Glusterfs free space with df

langner · 09-25-2010, 02:08 AM

Irrelevant of the amoung of data in the volume, for some reason I always see zero usage of my glusterfs volume using df:

Code:

me@machine:~$ df -h | grep gluster
glusterfs#/etc/glusterfs/glusterfs.vol
                      8.9T     0  8.9T   0% /glusterfs

I am using the 'distribute' mode. Is this normal, or can it be mu configuration? Has anyone else had this problem?

quanta · 11-01-2010, 11:25 AM

Please give us the output of `gluster volume info` command? Did you try to test by listing the content of mount dir?

langner · 11-02-2010, 01:55 AM

Quote:

Originally Posted by quanta

Please give us the output of `gluster volume info` command? Did you try to test by listing the content of mount dir?

I tested before, yes. And 'du' for example shows more than 1T of data:

Code:

me@machine# du -h --max-depth=1
80K     ./dir1
80K     ./dir2
80K     ./dir3
46G     ./dir4
1.1T    ./dir5
1.1T    .

While 'df -h' still gives a 0% reading, the inodes are counted up:

Code:

me@machine# df -ih | grep gluster
glusterfs#/etc/glusterfs/glusterfs.vol
                        1.2G    1.4M    1.2G    1% /glusterfs

I now realize that I have compiled glusterfs 2.0.9 and not 3.0.x, and I do not have the 'gluster' command. Is that new to 3.0.x? I suppose I should compile the new version to se if this behavior persists.

quanta · 11-02-2010, 02:43 AM

Quote:

Originally Posted by langner

I now realize that I have compiled glusterfs 2.0.9 and not 3.0.x, and I do not have the 'gluster' command. Is that new to 3.0.x? I suppose I should compile the new version to se if this behavior persists.

I've searched glusterfs available in CentOS repo and found the below:

Code:

glusterfs-client.i386 : GlusterFS Client
glusterfs-common.i386 : GlusterFS common files for both the client and the server
glusterfs-devel.i386 : GlusterFS Development Libraries
glusterfs-server.i386 : GlusterFS Server

Those have version 2.0.x. The package name is different from homepage. Therefore, I suppose that there are many changes from 2.0.x to the latest version - 3.1. In the 2.0.x there are only 3 commands: glusterfs, glusterfsd, glusterfs-volgen while the latest version added: gluster, glusterd, glusterfs-defrag. I suggest you install the latest version and let me know if it works.

langner · 11-03-2010, 04:31 AM

Quote:

Originally Posted by quanta

I've searched glusterfs available in CentOS repo and found the below:

Code:

glusterfs-client.i386 : GlusterFS Client
glusterfs-common.i386 : GlusterFS common files for both the client and the server
glusterfs-devel.i386 : GlusterFS Development Libraries
glusterfs-server.i386 : GlusterFS Server

Those have version 2.0.x. The package name is different from homepage. Therefore, I suppose that there are many changes from 2.0.x to the latest version - 3.1. In the 2.0.x there are only 3 commands: glusterfs, glusterfsd, glusterfs-volgen while the latest version added: gluster, glusterd, glusterfs-defrag. I suggest you install the latest version and let me know if it works.

Exactly, I am in fact using CentOS. I will try to compile and use a newer version.

langner · 11-09-2010, 06:21 AM

I compiled version 3.1.0 and saw exactly the same behavior. I haven't been able to fix it, but I now know it is caused by the features/quota translator. After turning that off the disk usage is reported correctly. Googling shows others that have found this problem, but no solution besides disable quota.