question about RAID10 near and far layouts

question about RAID10 near and far layouts

[plug bert]

hi peeps,

i've been reading through

http://en.wikipedia.org/wiki/Non-standard_RAID_levels

and just wanted to verify if my understanding is correct.

Is "near" safer than "far"?

e.g. given 4 drives in RAID10 array, n2:

4 drives

1 2 3 4
--------------
A1 A1 A2 A2
A3 A3 A4 A4
A5 A5 A6 A6
A7 A7 A8 A8

you'd lose the array if either 1&2 or 3&4 goes down at the same time.


With 4 drives in RAID10 array, f2:

4 drives
1 2 3 4
--------------------
A1 A2 A3 A4
A5 A6 A7 A8
A9 A10 A11 A12
.. .. .. ..
A4 A1 A2 A3
A8 A5 A6 A7
A12 A9 A10 A11

...there seems to be a lot more combinations that can result in a trashed array(1&2, 2&3, 3&4).

Is my analysis correct? Inputs are more than welcome, tia

Re: question about RAID10 near and far layouts

[Phil Turmel]

On 06/10/2012 09:26 PM, plug bert wrote:
> hi peeps,
> 
> i've been reading through
> 
> http://en.wikipedia.org/wiki/Non-standard_RAID_levels
> 
> and just wanted to verify if my understanding is correct.
> 
> Is "near" safer than "far"?
> 
> e.g. given 4 drives in RAID10 array, n2:
> 
> 4 drives
> 
> 1 2 3 4
> --------------
> A1 A1 A2 A2
> A3 A3 A4 A4
> A5 A5 A6 A6
> A7 A7 A8 A8
> 
> you'd lose the array if either 1&2 or 3&4 goes down at the same time.
> 
> 
> With 4 drives in RAID10 array, f2:
> 
> 4 drives
> 1 2 3 4
> --------------------
> A1 A2 A3 A4
> A5 A6 A7 A8
> A9 A10 A11 A12
> .. .. .. ..
> A4 A1 A2 A3
> A8 A5 A6 A7
> A12 A9 A10 A11
> 
> ...there seems to be a lot more combinations that can result in a trashed array(1&2, 2&3, 3&4).
> 
> Is my analysis correct? Inputs are more than welcome, tia

I think wikipedia might have this layout wrong.  I was under the
impression that a four-disk far2 layout would be:

> 1 2 3 4
> --------------------
> A1 A2 A3 A4
> A5 A6 A7 A8
> A9 A10 A11 A12
> .. .. .. ..
> A2 A1 A4 A3
> A6 A5 A8 A7
> A10 A9 A12 A11

I haven't checked the code, though.

Phil

metadata versions: 0.90 vs 1.2

[plug bert]

Hello,

i noticed this while twiddling with RAID1 arrays...

On our old CentOS 4.7 box, i can mount a RAID1 array member independently of the array -- i.e. mount /dev/sdc1 /mountpoint....but on CentOS 5.x and later i am met with an unknown filesystem: linux_raid_member error. 

Somebody hinted that this had something to do with the metadata versions, so i recreated the RAID1 array on CentOS 5.x with the --metadata=0.90 parameter...and was able to mount the array member w/o any problems.


Is this expected behavior? Are there any potential problems if i stick with metadata=0.90(apart from the 28 device and 2Tb disk space limit)?


tia

 

Re: metadata versions: 0.90 vs 1.2

[Mikael Abrahamsson]

On Wed, 13 Jun 2012, plug bert wrote:

> Is this expected behavior? Are there any potential problems if i stick 
> with metadata=0.90(apart from the 28 device and 2Tb disk space limit)?

From "man mdadm":

               1, 1.0, 1.1, 1.2 default
                      Use  the  new  version-1  format superblock.  This has few restrictions.  The different sub-versions store the superblock at
                      different locations on the device, either at the end (for 1.0), at the start (for 1.1) or 4K from the start (for 1.2).   "1"
                      is equivalent to "1.0".  "default" is equivalent to "1.2".

If you want to be able to mount the individual RAID1 members but still use 
the newer version superblock, you need to use version 1.0 (stored at the 
end of the device/partition).

I'd recommend to go with 1.0 instead of 0.90.

-- 
Mikael Abrahamsson    email: swmike@swm.pp.se

Re: metadata versions: 0.90 vs 1.2

[Emmanuel Noobadmin]

On 6/13/12, Mikael Abrahamsson <swmike@swm.pp.se> wrote:
> If you want to be able to mount the individual RAID1 members but still use
> the newer version superblock, you need to use version 1.0 (stored at the
> end of the device/partition).
>
> I'd recommend to go with 1.0 instead of 0.90.

Would metadata 1.0 boot with older grub? Since that's often why we're using 0.9.

Re: metadata versions: 0.90 vs 1.2

[Phil Turmel]

On 06/13/2012 10:46 AM, Emmanuel Noobadmin wrote:
> On 6/13/12, Mikael Abrahamsson <swmike@swm.pp.se> wrote:
>> If you want to be able to mount the individual RAID1 members but still use
>> the newer version superblock, you need to use version 1.0 (stored at the
>> end of the device/partition).
>>
>> I'd recommend to go with 1.0 instead of 0.90.
> 
> Would metadata 1.0 boot with older grub? Since that's often why we're using 0.9.

It does.  It's what I use myself.

Phil

Re: metadata versions: 0.90 vs 1.2

[Emmanuel Noobadmin]

On 6/13/12, Phil Turmel <philip@turmel.org> wrote:
> On 06/13/2012 10:46 AM, Emmanuel Noobadmin wrote:
>> On 6/13/12, Mikael Abrahamsson <swmike@swm.pp.se> wrote:
>>> If you want to be able to mount the individual RAID1 members but still
>>> use
>>> the newer version superblock, you need to use version 1.0 (stored at the
>>> end of the device/partition).
>>>
>>> I'd recommend to go with 1.0 instead of 0.90.
>>
>> Would metadata 1.0 boot with older grub? Since that's often why we're
>> using 0.9.
>
> It does.  It's what I use myself.

Thank you very much for that confirmation :)

Re: metadata versions: 0.90 vs 1.2

[David Brown]

On 13/06/2012 06:55, plug bert wrote:
> Hello,
>
> i noticed this while twiddling with RAID1 arrays...
>
> On our old CentOS 4.7 box, i can mount a RAID1 array member
> independently of the array -- i.e. mount /dev/sdc1 /mountpoint....but
> on CentOS 5.x and later i am met with an unknown filesystem:
> linux_raid_member error.
>
> Somebody hinted that this had something to do with the metadata
> versions, so i recreated the RAID1 array on CentOS 5.x with the
> --metadata=0.90 parameter...and was able to mount the array member
> w/o any problems.
>
>
> Is this expected behavior? Are there any potential problems if i
> stick with metadata=0.90(apart from the 28 device and 2Tb disk space
> limit)?
>

Yes, this is expected behaviour (as Mikael explained).

As for potential problems, the big one is if you mount a member of a 
raid1 array (with metadata 0.90, 1.0) directly, and write to it, you'll 
corrupt the raid1 array.  So make sure you only mount it read-only, 
unless you never want to see the raid1 again.

Typical uses of such mounts are for recovery purposes, or for accessing 
the raid from a bootloader (newer grub can understand more raid arrays, 
but raid1 with metadata 0.90 is often used with older grub).