[Gluster-users] Does glusterfs supports brick locality?

[Jeff Darcy]

On Mon, Apr 24, 2017, at 10:14 AM, atris adam wrote:
> I have two data centers in two different provinces. Each data centers
> have 3 servers. I want to setup cloud storage with glusterfs.> I want to make one glusterfs volume with these information.
> 
> province "a"==> 3 servers, each server has one 5TB brick (bricks
> number is from 1-3)> province "b"==> 3 servers, each server has one 5TB brick (bricks
> number is from 4-6)> 
> distributed gluster volume size: 30TB
> 
> Does glusterfs support servers with long distance?
> If yes, if an end user in province "a" create a file, where the file
> is allocated? I mean which brick is selected to write the file on it?
> I need the brick number from 1-3 to be> selected. Because I think if glusterfs select other bricks (num 4-6,
> which is in province "b"), higher latency and unnecessary network
> traffic will be consumed. Am I right?
The "nufa" translator/option exists for exactly this purpose, but it's a
bit more limited than what I think you want.   When creating a file, it
will create it in a brick *on the same node* if possible, but it doesn't
distinguish between other nodes near or far away.  This might be good
enough if you're using NFS as the access protocol, because in that case
the "same node" is the one running the NFS/Gluster proxy.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.gluster.org/pipermail/gluster-users/attachments/20170424/9441d6dc/attachment.html>