Re: block level vs. file level

[Andy Smith <andy@lug.org.uk>]

[-- Attachment #1: Type: text/plain, Size: 1505 bytes --]

On Sun, Feb 12, 2006 at 11:31:07AM -0800, it wrote:
> The hardware raid does the mirroring on the block level, so it's 
> actually /dev/sda mirroring /dev/sdb - the whole drive, and not 
> partitions. There is a way to set this up on software raid. It takes 
> more configuration tweaking, but the mirroring then includes the 
> partition table as well. This way, if a drive fails, one can replace it 
> without pre-partitioning it.

That can be less flexible though.  If I have say 4 drives then I
quite often want small /boot, / and swap under RAID-1 then the rest
as a single large partition in RAID-5, -6 or -10 as an LVM PV.

> This also raises another point, which is relevant for both cases - same 
> exact models of hard disks have different number of cylinders, so if a 
> RAID partition is created on a larger drive it cannot be mirrored to a 
> smaller drive.

Same exact models don't usually have different block counts, but
certainly if you replace a dead drive with a different one of the
same advertised capacity you can end up getting one slightly
smaller.

> Does anyone have any experience with this?

Yes, and it's a pain, but if you have to deal with it I think the
wealth of options in md leaves you better able to handle it than
with hardware RAID.  Here's something that happened to me:

http://strugglers.net/wiki/becks.strugglers.net

-- 
http://strugglers.net/wiki/Xen_hosting -- A Xen VPS hosting hobby
Encrypted mail welcome - keyid 0x604DE5DB

[-- Attachment #2: Digital signature --]
[-- Type: application/pgp-signature, Size: 189 bytes --]