[Gluster-users] How to write a volumebackup file with 4 nodes in replica 2

[Joe Julian]

On 09/01/2015 12:39 AM, Davy Croonen wrote:
> Hi
>
> We are currently running a gluster cluster with 2 nodes in replica 2. We mount the distributed volume with the gluster fuse client and a volumebackup file for failover.
>
> The volumebackup file looks as follows:
>
> volume node1
>    type protocol/client
>    option transport-type tcp
>    option remote-host 10.0.0.x
>    option remote-subvolume /mnt/volume/brick1
> end-volume
>
> volume node2
>    type protocol/client
>    option transport-type tcp
>    option remote-host 10.0.0.y
>    option remote-subvolume /mnt/volume/brick1
> end-volume
>
> volume replicate
>    type cluster/replicate
>    subvolumes node1 node2
> end-volume
>
> We now want the expand our volume with 2 extra nodes also in replica 2. The steps to achieve this are clearly explained in the manual, but what with the volumebackup file? Do we need to change this file and add the 2 extra nodes also or will the cluster manage this internally?
>
> If we have to add the 2 extra nodes to the volumebackup file, do we then have to do this as follows?
>
> volume node1
>    type protocol/client
>    option transport-type tcp
>    option remote-host 10.0.0.x
>    option remote-subvolume /mnt/volume/brick1
> end-volume
>
> volume node2
>    type protocol/client
>    option transport-type tcp
>    option remote-host 10.0.0.y
>    option remote-subvolume /mnt/volume/brick1
> end-volume
>
> volume node3
>    type protocol/client
>    option transport-type tcp
>    option remote-host 10.0.0.z
>    option remote-subvolume /mnt/volume/brick1
> end-volume
>
> volume node4
>    type protocol/client
>    option transport-type tcp
>    option remote-host 10.0.0.a
>    option remote-subvolume /mnt/volume/brick1
> end-volume
>
> volume replicate
>    type cluster/replicate
>    subvolumes node1 node2 node3 node4
> end-volume
>
> Kind regards
> Davy
> _

There's no need to write vol files since 3.0. If you're doing that, as 
you state "for failover", which I assume means you're using that in case 
the mount server is unavailable, that's unnecessary. Either use rrdns 
and specify multiple A records for a psuedo-hostname that points to each 
server, or use the backup-volfile-servers mount option.