[Dirvish] LVM2 and dirvish

Jason Boxman jasonb at edseek.com
Sat Feb 26 15:35:16 PST 2005

On Saturday 26 February 2005 18:02, Nathan Hunsperger wrote:
> Figure I'll chime in here too.  I'm also using LVM2, but in a different
> manner than what has been discussed so far; I'm using LVM for the
> backup volumes, rather than for the backed-up volumes.  In my setup, I
> have a dedicated backup server with a 600GB software raid5 filesystem,
> in addition to the root fs.  This server currently serves backups over
> NFS for 3 file systems.

I would've done this, too, but I created the backup array before I started 
using LVM2.  The only installation media I have that has LVM2 support and 
tools included is Debian's new DI, so I had been hesitant in the past to move 
wholesale to LVM.

> One note:  In several guides on using Dirvish, there is talk about
> changing the size of blocks to maximize storage.  If you do so, and
> are using linux software raid5, make certain that every filesystem on
> the raid5 volume has the same block size.  If you don't, your raid5
> performance will be pathetic, as linux will flush the block cache every
> time an I/O request is made for a different block size.  This also has
> the side effect of filling up /var/log quite quickly.

Yes, I remember experiencing that when I was running software RAID 5.  I had 
so many warnings in syslog I was contemplating removing it from the kernel 
source entirely.  I was using XFS, though.  As far as I know, you can only 
change the the blocksize and stride for ext3 filesystems.  I haven't found a 
way to do it with XFS, which is what I'm using on the backup server.

Now I'm using a 3Ware Escalade 7410 and hardware RAID with an unalterable 
stripe size of 64K.  (Soon to be a 7500-4LP.)


Jason Boxman
Perl Programmer / *NIX Systems Administrator
Shimberg Center for Affordable Housing | University of Florida
http://edseek.com/ - Linux and FOSS stuff

More information about the Dirvish mailing list