[Gluster-users] Unable to delete files but getfattr shows file is part of glusterfs

[SINCOCK John]

Hi Everybody,

 

We have recently setup a 2-node gluster, and are having some problems
deleting some files!

We haven't accessed the gluster bricks directly, all access has been
through the glusterfs mount 

The 1st node in our cluster also handles sharing the glusterfs volume
via samba, to windows clients.

 

Environment is:

Kernel: 2.6.32-431.1.2.0.1.el6.x86_64 #1 SMP Fri Dec 13 13:06:13 UTC
2013 x86_64 x86_64 x86_64 GNU/Linux

CentOS release 6.4 (Final)

 

glusterfs-libs-3.4.1-3.el6.x86_64

glusterfs-cli-3.4.1-3.el6.x86_64

glusterfs-fuse-3.4.1-3.el6.x86_64

glusterfs-server-3.4.1-3.el6.x86_64

glusterfs-3.4.1-3.el6.x86_64

 

               4 bricks on each node

               Gluster is setup without replication or striping

 

               Bricks on one of the two nodes (the other node is setup
the same):

/dev/sda1             9.1T  4.7T  4.5T  51% /mnt/glusterfs/bricks/1

/dev/sda2             9.1T  3.7T  5.5T  40% /mnt/glusterfs/bricks/2

/dev/sda3             9.1T  3.7T  5.5T  40% /mnt/glusterfs/bricks/3

/dev/sda4             9.1T  4.1T  5.0T  46% /mnt/glusterfs/bricks/4

 

Gluster volume mounted on 1st node:

g-unit-1:/vol00        73T   31T   43T  42% /gluster/vol00

               Mount options:

g-unit-1:/vol00 on /gluster/vol00 type fuse.glusterfs
(rw,nosuid,nodev,noatime,default_permissions,allow_other,max_read=131072
)

 

So the problem is this:

[root at g-unit-1 ~]# rm -Rf /gluster/vol00/data/CentOS3/CentOS_6.4_Final

rm: cannot remove
`/gluster/vol00/data/CentOS3/CentOS_6.4_Final/Packages': Directory not
empty

rm: cannot remove
`/gluster/vol00/data/CentOS3/CentOS_6.4_Final/EFI/BOOT': Directory not
empty

[root at g-unit-1 ~]#

 

And the delete leaves behind 2000 or so files:

[root at g-unit-1 ~]# find /gluster/vol00/data/CentOS3/CentOS_6.4_Final/
-type f | wc -l

2185

Almost all the left behind files are on the 2nd node, g-unit-2, but
there are a few on g-unit-1

 

>From the 2nd node, g-unit-2:

[root at g-unit-2 ~]# getfattr -m . -d -e hex
/mnt/glusterfs/bricks/2/data/CentOS3/CentOS_6.4_Final/EFI/BOOT/BOOTX64.e
fi

getfattr: Removing leading '/' from absolute path names

# file:
mnt/glusterfs/bricks/2/data/CentOS3/CentOS_6.4_Final/EFI/BOOT/BOOTX64.ef
i

trusted.gfid=0xa9b0a66cf39b49c7890b2da3f631929f

So the file is there, is tagged with some gluster info, and has
definitely been copied onto the volume properly, via the glusterfs
mount. It hasn't been copied directly onto the brick.

 

But from g-unit-1, accessing via the glusterfs mount, and trying to
delete this file:

[root at g-unit-1 ~]# rm
/gluster/vol00/data/CentOS3/CentOS_6.4_Final/EFI/BOOT/BOOTX64.efi

rm: remove regular file
`/gluster/vol00/data/CentOS3/CentOS_6.4_Final/EFI/BOOT/BOOTX64.efi'? y

rm: cannot remove
`/gluster/vol00/data/CentOS3/CentOS_6.4_Final/EFI/BOOT/BOOTX64.efi': No
such file or directory

 

What the hell!?!!!!!

 

Can someone please suggest the best course of action to resolve this and
prevent it happening in the future?

 

Thanks for any assistance.

 

John

 

 

 

 

 

 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://supercolony.gluster.org/pipermail/gluster-users/attachments/20140615/3de40562/attachment.html>