

http://www.ibm.com/developerworks/linux/library/l-4kb-sector -disks/index.html#tools

IE:
1. fdisk -c -u /dev/sda
2. create partition at 2048 to -0, of type fd
3. repeat for all drives or clone with sfdisk
4. create the array

mdadm -A /dev/md0 -v --raid-devices=4
--level=raid10 /dev/sda1 /dev/sdb1 /dev/sdc1 /dev/sdd1

5. add to mdadm.conf

mdadm -Es >> /etc/mdadm/mdadm.conf

After reboot, the array worked, and I formatted it and started loading
it with files. Only later did I realise that at some point (perhaps 2-3
reboots), the partitions were gone, and the array is running on the raw
devices:

root@venkman:/var/log/cacti# cat /proc/mdstat
Personalities : [linear] [multipath] [raid0] [raid1] [raid6] [raid5]
[raid4] [raid10]
md0 : active raid10 sdc[2] sda[0] sdd[3] sdb[1]
3907026944 blocks 64K chunks 2 near-copies [4/4] [UUUU]

unused devices:

fdisk shows the partition tables are gone too. I found NeilBrown's
comment ID #1857419, partway down this page:

http://www.issociate.de/board/post/463176/Superblocks.html

"If a partition starts a multiple of 64K from the start of the device,
and ends with about 64K of the end of the device, then a superblock on
the partition will also look like a superblock on the whole device.
This is one of the shortcomings of v0.90 superblocks. v1.0 doesn't
have this problem."

I used mdadm v2.6.7.1, as it comes on Ubuntu 10.10. Is this still a
shortcoming of superblock v1.2? Obviously I want to align to the sector
size, and use the entire disk.

Should I just accept this and not raw devices? Will it happen again if I
try to start over?

Regards,
Tyler

--
"If one would give me six lines written by the hand of the most honest
man, I would find something in them to have him hanged."
-- Armand Jean du Plessis, Cardinal Richelieu

--
To unsubscribe from this list: send the line "unsubscribe linux-raid" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html

Re: advanced partition format, mdadm, and raw disks

On Mon, 04 Jul 2011 17:15:33 +0100 "Tyler J. Wagner"
wrote:

> Everyone,
>
> I've created a RAID-10 array with mdadm. I created the array using
> partitions /dev/sd[abcd]1, where the partition starts at sector 2048 and
> uses the rest of the disk, as specified here:
>
> http://www.ibm.com/developerworks/linux/library/l-4kb-sector -disks/index.html#tools
>
> IE:
> 1. fdisk -c -u /dev/sda
> 2. create partition at 2048 to -0, of type fd
> 3. repeat for all drives or clone with sfdisk
> 4. create the array
>
> mdadm -A /dev/md0 -v --raid-devices=4
> --level=raid10 /dev/sda1 /dev/sdb1 /dev/sdc1 /dev/sdd1
>
> 5. add to mdadm.conf
>
> mdadm -Es >> /etc/mdadm/mdadm.conf
>
> After reboot, the array worked, and I formatted it and started loading
> it with files. Only later did I realise that at some point (perhaps 2-3
> reboots), the partitions were gone, and the array is running on the raw
> devices:

It must has stopped using partitions before you ran 'mkfs' which then
destroyed the partition tables. Probably on the first reboot.

>
> root@venkman:/var/log/cacti# cat /proc/mdstat
> Personalities : [linear] [multipath] [raid0] [raid1] [raid6] [raid5]
> [raid4] [raid10]
> md0 : active raid10 sdc[2] sda[0] sdd[3] sdb[1]
> 3907026944 blocks 64K chunks 2 near-copies [4/4] [UUUU]

No "super" is mentioned here so it must be using 0.90 metadata.


>
> unused devices:
>
> fdisk shows the partition tables are gone too. I found NeilBrown's
> comment ID #1857419, partway down this page:
>
> http://www.issociate.de/board/post/463176/Superblocks.html
>
> "If a partition starts a multiple of 64K from the start of the device,
> and ends with about 64K of the end of the device, then a superblock on
> the partition will also look like a superblock on the whole device.
> This is one of the shortcomings of v0.90 superblocks. v1.0 doesn't
> have this problem."
>
> I used mdadm v2.6.7.1, as it comes on Ubuntu 10.10. Is this still a
> shortcoming of superblock v1.2? Obviously I want to align to the sector
> size, and use the entire disk.
>

mdadm defaults to 0.90 until 3.0 when it starts defaulting to 1.2.


> Should I just accept this and not raw devices? Will it happen again if I
> try to start over?

I'm not really sure what "this" and "it" refer to. If it is working then
there is no immediate need to change anything.
If you want to make a change, be sure to back up your files.
If you create an array again, probably use "--metadata=1.2".

NeilBrown



>
> Regards,
> Tyler
>

--
To unsubscribe from this list: send the line "unsubscribe linux-raid" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html

Re: advanced partition format, mdadm, and raw disks

On Tue, 2011-07-05 at 10:30 +1000, NeilBrown wrote:
> No "super" is mentioned here so it must be using 0.90 metadata.

Ah. Damn that old Debian package.

I'm concerned that this means the filesystem isn't aligned to the 4K
sector, and is therefore less efficient. Is 0.90 metadata at the
beginning or end of the disk? Is it a multiple of 4K in size?

Regards,
Tyler

--
"I respect you too much to respect your ridiculous ideas."
-- Johann Hari

--
To unsubscribe from this list: send the line "unsubscribe linux-raid" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html

Re: advanced partition format, mdadm, and raw disks

On Tue, 05 Jul 2011 05:28:50 +0100 "Tyler J. Wagner"
wrote:

> On Tue, 2011-07-05 at 10:30 +1000, NeilBrown wrote:
> > No "super" is mentioned here so it must be using 0.90 metadata.
>
> Ah. Damn that old Debian package.
>
> I'm concerned that this means the filesystem isn't aligned to the 4K
> sector, and is therefore less efficient. Is 0.90 metadata at the
> beginning or end of the disk? Is it a multiple of 4K in size?
>

0.90 has data at the start and metadata at the end on a 64k aligned boundary.
So if the partition is 4K aligned, everything else will be too.

NeilBrown

--
To unsubscribe from this list: send the line "unsubscribe linux-raid" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html

Re: advanced partition format, mdadm, and raw disks

On Tue, 2011-07-05 at 14:56 +1000, NeilBrown wrote:
> 0.90 has data at the start and metadata at the end on a 64k aligned boundary.
> So if the partition is 4K aligned, everything else will be too.

OK, that's clear. Thanks very much for your help. It's a rare project
that the main developer helps users so directly. I appreciate it.

I'm rebuilding the array using aligned partitions and 1.2 metadata
anyway. It'll be easier to maintain, especially if I later have to
replace a drive.

Regards,
Tyler

--
"It is an interesting and demonstrable fact, that all children are atheists
and were religion not inculcated into their minds, they would remain so."
-- Ernestine Rose

--
To unsubscribe from this list: send the line "unsubscribe linux-raid" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html