Gluster users - Jan 2013 - DHT vs LVM for multiple bricks on a server

Hi,

I've been reading up on GlusterFS and I'm looking for best practices
around
using multiple disks as bricks in servers that will be part of a replicated
volume.

Say I start with a single disk each in two servers (/dev/sda1 mounted at /a)

gluster volume create test-volume replica 2 transport tcp server1:/a server2:/a


Then I add a second disk in each server (/dev/sdb1 mounted at /b)

gluster volume add-brick test-volume replica 2 transport tcp
server1:/b server2:/b


With this (after rebalancing), am I correct in understanding that I will
have a distributed replicated volume with GlusterFS providing the
equivalent of RAID1+0 for data on my volume.

Now as I understand, I will be restricted to adding disks (bricks) of the
same size whenever I need to extend the volume. What are the pros/cons of
instead using LVM to provide a single LV on each server and extending the
LV and filesystem each time I add additional storage? The other benefit to
LVM being the ability to take snapshots. The one downside I foresee is that
a concatenated LV will not use the second PV (disk) till the first PV is
full, though I could perhaps stripe?

More questions to follow, but I'm trying to think through this before I get
started with my first deployment.

TIA
Gaurav
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://supercolony.gluster.org/pipermail/gluster-users/attachments/20130107/013dd342/attachment.html>

*bump*


On Mon, Jan 7, 2013 at 8:13 PM, Gaurav P <gaurav.lists+gluster at
gmail.com>wrote:
> Hi,
>
> I've been reading up on GlusterFS and I'm looking for best
practices
> around using multiple disks as bricks in servers that will be part of a
> replicated volume.
>
> Say I start with a single disk each in two servers (/dev/sda1 mounted at
> /a)
>
> gluster volume create test-volume replica 2 transport tcp server1:/a
server2:/a
>
>
> Then I add a second disk in each server (/dev/sdb1 mounted at /b)
>
> gluster volume add-brick test-volume replica 2 transport tcp server1:/b
server2:/b
>
>
> With this (after rebalancing), am I correct in understanding that I will
> have a distributed replicated volume with GlusterFS providing the
> equivalent of RAID1+0 for data on my volume.
>
> Now as I understand, I will be restricted to adding disks (bricks) of the
> same size whenever I need to extend the volume. What are the pros/cons of
> instead using LVM to provide a single LV on each server and extending the
> LV and filesystem each time I add additional storage? The other benefit to
> LVM being the ability to take snapshots. The one downside I foresee is that
> a concatenated LV will not use the second PV (disk) till the first PV is
> full, though I could perhaps stripe?
>
> More questions to follow, but I'm trying to think through this before I
> get started with my first deployment.
>
> TIA
> Gaurav
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://supercolony.gluster.org/pipermail/gluster-users/attachments/20130110/542183aa/attachment.html>