[Gluster-devel] write-behind mtime workaround

[Steffen Grunewald]

On Thu, Apr 26, 2007 at 10:57:20PM -0400, Brent A Nelson wrote:
> Setting "option aggregate-size 0" does not eliminate the mtime glitch 
> 100%.  Copying /usr to an afr/unify glusterfs, I found that almost 
> everything was okay, but a handful of files on one server had updated 
> mtimes, and a different handful on the other server did, as well.

This is kind of crucial since self-healing relies on inode metadata
(including mtime)...

> So, even with aggregation out of action, write-behind will still 
> occasionally alter mtimes on a "cp -a".

An easy way to detect this should be to use rsync in both directions, with
the -n option. (rsync -avn source_dir/. target_dir/.)

Cheers
 Steffen