[Gluster-users] Is anyone using sharding?

[Krutika Dhananjay]

Hi Elvind, 

At the moment, the focus is on making sharding work with gluster-as-VM-store use case (which is a typical large-files-with-single-writer workload). 
After this, I do plan to make it work for general-purpose use cases too. 

With respect to this specific issue, could you set performance.strict-write-ordering to "on" and try your test again? 

Here's what you need to do: 
#gluster volume set <VOLNAME> performance.strict-write-ordering on 

-Krutika 

----- Original Message -----

> From: "Eivind Sarto" <eivindsarto at gmail.com>
> To: Gluster-users at gluster.org
> Sent: Tuesday, September 15, 2015 3:35:11 AM
> Subject: [Gluster-users] Is anyone using sharding?

> I noticed that several sharding bugs relating to file size got closed because
> 3.7.4 got released. However, file sizes are still messed up in 3.7.4. I know
> this is a beta feature, but you cannot even write a single file without
> getting the wrong file size.

> Neither buffered or direct IO works:

> # dd if=/dev/zero of=/voltest/dd.buffered bs=1024k count=1024
> 1024+0 records in
> 1024+0 records out
> 1073741824 bytes (1.1 GB) copied, 1.66046 s, 647 MB/s
> # dd if=/dev/zero of=/voltest/dd.direct oflag=direct bs=1024k count=1024
> 1024+0 records in
> 1024+0 records out
> 1073741824 bytes (1.1 GB) copied, 1.91092 s, 562 MB/s
> root at science08:~/test# ls -l /voltest
> total 2104384
> -rw-r--r-- 1 root root 1075052544 Sep 14 14:31 dd.direct
> -rw-r--r-- 1 root root 1074266112 Sep 14 14:18 dd.buffered

> # gluster vol info voltest
> Volume Name: voltest
> Type: Distribute
> Volume ID: 849e2d6b-faaa-43d7-b06e-3c6d3057dff3
> Status: Started
> Number of Bricks: 4
> Transport-type: tcp
> Bricks:
> Brick1: sc03:/brick_ssd_0/gv
> Brick2: sc04:/brick_ssd_0/gv
> Brick3: sc03:/brick_ssd_1/gv
> Brick4: sc04:/brick_ssd_1/gv
> Options Reconfigured:
> features.shard: on
> performance.readdir-ahead: on
> root at science03:~# gluster --version
> glusterfs 3.7.4 built on Sep 1 2015 12:08:58

> _______________________________________________
> Gluster-users mailing list
> Gluster-users at gluster.org
> http://www.gluster.org/mailman/listinfo/gluster-users
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.gluster.org/pipermail/gluster-users/attachments/20150914/325a361c/attachment.html>