Bug 762529 (GLUSTER-797)

Summary:	DHT layout and posix generation number xattr's should be filtered
Product:	[Community] GlusterFS	Reporter:	Vikas Gorur <vikas>
Component:	core	Assignee:	Sachidananda Urs <sac>
Status:	CLOSED CURRENTRELEASE	QA Contact:
Severity:	high	Docs Contact:
Priority:	urgent
Version:	mainline	CC:	allen, amarts, chida, gluster-bugs, krishna, lakshmipathi, vijay
Target Milestone:	---
Target Release:	---
Hardware:	All
OS:	All
Whiteboard:
Fixed In Version:		Doc Type:	Bug Fix
Doc Text:		Story Points:	---
Clone Of:		Environment:
Last Closed:		Type:	---
Regression:	RTP	Mount Type:	---
Documentation:	---	CRM:
Verified Versions:		Category:	---
oVirt Team:	---	RHEL 7.3 requirements from Atomic Host:
Cloudforms Team:	---	Target Upstream Version:

Description Vikas Gorur 2010-03-31 18:26:59 UTC

The DHT layout and posix generation number xattr's are visible on the mountpoint. This can cause problems if someone does:

rsync -X /mnt/glusterfs1 /mnt/glusterfs2

As this will overwrite the generation numbers and cause ESTALE errors.

Comment 1 Krishna Srinivas 2010-04-06 18:19:57 UTC

Vikas found following:
-----
ESTALE because of this:

[2010-03-31 18:59:45] D [dht-layout.c:649:dht_layout_dir_mismatch] distribute: / - disk
layout missing
[2010-03-31 18:59:45] D [dht-common.c:274:dht_revalidate_cbk] distribute: mismatching
layouts for /

On the backend, layout xattr is missing on the first subvolume (replicate pair) and present
on the second subvolume. Value on second subvolume:

trusted.glusterfs.dht=0x00000001000000007fffffffffffffff

This happens when rsync -rav -X is done.
-----

Comment 2 Amar Tumballi 2010-04-20 08:04:41 UTC

Needs a small fix in dht_getxattr ()

Comment 3 Anand Avati 2010-04-23 10:33:01 UTC

PATCH: http://patches.gluster.com/patch/3156 in master (Filter out extended attribute visibility on the mount point.)

Comment 4 Anand Avati 2010-04-23 10:33:18 UTC

PATCH: http://patches.gluster.com/patch/3163 in release-3.0 (Filter out extended attribute visibility on the mount point.)

Comment 5 Allen Lu 2010-04-26 13:00:39 UTC

Do we have this fixed or not? This is a P1 bug.

Comment 6 Vijay Bellur 2010-04-26 13:16:03 UTC

(In reply to comment #5)
This has been fixed.

Comment 7 Lakshmipathi G 2010-06-11 02:28:11 UTC

verified with 3.0.5rc6. rsync  files from native fs to gluster mount pt and also tested between two gluster mount pts - didn't produce the "Stale NFS file handle"  error.

rsync -Xav /lib /home/laks/workpath/regression/797/regr/mnt/client1
rsync -Xav /usr/bin/ /home/laks/workpath/regression/797/regr/mnt/client1
rsync -Xav /lib /home/laks/workpath/regression/797/regr/mnt/client1