Gluster users - Jan 2014 - distribute-replicate with 3 storage node with different disk size

Hi all,

I have not tested this yet, but would like to get your comment.

I have a customer that is currently running BackupPC system, and the local
disk is steadily going to run out of space soon.

So I'm thinking of helping this customer move the storage to gluster

They have limited budget, so buying new machine with big disk is not an
option. But, they have a couple of old machine available, and the disk size
differs in each of them

So, I'm planning to grow the storage slowly for them.

The first step that I think is to get 2 replicated storage nodes, so that
redundancy can be increased.

Once that work, I'm planning to introduce another node to get a
distributed-replicated system. This is the part that I'm not sure.

Let say the customer have nodes like this:
node1: 100GB (p1n1)
node2: 200GB (p1n2,p2n2)
node3: 300GB (p1n3)

node1 is currently the single BackupPC storage. I can partition node2 to
get a 100GB (p1n2), then join with node1 (p1n1) to get a replicated gluster
storage.

When I introduce node3 later, I plan to use the extra 100GB from node2
(p2n2), to join with 100GB from node3 (p1n3). Now, I have 4 bricks defined.
1 brick in node1 (p1n1), 2 brick in node2 (p1n2,p2n2), and 1 brick in node3
(p1n3).

The question is, how to change from replicated system (p1n1-p1n2), to
distributed-replicated [(p1n1-p1n2)(p2n2-p1n3)] in a running system?

Ideas?

Thanks.


-- 
Sharuzzaman Ahmat Raslan
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://supercolony.gluster.org/pipermail/gluster-users/attachments/20140108/0612d770/attachment.html>

Seems that nobody is answering my concern :)

I think these post partially answer my question, but I think I still need
to perform the experiment to understand more:

http://gluster.org/pipermail/gluster-users/2011-February/006649.html
http://gluster.org/pipermail/gluster-users/2011-February/006642.html
http://review.gluster.com/#/c/158/
http://www.ashishnepal.com/glusterfs-centos/

Review-158 is the most detail info I found on the Internet, and did not
found it in any Gluster documentation.

Maybe I should start my own wiki collecting all the information found on
the Internet.

Thanks.






On Wed, Jan 8, 2014 at 10:48 AM, Sharuzzaman Ahmat Raslan <
sharuzzaman at gmail.com> wrote:
> Hi all,
>
> I have not tested this yet, but would like to get your comment.
>
> I have a customer that is currently running BackupPC system, and the local
> disk is steadily going to run out of space soon.
>
> So I'm thinking of helping this customer move the storage to gluster
>
> They have limited budget, so buying new machine with big disk is not an
> option. But, they have a couple of old machine available, and the disk size
> differs in each of them
>
> So, I'm planning to grow the storage slowly for them.
>
> The first step that I think is to get 2 replicated storage nodes, so that
> redundancy can be increased.
>
> Once that work, I'm planning to introduce another node to get a
> distributed-replicated system. This is the part that I'm not sure.
>
> Let say the customer have nodes like this:
> node1: 100GB (p1n1)
> node2: 200GB (p1n2,p2n2)
> node3: 300GB (p1n3)
>
> node1 is currently the single BackupPC storage. I can partition node2 to
> get a 100GB (p1n2), then join with node1 (p1n1) to get a replicated gluster
> storage.
>
> When I introduce node3 later, I plan to use the extra 100GB from node2
> (p2n2), to join with 100GB from node3 (p1n3). Now, I have 4 bricks defined.
> 1 brick in node1 (p1n1), 2 brick in node2 (p1n2,p2n2), and 1 brick in node3
> (p1n3).
>
> The question is, how to change from replicated system (p1n1-p1n2), to
> distributed-replicated [(p1n1-p1n2)(p2n2-p1n3)] in a running system?
>
> Ideas?
>
> Thanks.
>
>
> --
> Sharuzzaman Ahmat Raslan
>


-- 
Sharuzzaman Ahmat Raslan
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://supercolony.gluster.org/pipermail/gluster-users/attachments/20140109/71d99c43/attachment.html>