[Gluster-devel] AFR write completion? AFR read redundancy?

[Martin Fick]

--- Anand Avati <avati at zresearch.com> wrote:
> >
> > First, I am curious as to the write semantics, 
> > when does AFR report a write complete?  Does a 
> > write have to make it to all of the subvolumes 
> > or just one before it claims a write complete?
> 
> AFR waits for completion (either success or failure)
> from all subvolumes before returning. 

Cool. In the case of mixed results (some success, some
failures), what is the result?  Anyway to distinguish
this from a complete success?

> If you need one of your subvolumes to be written in 
> background, you need to load write-behind in that 
> path.

Makes sense.


> > Also, what about reads, does it only read from one
> > subvolume?  Is there any way to make it read from
> > all subvolumes to check/correct for errors?
> 
> it does a read load balancing per file. it does not
> check/correct errors.

Any plans for such a feature?  HA is great, but so is
error resilience/correction. :)

Thanks for the quick responses!

-Martin



      ____________________________________________________________________________________
Never miss a thing.  Make Yahoo your home page. 
http://www.yahoo.com/r/hs