CentOS - Jul 2010 - redundant networked secure file system recommendation

Hi all,

We are currently running a NFS-based server centric setup. I would
like to set up something where I can easily have more than one
redundant server, security/authentication (this part seems a little
flaky with NFS, at least did several years ago), with the capability
to easily add/remove servers as necessary, take redundant servers down
for maintenance, etc. Total volume we expect to run on the server side
will be somewhere between 10-30 TB. The servers will most likely be
CentOS machines, the clients mostly Linux machines with some Macs and
possibly Windows (the latter part not that important).

Any insight, thoughts and recommendations will be much appreciated.

Thanks.

Boris.

Boris Epstein <borepstein at gmail.com> wrote:
> We are currently running a NFS-based server centric setup. I would
> like to set up something where I can easily have more than one
> redundant server, security/authentication
Have you considered an NFS cluster based on pacemaker/openais/corosync?
See <http://www.clusterlabs.org/>.  One possible config is an
active/passive
cluster using NFS on top of DRBD.  I've never clustered samba, but it
looks like others have done it:
<http://dev.gentoo.org/~wschlich/src/sys-cluster/heartbeat-scripts-4/ocf-ra/samba>

Google will show you various problems with active/active clusters related to
both GFS2 and samba.  Don't try to have your NFS servers also be NFS
clients.

For reasonable security/authentication, you'd probably want to integrate
with kerberos.

Devin

Am 19.07.2010 18:02, schrieb Boris Epstein:
> Hi all,
>
> We are currently running a NFS-based server centric setup. I would
> like to set up something where I can easily have more than one
> redundant server, security/authentication (this part seems a little
> flaky with NFS, at least did several years ago), with the capability
> to easily add/remove servers as necessary, take redundant servers down
> for maintenance, etc. Total volume we expect to run on the server side
> will be somewhere between 10-30 TB. The servers will most likely be
> CentOS machines, the clients mostly Linux machines with some Macs and
> possibly Windows (the latter part not that important).
>
> Any insight, thoughts and recommendations will be much appreciated.
>
> Thanks.
>
> Boris.
>    
I think about buying a HP LeftHand SAN wich is an iSCSI cluster sytem. 
One could use that as backend for two failover hosts serving GFS via NFS.

+C