Gluster users - Feb 2014 - How to add a new node to a distributed replicated volume

Hello gluster users and experts,

I've been using gluster for several distributed volumes for a long while,
and am very happy with it.

However, I have stumbled upon a problem I was unable to find a solution to
in the documentation and the archives of this list. Would be happy to find
the right way to approach it.

I have a distributed-replicated volume that looks something like that (with
8 machines, rather then the 3 I put in the example):
Bricks:
Brick1: machine1:/data
Brick2: machine2:/data_replica
Brick3: machine2:/data
Brick4: machine3:/data_replica
Brick5: machine3:/data
Brick6: machine1:/data_replica

I want to add machine4 with directories /data and /data_replica too, but I
don't want it to be a replica of itself, so I want the two new bricks to be
added as brick6 and brick7, and push current brick6 to be brick8.

Is there a way to do this?

My current solution was to try and remove-brick my brick5 and brick6 and
later add them back with the two new bricks in the correct order. I started
this process and it's been more than a day now and still the state of the
removal (using "gluster volume remove-brick ... status") is that two
machines have completed the rebalance and the rest have not started.

Calling "gluster volume info my_volume" still shows the removed bricks
as
part of the volume, so I assume it's not finished.

Many thanks in advance for your advice,
Ayelet
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://supercolony.gluster.org/pipermail/gluster-users/attachments/20140206/6ff072bc/attachment.html>