Re: [Gluster-devel] Adding nodes

[Steffen Grunewald]

On Tue, May 27, 2008 at 02:22:31PM -0700, Amar S. Tumballi wrote:
> >
> > To answer your original question, files within an AFR are healed from one
> > node to the other when the file is accessed (actually read) through the AFR
> > and one node is found to have more recent data than others.
> >
> 
> Just for understanding, this should be "AFR self-heal done when files are
> accessed  (actually open()'d)", it may not be even a read call. if you have
> a tool, which just does  'open()/close()' on a file, it gets synced.

What could a half-automated self-heal procedure look like? Since self-heal
would also have to handle file deletions during the absence of the brick to
be healed, or there may have been a split-brain situation in between, this
doesn't seem too trivial to me. 

Otherwise, I could imagine a 
        find directory -type f | xargs -n1 -i% dd if=% of=/dev/null count=1
run on the right brick... (even count=0 would do -- according to strace, dd 
would perform the full set of open(), llseek(), and close() -- thus it should
do the trick)


Steffen

-- 
Steffen Grunewald * MPI Grav.Phys.(AEI) * Am Mühlenberg 1, D-14476 Potsdam
Cluster Admin * http://pandora.aei.mpg.de/merlin/ * http://www.aei.mpg.de/
* e-mail: steffen.grunewald(*)aei.mpg.de * +49-331-567-{fon:7233,fax:7298}
No Word/PPT mails - http://www.gnu.org/philosophy/no-word-attachments.html