[Gluster-users] Order of subvolumes in afr

[Dave Hall]

On Fri, 2010-02-26 at 12:04 -0800, Vikas Gorur wrote:
> On Feb 25, 2010, at 2:25 AM, Paras Fadte wrote:
> > 
> > 
> > Would that mean that when I remount  */mnt/gluster/data  ,*   data listed
> > under it would be taken from*  /home/data_backup*  directory from Server B
> > ?. In other words  "B" would  become "primary"  gluster server and "A" would
> > be "secondary" ?
> > 
> 
> 
> There is no "primary" or "secondary" servers when it comes to replicate.
> Writes are done to all servers synchronously and reads are distributed
> across all servers. Changing the order would not do anything.

You can make gluster only read from a designated node/volume but adding 
option read-subvolume <vol-name>
to your replicate definition.

something like this

volume vol-a
  type protocol/client
  option transport-type tcp
  option remote-host 1.2.3.4
  option remote-subvolume brick
end-volume
[...]
volume replication
  type cluster/replicate
  subvolumes vol-a vol-b vol-c [...]
  option read-subvolume vol-a
end-volume

Cheers

Dave