Gluster users - Aug 2017 - [Gluster-devel] High load on glusterfs!!

On Thu, Aug 17, 2017 at 12:03:02PM +0530, ABHISHEK PALIWAL
wrote:
> Hi Team,
> 
> I have an query regarding the usage of ACL on gluster volume. I have
> noticed that when we use normal gluster volume (without ACL) CPU load is
> low, but when we apply the ACL on gluster volume which internally uses Fuse
> ACL, CPU load gets increase about 6x times.
> 
> Could you please let me know is this expected or we can do some other
> configuration to reduce this type of overhead on gluster volume with ACLs.
> 
> For more clarification we are using kernel NFS for exporting the gluster
> volume.
Exporting Gluster volumes over FUSE and kernel NFS is not something we
suggest and test. There are (or at least were) certain limitations in
FUSE that prevented good support for this.

Please use NFS-Ganesha instead, that is the NFS server we actively
develop with. Gluster/NFS is still available too, but is only receiving
the occasional fixes and is only suggested for legacy users that did not
move to NFS-Ganesha yet.

HTH,
Niels
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 801 bytes
Desc: not available
URL:
<http://lists.gluster.org/pipermail/gluster-users/attachments/20170830/f05efd7b/attachment.sig>

What is Gluster/NFS and how can we use this.

On Wed, Aug 30, 2017 at 1:24 PM, Niels de Vos <ndevos at redhat.com>
wrote:
> On Thu, Aug 17, 2017 at 12:03:02PM +0530, ABHISHEK PALIWAL wrote:
> > Hi Team,
> >
> > I have an query regarding the usage of ACL on gluster volume. I have
> > noticed that when we use normal gluster volume (without ACL) CPU load
is
> > low, but when we apply the ACL on gluster volume which internally uses
> Fuse
> > ACL, CPU load gets increase about 6x times.
> >
> > Could you please let me know is this expected or we can do some other
> > configuration to reduce this type of overhead on gluster volume with
> ACLs.
> >
> > For more clarification we are using kernel NFS for exporting the
gluster
> > volume.
>
> Exporting Gluster volumes over FUSE and kernel NFS is not something we
> suggest and test. There are (or at least were) certain limitations in
> FUSE that prevented good support for this.
>
> Please use NFS-Ganesha instead, that is the NFS server we actively
> develop with. Gluster/NFS is still available too, but is only receiving
> the occasional fixes and is only suggested for legacy users that did not
> move to NFS-Ganesha yet.
>
> HTH,
> Niels
>


-- 




Regards
Abhishek Paliwal
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.gluster.org/pipermail/gluster-users/attachments/20170830/a51783f8/attachment.html>

On Wed, Aug 30, 2017 at 01:52:59PM +0530, ABHISHEK PALIWAL
wrote:
> What is Gluster/NFS and how can we use this.
Gluster/NFS (or gNFS) is the NFS-server that comes with GlusterFS. It is
a NFSv3 server and can only be used to export Gluster volumes.

You can enable it:
 - install the glusterfs-gnfs RPM (glusterfs >= 3.11)
 - the glusterfs-server RPM might contain the NFS-server (glusterfs < 3.11)
 - build with "./configure --enable-gnfs"
 - enable per volume with: gluster volume set $VOLUME nfs.disable false
 - logs are in /var/log/gluster/nfs.log

But really, NFS-Ganesha is the recommendation. It has many more features
and will receive regular updates for improvements.

Niels

> 
> On Wed, Aug 30, 2017 at 1:24 PM, Niels de Vos <ndevos at redhat.com>
wrote:
> 
> > On Thu, Aug 17, 2017 at 12:03:02PM +0530, ABHISHEK PALIWAL wrote:
> > > Hi Team,
> > >
> > > I have an query regarding the usage of ACL on gluster volume. I
have
> > > noticed that when we use normal gluster volume (without ACL) CPU
load is
> > > low, but when we apply the ACL on gluster volume which internally
uses
> > Fuse
> > > ACL, CPU load gets increase about 6x times.
> > >
> > > Could you please let me know is this expected or we can do some
other
> > > configuration to reduce this type of overhead on gluster volume
with
> > ACLs.
> > >
> > > For more clarification we are using kernel NFS for exporting the
gluster
> > > volume.
> >
> > Exporting Gluster volumes over FUSE and kernel NFS is not something we
> > suggest and test. There are (or at least were) certain limitations in
> > FUSE that prevented good support for this.
> >
> > Please use NFS-Ganesha instead, that is the NFS server we actively
> > develop with. Gluster/NFS is still available too, but is only
receiving
> > the occasional fixes and is only suggested for legacy users that did
not
> > move to NFS-Ganesha yet.
> >
> > HTH,
> > Niels
> >
> 
> 
> 
> -- 
> 
> 
> 
> 
> Regards
> Abhishek Paliwal