[Gluster-devel] nfs over glusterfs/fuse

[Dale Dude]

Sent offlist. Sorry Brent.

-------- Original Message --------
Subject: 	Re: [Gluster-devel] nfs over glusterfs/fuse
Date: 	Thu, 14 Jun 2007 17:52:46 -0400
From: 	Dale Dude <dale at oc3networks.com>
To: 	Brent A Nelson <brent at phys.ufl.edu>
References: 	<4671B0DF.4020001 at oc3networks.com> 
<Pine.LNX.4.64.0706141721330.11166 at atma.phys.ufl.edu>



I didnt run it long enough to notice if memory grew. Fuse-dev says 
nfs-kernel-server is depricated and suggested unfs3 which also didnt 
work. I was hoping its 'removable' option would help, but no. 
nfs-user-server had an interesting option of re-exporting other nfs 
mounts, but damn if I couldnt get the client to even mount using it 
without permission denied no matter what I did.

Odd that the userspace nfsd's have the same issue. Samba is the only 
thing that works. I cant put glusterfs on these boxes either. Hurts.

Brent A Nelson wrote:
> I get the same and a host of other symptoms (I've only tried 
> nfs-kernel-server); it's not really usable.  Do you also see 
> significant growth in the memory consumed by the glusterfs process 
> that is being reexported?
>
> I wonder if the new inode-based release tomorrow will fix the NFS 
> reexport issue...
>
> On Thu, 14 Jun 2007, Dale Dude wrote:
>
>> This may be more for fuse but I thought Id post here first since it 
>> seems someone is/has done this.
>>
>> Ubuntu Feisty Kernel 2.6.20-16. Fuse 2.6.5.
>>
>> I tried these userspace servers: unfs3, nfs-user-server
>> I tried the nfs-kernel-server
>>
>> With each of the above nfs servers I get spuradic Input/Output errors 
>> when I try to 'ls' or 'find' on some directories on the nfs client. I 
>> tried fsid option without success. Im using samba/cifs now which 
>> works but proving to be way to slow.
>>
>> Any advice would greatly appreciated.
>>
>> Regards,
>> Dale
>