[Gluster-users] replica count documentation

Jeff Darcy jdarcy at redhat.com
Fri Apr 15 13:56:39 UTC 2011


On 04/14/2011 05:13 PM, Whit Blauvelt wrote:
> On similar systems, there are options to have control return to the user
> either after the first of two mirrored storage locations is written to, or
> to have it wait until after the second is written to. Does Gluster have such
> an option? If so, what does it default to? 

The replication that's in GlusterFS right now (AFR) is of a
synchronous/local type.  Gluster is working on "geosync" which provides
a sort of DR-oriented asynchronous replication with very weak
consistency/ordering properties.  I'll let someone from Gluster comment
on their plans beyond that.  Separately, as part of the CloudFS project
which is based on GlusterFS, I'm working on fully ordered asynchronous
replication between multiple peer sites.  Since that's a more
challenging problem, and also somewhat blocked behind other more
pressing priorities, that will take longer to complete.



More information about the Gluster-users mailing list