Re: [Gluster-devel] AFR

[Amar S. Tumballi]

Actually GlusterFS can be configured by volume spec files. It doesn't have 
any hardcoded values. So, it all depends on how you write your spec file. 

Say you export 3 harddisk partition from a single machine, make sure you don't 
have the same volumes as your child node for AFR. (Finding it weird? i 
strongly suggest you to understand our translator/spec file design).

Dale,
 the information about where the file replica is there is known by spec file 
(the order of subvolumes, as of now). with the introduction of namespace it 
should not be a problem.

- bulde

On Fri, Feb 16, 2007 at 07:42:01PM -0800, Dale Dude wrote:
> Amar,
> 
> I think Brent is asking for #4 whether replicas will on different 
> servers/bricks so no single server/brick will have the same replica in 
> case of server/brick failure?
> 
> Added question...How will glusterfs know what servers/bricks contain 
> particular replicas? Is there a distributed dbase with this info? I 
> would assume something like this is needed esepcially for the fsck 
> (which I was interested in working on if noone has already)? Maybe you 
> guys have an interesting approach to share before release or should I go 
> to the code?
> 
> Regards,
> Dale
>