Gluster users - May 2015 - Why have problems gluster and kernel nfs

Hi all,
I Install gluster in one node,disable gluster nfs and export nfs by kernel nfs
server,they are in the same linux system;
another linux client mount nfs path,when dd file on the nfs mount
point;Sometimes the io return error;

Who knows what is the reason, if there is a corresponding kernel patch.

thanks!

best regards
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://www.gluster.org/pipermail/gluster-users/attachments/20150509/006dad26/attachment.html>

What are you trying to export? The gluster volume?  You should do that with the 
gluster nfs module or the newer NFS ganesha module, not via a kernel NFS export.

Regards,

Ric


On 05/09/2015 08:00 AM, ?? wrote:
> Hi all,
> I Install gluster in one node,disable gluster nfs and export nfs by kernel
nfs
> server,they are in the same linux system;
> another linux client mount nfs path,when dd file on the nfs mount 
> point;Sometimes the io return error;
>
> Who knows what isthe reason,if there is a corresponding kernelpatch.
>
> thanks!
>
> best regards
>
>
>
>
>
> _______________________________________________
> Gluster-users mailing list
> Gluster-users at gluster.org
> http://www.gluster.org/mailman/listinfo/gluster-users

On Sat, May 09, 2015 at 01:00:31PM +0800, ?? wrote:
> Hi all,
> I Install gluster in one node,disable gluster nfs and export nfs by kernel
nfs server,they are in the same linux system;
> another linux client mount nfs path,when dd file on the nfs mount
point;Sometimes the io return error;
It depends a little on how you have set this up. You should only write
to the bricks of the Gluster volume through a Gluster client. Gluster
clients are the FUSE mount and the Gluster/NFS server. You should not
export the bricks directly over the Linux kernel NFS-server.

Exporting a FUSE mountpoint over the Linux kernel NFS-server is not
something that is commonly done or tested. At least in the past, the
fuse kernel-module did not support the NFS-server sufficiently. It seems
that newer versions of the kernel provide all the required features, but
there are still some things to take into consideration:

    http://sourceforge.net/p/fuse/fuse/ci/master/tree/README.NFS


The advised way to export Gluster volumes over NFS, is by using the
Gluster/NFS server or with NFS-Ganesha. Gluster/NFS is enabled by
default (you already mentioned you disabled it, so I assume you know
about it). NFS-Ganeshe is an other userspace NFS-server that supports
many more features than Gluster/NFS:

    http://nfs-ganesha.github.io/
   
http://glusterants.wordpress.com/2014/09/03/glusterfs-and-nfs-ganesha-integration/
    https://github.com/nfs-ganesha/nfs-ganesha/wiki/GLUSTER
   
http://www.gluster.org/community/documentation/index.php/Features/HA_for_ganesha

NFS-Ganesha version 2.2 has been released about a week ago. Fedora 22
(when released) will contain the packaged version. The CentOS Storage
SIG will likely provide this version shortly too.

> Who knows what is the reason, if there is a corresponding kernel patch.
Without more details it will be very difficult to help you with this. My
strong advise is to use Gluster/NFS or NFS-Ganesha instead as that is
where the Gluster developers focus on.

Good luck!
Niels
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 181 bytes
Desc: not available
URL:
<http://www.gluster.org/pipermail/gluster-users/attachments/20150510/c000ee70/attachment.sig>