[sclug] RAID and/or LVM for ganging together HDDs to expand capacity?

[Tom Dawes-Gamble]

On Fri, 2007-10-19 at 01:16 +0100, John Stumbles wrote:
> 
> With LVM when one disk fails do you lose all the data on the entire set 
> of disks or just that which is physically on the failed disk?
> 

It depends. :-)

If you get a bad spot then it's just like getting a bad spot on a none
LVM file system.

If you loose the whole disk then it's like getting a whole hosts of bad
spots.

When the a disk fails you can replace it and use vgcfgrestore to put the
LVM meta back on the drive.  Then it would be up to fsck as to how much
you loose.

If you loose the disk with the Super Block on it then it's down to how
good you are at rescuing data from broken file systems.

If you limit yourself to only having all the extents of an LV on the
same PV then it's really no worse than using a none LVM configuration.


> I realise the point about MTBF. For that reason - as well as limitations 
> in how many discs I can run inside a box - I wouldn't plan to have more 
> than 2 drives permanently running at a time.
> 

The MTBF is always a good one.  The more bits you have the more there is
to go wrong.   Is it safer to fly a single engined plane?  You have
twice the chance of an engine failure on a twin.  YMMV.

Tom.