Gluster users - Nov 2015 - Can I create a gluster volume from NFS mounted partitions?

Hi team

I have a special case here: I am trying to create a glusterFS volume from NFS
mounts offered by third party.
But it fails. The long story has as follows:

I have two 50 GB NFS volumes offered from two data centres:
In DC1:
sbox1.mydomain:/q_files
IN DC2:
sbox2.mydomain:/q_files

I dont have even ssh access to those sboxes so I cannot install or do anything,
but I am allowed to mount their directories with r/w over tcp.

I have setup  a pair of VM nodes running glusterFs. One node on DC1, one on DC2.
On each of MY nodes, I mount the offered NFS partitions:

The node in DC1  mybox1.mydomain.corp mounts the partition from DC1 machine:
mount -t nfs sbox1.mydomain:/q_files   /nfs

The node in DC2 mybox2.mydomain.corp mounts the partition from DC2 machine:
mount -t nfs sbox2.mydomain:/q_files   /nfs

I do the peering normally...
root@ mybox1:~ # gluster peer probe mybox2.mydomain
peer probe: success. Host mybox2.mydomain port 24007 already in peer list

ok so far - no problem.

Then I try to create a glusterFS volume from those mounts:
root@ mybox1:~ # gluster volume create mynfs replica 2 mybox1.mydomain:/nfs
mybox2.mydomain:/nfs force
volume create: contentstore_nfs: failed: Glusterfs is not supported on brick:
mybox1.mydomain:/nfs.
Setting extended attributes failed, reason: Operation not supported.

I get this message: Operation not supported.

The question is obvious:
Does GlusterFS supports such an operation? In other words, can I create a
glusterFS volume from bricks created from NFS mounts?
If it is feasible, can you please specify what am I doing wrong?

Best Regards
mike
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://www.gluster.org/pipermail/gluster-users/attachments/20151127/09960c3f/attachment.html>

-Atin
Sent from one plus one
On Nov 27, 2015 11:01 PM, "Mountrakis, Michael" <michael.mountrakis
at sap.com>
wrote:
>
> Hi team
>
>
>
> I have a special case here: I am trying to create a glusterFS volume from
NFS mounts offered by third party.
>
> But it fails. The long story has as follows:
>
>
>
> I have two 50 GB NFS volumes offered from two data centres:
>
> In DC1:
>
> sbox1.mydomain:/q_files
>
> IN DC2:
>
> sbox2.mydomain:/q_files
>
>
>
> I dont have even ssh access to those sboxes so I cannot install or do
anything, but I am allowed to mount their directories with r/w over
tcp.
>
>
>
> I have setup  a pair of VM nodes running glusterFs. One node on DC1, one
on DC2.
>
> On each of MY nodes, I mount the offered NFS partitions:
>
>
>
> The node in DC1  mybox1.mydomain.corp mounts the partition from DC1
machine:
>
> mount -t nfs sbox1.mydomain:/q_files   /nfs
>
>
>
> The node in DC2 mybox2.mydomain.corp mounts the partition from DC2
machine:
>
> mount -t nfs sbox2.mydomain:/q_files   /nfs
>
>
>
> I do the peering normally?
>
> root@ mybox1:~ # gluster peer probe mybox2.mydomain
>
> peer probe: success. Host mybox2.mydomain port 24007 already in peer list
>
>
>
> ok so far ? no problem.
>
>
>
> Then I try to create a glusterFS volume from those mounts:
>
> root@ mybox1:~ # gluster volume create mynfs replica 2
mybox1.mydomain:/nfs mybox2.mydomain:/nfs force
>
> volume create: contentstore_nfs: failed: Glusterfs is not supported on
brick: mybox1.mydomain:/nfs.
>
> Setting extended attributes failed, reason: Operation not supported.
>
>
>
> I get this message: Operation not supported.
Underlying file system of gluster bricks need extended attribute support
and hence this failure.
>
>
>
> The question is obvious:
>
> Does GlusterFS supports such an operation? In other words, can I create a
glusterFS volume from bricks created from NFS mounts?
>
> If it is feasible, can you please specify what am I doing wrong?
>
>
>
> Best Regards
>
> mike
>
>
> _______________________________________________
> Gluster-users mailing list
> Gluster-users at gluster.org
> http://www.gluster.org/mailman/listinfo/gluster-users
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://www.gluster.org/pipermail/gluster-users/attachments/20151127/bebb89d1/attachment.html>

On Fri, Nov 27, 2015 at 05:25:32PM +0000, Mountrakis, Michael
wrote:
> Hi team
> 
> I have a special case here: I am trying to create a glusterFS volume from
NFS mounts offered by third party.
> But it fails. The long story has as follows:
> 
> I have two 50 GB NFS volumes offered from two data centres:
> In DC1:
> sbox1.mydomain:/q_files
> IN DC2:
> sbox2.mydomain:/q_files
> 
> I dont have even ssh access to those sboxes so I cannot install or do
anything, but I am allowed to mount their directories with r/w over tcp.
> 
> I have setup  a pair of VM nodes running glusterFs. One node on DC1, one on
DC2.
> On each of MY nodes, I mount the offered NFS partitions:
> 
> The node in DC1  mybox1.mydomain.corp mounts the partition from DC1
machine:
> mount -t nfs sbox1.mydomain:/q_files   /nfs
> 
> The node in DC2 mybox2.mydomain.corp mounts the partition from DC2 machine:
> mount -t nfs sbox2.mydomain:/q_files   /nfs
> 
> I do the peering normally...
> root@ mybox1:~ # gluster peer probe mybox2.mydomain
> peer probe: success. Host mybox2.mydomain port 24007 already in peer list
> 
> ok so far - no problem.
> 
> Then I try to create a glusterFS volume from those mounts:
> root@ mybox1:~ # gluster volume create mynfs replica 2 mybox1.mydomain:/nfs
mybox2.mydomain:/nfs force
> volume create: contentstore_nfs: failed: Glusterfs is not supported on
brick: mybox1.mydomain:/nfs.
> Setting extended attributes failed, reason: Operation not supported.
> 
> I get this message: Operation not supported.
> 
> The question is obvious:
> Does GlusterFS supports such an operation? In other words, can I create a
glusterFS volume from bricks created from NFS mounts?
> If it is feasible, can you please specify what am I doing wrong?
Gluster requires support for extended attributes on the filesystem. NFS
does not support this (yet). At the moment, it is not possible to use
NFS exports as Gluster bricks.

Once the NFS specification for supporting extended attributes is in
place, NFS-clients and NFS-servers need to start implementing this. Only
after that, it will be possible to use NFS mountpoints as Gluster
bricks. see https://tools.ietf.org/html/draft-ietf-nfsv4-xattrs-01 for
more technical details on the (current in draft) specification.

HTH,
Niels
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 819 bytes
Desc: not available
URL:
<http://www.gluster.org/pipermail/gluster-users/attachments/20151127/c30249ac/attachment.sig>