[Gluster-devel] Replication resync after disconnection

[gordan at bobich.net]

On Wed, 23 Apr 2008, Gareth Bult wrote:

> Which node are you mounting against?

Each node mounts against localhost (for speed) and replicated to remote 
node(s).

> If each node mounts against the local server, you might like to check out this bug;
> http://savannah.nongnu.org/bugs/?21918

That sounds _exactly_ like what I'm seeing. Surely, mounting against a 
remote server is daft when there is data locally?!

Would this be worked around if AFR were implemented on the server rather 
than client side?

> It's been outstanding for nearly 4 months and there's no indication this 
> issue has been addressed. Essentially to ensure a consistent view, all 
> clients need to mount against the same server .. which is sort of Ok as 
> they should both fail-over to the (same) next available node in the 
> event of a server crash.

Except in the case where you have local copy of the data, getting the data 
off the remote server seems... suboptimal.

Does anyone know when this will be fixed?

Gordan