[Gluster-devel] AFR Replication

[Christopher Hawkins]

> You mean, have servers AFR as clients, then re-export the 
> AFR-ed volume again? GlusterFS on top of GlusterFS?

Yup. The distinction between server and client is a gray area. 
There are exports and imports, and they mix and match. 

> Sure, I accept that. I was just asking if there was a way to 
> make the additional writes server-side, because servers are 
> few and clients are many, so n* the server bandwdth will 
> generally be smaller than server*client bandwidth.

True enough. I think the cost / benefit is an open question...
The more you leave to the clients, the less dependent they are 
on the servers. Usually you don't have a choice, but here you do.  
 
> I can see that it's less of an issue than block-level 
> split-brain, because this would at most lead to the odd file 
> getting corrupted, whereas block-level split-brain would 
> destroy the entire FS very quickly.

Good distinction. You are certainly right that there are
more possibilities for split brain than I am thinking of, but wouldn't
the worst case scenario be that two versions of the same file(s) are
getting written, and then when you sync only one of them will remain?

In that case you have no corruption, just lost changes on the file that got
tossed. 
I'm not sure that corruption is out of the question, just thinking out loud.


Chris