Gluster users - Mar 2013 - About adding bricks ...

Hi @all,

I've created a Distributed-Replicated Volume consisting of 4 bricks on
2 servers.

# gluster volume create glusterfs replica 2 transport tcp \
gluster0{0..1}:/srv/gluster/exp0 gluster0{0..1}:/srv/gluster/exp1

Now I have the following very nice replication schema:

+-------------+   +-------------+
|  gluster00  |   |  gluster01  |
+-------------+   +-------------+
| exp0 | exp1 |   | exp0 | exp1 |
+------+------+   +------+------+
    |      |          |      |
    +------+----------+      |
           |                 |
           +-----------------+

If one HD goes down, I'm backed! I'm even backed if one server goes
down!
Great! But how will it be if I just add another Server with two bricks and
rebalance?

# gluster volume add-brick glusterfs gluster02:/srv/gluster/exp{0..1}
# gluster volume rebalance glusterfs start

The replication schema I obviously want is this:

+-------------+   +-------------+   +-------------+
|  gluster00  |   |  gluster01  |   |  gluster02  |
+-------------+   +-------------+   +-------------+
| exp0 | exp1 |   | exp0 | exp1 |   | exp0 | exp1 |
+------+------+   +------+------+   +------+------+
    |      |          |      |          |      |
    |      +----------+      +----------+      |
    |                                          |
    +------------------------------------------+

But I think I'll maybe get this one:

+-------------+   +-------------+   +-------------+
|  gluster00  |   |  gluster01  |   |  gluster02  |
+-------------+   +-------------+   +-------------+
| exp0 | exp1 |   | exp0 | exp1 |   | exp0 | exp1 |
+------+------+   +------+------+   +------+------+
    |      |          |      |          |     |
    +------+----------+      |          +-----+
           |                 |
           +-----------------+

So how can I get the schema I want? Is this even possible? The examples
in the official 'Admin Guide' are quite trivial and don't cover this
scenario.

Cheers, Markus
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://supercolony.gluster.org/pipermail/gluster-users/attachments/20130320/0d8fac85/attachment.html>

Hi Markus,

Markus Hubig wrote on 20130320:
...
> # gluster volume add-brick glusterfs gluster02:/srv/gluster/exp{0..1}
...
> But I think I'll maybe get this one:
> 
> +-------------+   +-------------+   +-------------+
> |  gluster00  |   |  gluster01  |   |  gluster02  |
> +-------------+   +-------------+   +-------------+
> | exp0 | exp1 |   | exp0 | exp1 |   | exp0 | exp1 |
> +------+------+   +------+------+   +------+------+
>     |      |          |      |          |     |
>     +------+----------+      |          +-----+
>            |                 |
>            +-----------------+
> 
> So how can I get the schema I want? Is this even possible?
Afaik it involves replace-brick commands (use this on >= 3.3.1).

So instead of the add-brick you'd do something like this :

- replace-brick gluster00:exp0 gluster02:exp0

You go to sleep for 2 days ;-)

- replace-brick gluster01:exp0 gluster02:exp1

You sleep another 2 days, then

- add-brick gluster00:exp0 gluster01:exp0

and then start the 2 rebalance steps.

regards,
   Hans Lambermont
-- 
Hans Lambermont | Senior Architect
(t) +31407370104 (w) www.shapeways.com