----- Original Message -----> From: "shacky" <shacky83 at gmail.com>
> To: "Rajesh Joseph" <rjoseph at redhat.com>
> Cc: "Soumya Koduri" <skoduri at redhat.com>,
"gluster-users" <gluster-users at gluster.org>
> Sent: Thursday, August 20, 2015 3:42:02 PM
> Subject: Re: [Gluster-users] NFS mount
>
> 2015-08-20 12:03 GMT+02:00 Rajesh Joseph <rjoseph at redhat.com>:
>
> > AFAIK, you are not suppose to do NFS mount on the same node where
> > gluster NFS server is running due to NLM locking.
>
> I have three nodes which holds data and computing.
> I need to have data replicated on all three nodes for failover
> features, but every nodes needs to use its own local volume.
> With the Gluster native mount client this works, but I have very poor
> performance (5MB/sec!).
> The network speed is actually about 950 Mbit/sec, and if I copy a file
> locally I get about 300 MB/sec.
> So the performance problems are related on Gluster, and this is
> because I wish to try NFS on Gluster.
>
> What could you advise me?
> Thank you very much for your help!
>
One way is to have your storage node separate and mount them where you are doing
the compute.
The other way is to mount Gluster NFS with "-o nolock" option. AFAIK,
with nolock
the conflict from NLM will go away. But I need someone from NFS team to confirm
this.
Best Regards,
Rajesh