[Gluster-devel] oddity with the mountpoint

[Daniel Maher]

On Thu, 10 Apr 2008 09:44:00 +0200 "Daniel Maher"
<dma+gluster at witbe.net> wrote:

> On Wed, 9 Apr 2008 21:20:20 +0530 "Anand Avati" <avati at zresearch.com>
> wrote:
> 
> > Daniel,
> >  please send the client logs from that session.
> > 
> > avati
> 
> I am mounting the client via autofs (as per the wiki), and i do not
> know how to enable debug logging using this method.  As a result,
> there are literally no client logs to send you - the period of time
> during which this condition is apparent produces no log entries.
> 
> If somebody could tell me how to enable debug logging via autofs, i'd
> be more than happy to try that out.

Just to follow up on this, the condition is reliably re-created only
when mounting via autofs - if i mount from the commandline using
glusterfs manually, there is no problem.

Furthermore, the condition only occurs when i pass the --ghost
parameter to automount in the master config :
/opt  /etc/auto.gfs   --timeout=5 --ghost

If i don't pass --ghost, the problem doesn't occur; however, the
directory "disappears" if it's not being accessed.  After 5 seconds
(i.e. the value specified by timeout) /opt/gfs-mount ceases to exist :
# ls -l /opt
total 0

Oddly, if i cd to it, it works :
# cd /opt/gfs-mount
# ls
textfile.txt

Clearly, the --ghost feature of autofs is the culprit.  Has anybody
else had similar results using autofs with --ghost to automount Gluster
shares ?


-- 
Daniel Maher <dma AT witbe.net>