I don't disagree, but is there any *particular* reason setxattr would be preferred?

(In reply to comment #1)
> I don't disagree, but is there any *particular* reason setxattr would be
> preferred?

We are overhauling gluster volume rebalance. Part of that is to be able to have fine grained control of how a new layout should be. Eg. to span a directory across N nodes, or exclude host XYZ etc. setxattr's value will be used to specify such parameters for the 'fix.layout' key.

Avati

PATCH: http://patches.gluster.com/patch/7597 in master (distribute: handle 'fix.layout' key in setxattr() instead of getxattr())

verified with 3.2.3qa3 - it didn't show any invalid argument error message.

with 3.2.2.
#setfattr -n trusted.glusterfs.fix.layout -v something /mnt/
#getfattr -d -m . /mnt
gave error message like 

"
[2011-08-19 14:31:59.36928] E [stat-prefetch.c:695:sp_remove_caches_from_all_fds_opened] (-->/opt/glusterfs/3.2.2//lib/glusterfs/3.2.2/xlator/mount/fuse.so(fuse_setxattr_resume+0x427) [0x7f28baa1aec2] (-->/opt/glusterfs/3.2.2//lib/glusterfs/3.2.2/xlator/debug/io-stats.so(io_stats_setxattr+0x2b0) [0x7f28b3df86a2] (-->/opt/glusterfs/3.2.2//lib/glusterfs/3.2.2/xlator/performance/stat-prefetch.so(sp_setxattr+0x151) [0x7f28b804ec22]))) 0-bug764807-stat-prefetch: invalid argument: inode
"