Re: Gluster quota showing wrong space utilization (showing 16384.0 PB)

[Vijaikumar M <vmallika@xxxxxxxxxx>]

Hi Omar,

We will run this test-case in our lab machines.
Meanwhile can you provide the brick logs and execute the attached script for each brick-path, and provide the output from the script

Thanks,
Vijay

On Tuesday 03 February 2015 08:06 AM, Omkar Kulkarni wrote:

Hi Vijay,

I was recreate the issue. 

1) Create a replicated volume with 2 CentOS 6.5 machines 

i)gluster volume create homedir replica 2 transport tcp node1:/brick1/homedir node2:/brick1/homedir

gluster  volume set homedir network.ping-timeout 10

gluster  volume set homedir performance.cache-size 1GB

gluster  volume set homedir nfs.rpc-auth-allow 10.99.30.20,10.99.30.21

gluster  volume set homedir auth.allow 10.99.30.20,10.99.30.21

gluster  volume set homedir features.quota on

gluster  volume set homedir features.quota-timeout 0

gluster  volume set homedir features.quota-deem-statfs on

ii) setup quotas on around 900 folders

node1:~# gluster volume quota homedirlist |wc -l

903

iii) Created a rsync to create/delete files with different names  in 20 different folders and ran every 2 minutes

iv) in Around 2 days some directories showed PB in files used

On Fri, Jan 23, 2015 at 1:01 AM, Vijaikumar M <vmallika@xxxxxxxxxx> wrote:

Hi Omar,

If the issue is happening consistently, can we get a re-creatable test-case for this same?

Thanks,
Vijay

On Wednesday 21 January 2015 02:47 AM, Omkar Kulkarni wrote:

Hi Guys,

I am using gluster 3.5.2 and for couple of directories I am getting used space as 16384.0 PB

 node1:~# gluster volume quota homedir list |grep PB

/storage/home1                          3.0GB       90%   16384.0PB   3.0GB

/storage/home2                             1.0GB       90%   16384.0PB   1.0GB

DU on actual directory

node1:~#  du -hs /data/storage/home1/

368K    /data/storage/home1/

node1:~#  du -hs /data/storage/home2

546K    /data/storage/home2

I am using replicated volume (replica count 2) and mounted volume as nfs. Individual bricks are formatted as ext4. If I remove quotas and restart replicated volume and then re-add quotas then gluster quota shows correct value (instead of 16384.0 PB) but once file transfer starts and files are added/deleted again these directories start showing incorrect values. 

Just wanted to check if this is a bug in gluster quota and/or if there is any way to mitigate this?

I do see a similar bug but for 3.4

https://bugzilla.redhat.com/show_bug.cgi?id=1061068

Thanks

 

 

_______________________________________________
Gluster-users mailing list
Gluster-users@xxxxxxxxxxx
http://www.gluster.org/mailman/listinfo/gluster-users

Attachment: quota-verify.gz
Description: application/gzip

_______________________________________________
Gluster-users mailing list
Gluster-users@xxxxxxxxxxx
http://www.gluster.org/mailman/listinfo/gluster-users