[Linux-cluster] NFS Server Cluster

[Robert S Peterson]

On Tue, 2006-05-23 at 12:43 -0400, Mark Harrington wrote:
> Hi All,
> 
> I am playing around with a 2 node NFS Server cluster with RHCS/GFS 4 Update
> 3 and had a few questions. 
> 
> I could not find much documentation on active-active NFS clusters with RHCS
> 4 so I set up a pseudo active-active cluster as follows: 
>  - Created 2 RHCS services that export a single GFS volume with NFS.  
>  - Each service has a prioritized fail-over domain so if:
>     - Both nodes are up each node will export the GFS volume
>     - A node goes down the service (& VIP) will fail over to the other node.
> 
>  - 1/2 the NFS clients (Web server farm) will mount off one VIP and the 
>    other 1/2 off the other.
>  - The NFS clients are only permitted to mount 1 service
> 
> 1) We are still testing this configuration but has anyone else setup an NFS
> cluster like this?  Is there a better way?  Can you foresee any issues with
> this setup?

I don't see any issues.  As a matter of fact, I recommend exactly this
in the "NFS/GFS cookbook" document I posted a few minutes ago to
linux-cluster:

http://sources.redhat.com/cluster/doc/nfscookbook.pdf 

> 3) Do we need to put /var/lib/nfs on a GFS volume so both NFS servers can
> update the locks file?  Will there be any issues if both NFS servers try to
> update the locks file at the same time?

I don't think so, but I could be wrong.

> 3) Is it possible to do this with NFSv4?

I'm pretty sure you can.

> Thanks for any help,
> Mark

Bob Peterson
Red Hat Cluster Suite