[Lxc-users] [NFS] nfs server

Rob Landley rob at landley.net
Tue Aug 9 15:44:23 UTC 2011


On 08/09/2011 08:27 AM, Kirill A. Shutemov wrote:
> On Tue, Aug 09, 2011 at 03:10:44PM +0200, Daniel Lezcano wrote:
>> On 08/09/2011 10:03 AM, Jeremy MAURO wrote:
>>> Hi everyone,
>>>
>>> Any updates about this request, because I am unable to setup a nfs 
>>> server on the container without a nfs-server on the main server.
>>> Once the nfs-kernel-server is setup on the main server, I am still 
>>> unable to mount the nfs exports:
>>> On the main server:
>>> [root at server]:/var/lib/lxc/nfsroot/rootfs/var/lib # showmount -e localhost
>>> Export list for localhost:
>>> [root at server]:/var/lib/lxc/nfsroot/rootfs/var/lib #
>>>
>>> On the container:
>>> [root at nfsroot]:~ # showmount -e localhost
>>> Export list for localhost:
>>> /var/nfsroot 10.10.0.0/16
>>> [root at nfsroot]:~ # showmount -e 10.10.200.31
>>> Export list for 10.10.200.31:
>>> /var/nfsroot 10.10.0.0/16
>>>
>>> And when trying to mount the export, the clients just hangs
>>> [root at client-server]:/ # mount 10.10.200.31:/var/nfsroot /mnt/
>>>
>>> Any idea?
>>
>> I think the nfs server is not yet supported by the kernel.
>> AFAIR Kirill had a patchset to be merged mainline but I don't know what
>> is the status.
> 
> AFAIK, upstream has no objections on this patchset, but they don't merge
> it. I don't know why.
> 
> Anyway, it's only a small part of work which need to be done to get nfs work
> properly in containers.
> 
> Rob, do you have any status update? I don't follow nfs maillist currently.
> 

P.S.  I believe the fies to make cifs and p9 work already went in, and
FUSE already did, but it's been a while and I'll have to retest.  (I
know I got 'em all to work, I _think_ all changes necessary to do so
went upstream.)  If you're not tied to NFS, you have several options.

(Getting NFSv4 to work was a crawling horror due to its horrible
overcomplicated design wanting to merge different mount points into the
same superblock without even using the --bind mount mechanism, make
callbacks to kernel threads and userspace with no obvious ownership
rules...  NFSv3 was a piece of cake in comparison, and I don't think I
ever got lockd to work properly there either.  Of course, other network
filesystems never needed it...)

Rob




More information about the lxc-users mailing list