[Gluster-users] Why does glusterfs not automatically fix these kinds of problems?

[Will Rouesnel]

Simple unify configuration with bricks running on the same system (multiple
hard disks):
 
2008-10-02 02:11:14 E [fuse-bridge.c:715:fuse_fd_cbk] glusterfs-fuse:
1534391: (12) /home/will/documents/University/PhD 2008/Experiments/Silver
nanocube rapid polyol synthesis/Photos/P8130384.JPG => -1 (5)
2008-10-02 02:11:14 E [unify.c:882:unify_open] unify:
/home/will/documents/University/PhD 2008/Experiments/Silver nanocube rapid
polyol synthesis/Photos/P8130385.JPG: entry_count is 1
2008-10-02 02:11:14 E [unify.c:885:unify_open] unify:
/home/will/documents/University/PhD 2008/Experiments/Silver nanocube rapid
polyol synthesis/Photos/P8130385.JPG: found on brick7
 
This only seems to happen with specific directories in my home directory
(which is hosted on glusterfs and I keep synchronized with Unison). For
whatever reason, this seems to be a recurring error - the files are
physically present on the bricks, but simply will not show up in the
namespace or under the actual glusterfs mount.
 
I've tried deleting the namespace brick (with gluster offline) and I am
definitely running the same server and client version (1.3.12). Does anyone
have any ideas *why* this keeps happening and why it would keep happening to
(apparently) only files in certain directories? I have about 400-500 files
not showing up in this fashion. Moreover, how come despite finding the
files, does glusterfs apparently not self-heal them to the namespace brick?
Can I adjust the trace translater to give an idea on whether it can access
the files or not?
 
- Will
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://supercolony.gluster.org/pipermail/gluster-users/attachments/20081002/d836af16/attachment.html>