Re: LVM and disk failure
On Sat, 2006-01-07 at 22:15, Daniel Webb wrote:
> On Sat, Jan 07, 2006 at 09:02:20PM -0800, Mike Bird wrote:
> Well, yes, but supposing you *do* have a failure? Then what? Half the
> filesystem is still there on the second disk, is it recoverable, and if not,
> why not?
You may get some of the data. You probably won't get all of it.
I've had the great good fortune to have a group of bad blocks
develop at a place on a drive where no data was currently
stored, and I've lost data to bad blocks too.
> I'm getting the impression that spanning volume groups with a logical volume
> is a *very* bad idea unless the physical volumes are RAID.
A VG built on two single drive PV's is twice as large but roughly
half as reliable as a single drive. Depending upon the kind
of data, that may or may not be a bad idea.
Almost all of our VG's are built on RAID-1 PV's.
However, we have a VG which contains the LV which contains our
partial Debian mirror, together with a bunch of similar stuff.
It's huge but it can be rebuilt if a drive dies. It's not
worth it to us to double up the 1.8TB to get RAID-1 protection
that isn't needed for that particular kind of data.
--Mike Bird
Reply to: