[Lustre-discuss] Lustre 1.6.6 - cannot get rid of user_xattr on Sles10

Jeffrey Alan Bennett jab at sdsc.edu
Thu Dec 4 14:32:58 PST 2008

Hi Andrei,
What type of filesystem is there on /tmp ?

If you had, say an NTFS or FAT filesystem, you would get an error like that, since these filesystems dont support the same attributes as Unix-based filesystem, that is "user, group, others, etc..."



	From: lustre-discuss-bounces at lists.lustre.org [mailto:lustre-discuss-bounces at lists.lustre.org] On Behalf Of Andrei Maslennikov
	Sent: Thursday, December 04, 2008 2:26 PM
	To: lustre-discuss at lists.lustre.org
	Subject: [Lustre-discuss] Lustre 1.6.6 - cannot get rid of user_xattr on Sles10
	On an only RHEL 4.7 OSS/MDT machine used in some standalone setup we have:
	# mount
	/dev/sdc8 on /lusys/mgt type lustre (rw,noacl,nouser_xattr)                  <--- MDT
	/dev/lustre/united on /lusys/ost1 type lustre (rw,noacl,nouser_xattr)       <--- OST at o2ib:/lustre on /lustre type lustre (rw,noacl,nouser_xattr)    <--- Client 
	On this machine I can create files inside Lustre(/lustre) and then move them 
	to any other local file system withut any problem.
	Instead, on a Suse SLES10SP2 client machine with unmodified kernel we got:
	# mount at o2ib:/lustre on /lustre type lustre (rw,noacl,nouser_xattr)
	However, any attempt to move any file from inside /xenfs to, say, /tmp
	ends up correctly (file is being moved), but insists on the following error:
	# mv /lustre/uuu /tmp
	mv: setting attributes for `/tmp/uuu': Operation not supported
	Could someone advice us on how to get rid of this? Not so long ago somebody
	had a similar issue, but the problem apparently remained unsolved. There should
	be some setting overlooked by us on Suse only (on Red Hat it works)... but
	what may this be?...
	PS tunefs.lustre accepts the persistent setting of nouser_xattr, but puts it *after* 
	Persistent mount opts: errors=remount-ro,iopen_nopriv,user_xattr,nouser_xattr
	This however does not change anything as we mount /lustre with nouser_xattr
	options on the common line. The issue is still there...

More information about the lustre-discuss mailing list