GPT on MD raid

GPT on MD raid

[Jeremy Sanders]

Hi - I was thinking of using a GPT partition on top of a MD raid device. It 
occurs to me that because GPT also puts data at the end of the device, it 
might get confused if the raid device is resized by adding more drives, etc.

Has anyone tried this? Should it work? How should the GPT header get moved 
to the end of the raid device? Will parted fix it?

Should I ditch this idea and delve into the complexities of LVM?

Thanks

Jeremy

Re: GPT on MD raid

[Phil Turmel]

On 04/18/2011 12:11 PM, Jeremy Sanders wrote:
> Hi - I was thinking of using a GPT partition on top of a MD raid device. It 
> occurs to me that because GPT also puts data at the end of the device, it 
> might get confused if the raid device is resized by adding more drives, etc.
> 
> Has anyone tried this? Should it work? How should the GPT header get moved 
> to the end of the raid device? Will parted fix it?

I've parted fix cases of this in the past, but I don't know how robust it is.

> Should I ditch this idea and delve into the complexities of LVM?

Yes, in my not-entirely-humble opinion.  I suspect you'll find LVM isn't all that complex, at least if all you are trying to do is replace GPT.  And you get some serious advantages without extra effort, like non-contiguous volumes and snapshotting.

While my copy of gparted won't work with it, there are gui tools to manage LVM.  I use kvpm on my gentoo/KDE laptop.

> 
> Thanks

HTH,

Phil

Re: GPT on MD raid

[Jeremy Sanders]

On 18/04/11 17:47, Phil Turmel wrote:
> On 04/18/2011 12:11 PM, Jeremy Sanders wrote:
>> Hi - I was thinking of using a GPT partition on top of a MD raid device. It
>> occurs to me that because GPT also puts data at the end of the device, it
>> might get confused if the raid device is resized by adding more drives, etc.
>>
>> Has anyone tried this? Should it work? How should the GPT header get moved
>> to the end of the raid device? Will parted fix it?
>
> I've parted fix cases of this in the past, but I don't know how robust it is.
>
>> Should I ditch this idea and delve into the complexities of LVM?
>
> Yes, in my not-entirely-humble opinion.  I suspect you'll find LVM isn't all that complex, at least if all you are trying to do is replace GPT.  And you get some serious advantages without extra effort, like non-contiguous volumes and snapshotting.
>
> While my copy of gparted won't work with it, there are gui tools to manage LVM.  I use kvpm on my gentoo/KDE laptop.

Thanks for all the replies - I will look at lvm...

Jeremy

Re: GPT on MD raid

[Mikael Abrahamsson]

On Mon, 18 Apr 2011, Jeremy Sanders wrote:

> Should I ditch this idea and delve into the complexities of LVM?

Yes. LVM isn't all that complex and a lot more flexible, plus you can do 
"everything" online without having to unmount filesystems to re-partition.

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

Re: GPT on MD raid

[Roman Mamedov]

[-- Attachment #1: Type: text/plain, Size: 923 bytes --]

On Mon, 18 Apr 2011 17:11:56 +0100
Jeremy Sanders <jeremy@jeremysanders.net> wrote:

> Hi - I was thinking of using a GPT partition on top of a MD raid device. It 
> occurs to me that because GPT also puts data at the end of the device, it 
> might get confused if the raid device is resized by adding more drives, etc.
> 
> Has anyone tried this? Should it work? How should the GPT header get moved 
> to the end of the raid device? Will parted fix it?
> 
> Should I ditch this idea and delve into the complexities of LVM?

I'd suggest that you reconsider if you really need to partition a single
mdraid device, or maybe it would be better to create several mdraid devices
(on regular partitions) for different needs, possibly with differing shapes
and settings. E.g. over the same set of disks I use three RAIDs, two small
arrays for boot and root FS, and the rest for data.

-- 
With respect,
Roman

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 198 bytes --]