[Linux-cluster] GFS+EXT3 via NFS?

[Lon Hohberger]

On Fri, 2007-01-12 at 10:09 -0500, Kovacs, Corey J. wrote:
> Lon, thanks. I could manually type the cluster.conf in but it would likely be
> riddled with typos :)
> 
> Suffice it to say that the exports are managed by the cluster which I think
> is the problem in our particular case as we have mixed GFS/EXT3 filesystems
> being exported from the same services. 

I wouldn't think this should matter, but it does depend on how they're
configured.  It should look something like:

  <service>
    <fs>
      <nfsexport>
        <nfsclient/>
        ...
      </nfsexport>
    </fs>
    <clusterfs>
      <nfsexport>
        <nfsclient/>
        ...
      </nfsexport>
    </clusterfs>
    <ip/>
  </service>


> That being said, what will the effect be of seperating the services by
> filesystem type if a service exporting EXT3 fails over to a node exporting
> non-cluster-managed GFS exports? Will the mechanics of moving a
> cluster-managed export to a node with non-managed exports collide?

They shouldn't - as long as the exports don't overlap.

-- Lon