[Gluster-users] different size of file in replicated volume using sparse file

Marco Agostini comunelevico at gmail.com
Thu Jun 5 09:22:44 UTC 2014


I'm using Gluster 3.4.2 to manage qcow2 file used by some virtual machines.

I'm using Proxmox as virtualization system (3 node) with 2 gluster
external server (san1 and san2).

I'm storing the .qcow2 file (virtual image of virtual server) to gluster.

The .qcow2 are sparse file.

These the configuration of the volume:
# gluster volume info glusterb1

Volume Name: glusterb1
Type: Replicate
Volume ID: 76106c18-e7eb-4f02-bd86-cb90d41d8960
Status: Started
Number of Bricks: 1 x 2 = 2
Transport-type: tcp
Brick1: san1:/glusterfs2/glusterb1
Brick2: san2:/glusterfs2/glusterb1

These the mountpoint on my Proxmox (3 node):
san1:/glusterb1 on /mnt/pve/glusterb1 type fuse.glusterfs

Everything seems to work correctly, but when I try to check the size
of the .qcow2 file on every gluster node, the dimensions do not match:

# ssh root at san1 du /glusterfs2/glusterb1/images/119 && ssh root at san2
du /glusterfs2/glusterb1/images/119
19124304        /glusterfs2/glusterb1/images/119
19124108        /glusterfs2/glusterb1/images/119

19124304 the file dimension on san1
19124108 the file dimension on san2

The volume are not in split-brain mode:
# gluster volume heal glusterb1 info split-brain
Gathering Heal info on volume glusterb1 has been successful

Brick san1:/glusterfs2/glusterb1
Number of entries: 0

Brick san2:/glusterfs2/glusterb1
Number of entries: 0

Is it possible that the two files on two nodes Gluster, have different
sizes ? or something is not working ?

sorry for my poor english
thank you very much

More information about the Gluster-users mailing list