On Thu, May 15, 2003 at 06:42:18PM -0500, Nate Carlson wrote:
> On Thu, 15 May 2003, Nate Carlson wrote:
> > Just had a drive go dead (one of the Maxtor 160gb drives) that isn't
> > in a hotswap sled.. have to take Gladiator down to swap the drive out.
> >
> > Should be back up shortly.
> 
> Well, looks like we lost a second drive; the controller started throwing
> many CRC errors on a second one of the Maxtor drives when I started the
> rebuild. Ugh! Talked to 3ware tech support; they say these Maxtor drives
> go bad all the time, and gave us a few suggestions.. waiting to see what
> happens on a rebuild.
> 
> The way we've got Gladiator set up, there are two arrays (~600gb Western
> Digital, ~800gb Maxtor), and they are in a LVM VG with one big EXT3
> partition across it. The WD drives are first in the VG, then the Maxtor
> drives. We've got a total of 320gb of data on Gladiator, so if data is
> stored sequentially, all the data *should* be on the WD drives. Anyone
> know if it's possible to shrink an ext3 partition when the second half of
> it's on a dying disk? And then have to reduce the LVM partition, of
> course..

I would not bet on the data being stored sequentially: the inode
allocator tries to spread the directory inodes around to reduce
fragmentation - if you leave some space between directories A and B, A
can grow without getting fragmented.

Cheers,
florin

-- 

"NT is to UNIX what a doughnut is to a particle accelerator."
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: not available
Url : http://shadowknight.real-time.com/pipermail/tclug-list/attachments/20030515/80f540c1/attachment.pgp