Gluster users - Jun 2011 - Suggestions welcome for expanded single array and redundancy

Hi

We have started evaluating Glusterfs and I am very impressed with what we have
seen and tested.  I am hoping someone can point me in the right direction or
offer a solution.  We have multiple servers which we wish to use for storage. 
We would like to have a single storage array accessible by client machines on
the network.  We would like to achieve N+2 redundancy but have the ability to
keep adding servers to increase the available overall storage.  We basically
want a big NFS accessible network storage with underlying redundancy across
storage on different servers.

Example
3 x Servers - each with Gluster configured with 1TB of storage
Configured for replication
Available storage 1TB
Redundancy 2 x copies


The above is straight forward.  What we want to achieve is
Add more servers (for the sake of the exercise specs will be identical)
Keep the same level of redundancy N+2
Increase storage
If 2 more servers were installed (5 x 1TB) - (2 redundant) = 3TB Available
If 4 more servers were installed (7 x 1TB) - (2 redundant) = 5TB Available
and so on

Is this possible with glusterfs and any suggestions on how to achieve this. 
Please keep in mind we are expecting data growth and would like to keep adding
servers to increase storage.

Regards
Stewart

All you need to do is add 2 new bricks to existing volume. Then run
fix-layout because existing directories will still have old hash
mapping and then run re-balance which will copy files that hash to new
hash mapping on the new bricks, this is only required if you want
existing directories to hash to new bricks. Other option could be to
create new dirs. But it depends on your requirements.

http://www.gluster.com/community/documentation/index.php/Gluster_3.2:_Rebalancing_Volume_to_Fix_Layout_Changes


On Fri, Jun 3, 2011 at 1:38 AM, Stewart Campbell
<stewartcampbell at assureprograms.com.au> wrote:
> Hi
>
> We have started evaluating Glusterfs and I am very impressed with what we
have seen and tested. ?I am hoping someone can point me in the right direction
or offer a solution. ?We have multiple servers which we wish to use for storage.
?We would like to have a single storage array accessible by client machines on
the network. ?We would like to achieve N+2 redundancy but have the ability to
keep adding servers to increase the available overall storage. ?We basically
want a big NFS accessible network storage with underlying redundancy across
storage on different servers.
>
> Example
> 3 x Servers - each with Gluster configured with 1TB of storage
> Configured for replication
> Available storage 1TB
> Redundancy 2 x copies
>
>
> The above is straight forward. ?What we want to achieve is
> Add more servers (for the sake of the exercise specs will be identical)
> Keep the same level of redundancy N+2
> Increase storage
> If 2 more servers were installed (5 x 1TB) - (2 redundant) = 3TB Available
> If 4 more servers were installed (7 x 1TB) - (2 redundant) = 5TB Available
> and so on
>
> Is this possible with glusterfs and any suggestions on how to achieve this.
?Please keep in mind we are expecting data growth and would like to keep adding
servers to increase storage.
>
> Regards
> Stewart
> _______________________________________________
> Gluster-users mailing list
> Gluster-users at gluster.org
> http://gluster.org/cgi-bin/mailman/listinfo/gluster-users
>