[Linux-HA] Fail-over NFS on RHEL4

[Greg Dickie]

We just hit this same problem the other day. I haven't tested it yet but
my approach is to change the location of that directory
to /var/lib/rpc_pipefs in /etc/idmapd.conf and /etc/modprobe.conf.dist
so its not under /var/lib/nfs any longer.

I'll post back if that works when its failed over....

Greg


On Fri, 2007-01-19 at 15:14 +0100, Ingvar Hagelund wrote:
> We want to set up a fail-over NFS service with a two-node heartbeat
> cluster on RHEL4 with shared storage on SAN. This should be quite
> strightforward. Looking at the docs, it seems a good idea to symlink
> /var/lib/nfs to some /path/to/failover-volume/nfs-state/, to make sure
> that NFS state follows the service. But on RHEL4, there are a kernel
> filesystem (sunrpc) mounted inside /var/lib/nfs.
> 
> # grep sunrpc /proc/mounts
> sunrpc /var/lib/nfs/rpc_pipefs rpc_pipefs rw 0 0
> 
> (And as a side note: Why!!?)
> 
> So, we thought, what about just symlinking the necessary files inside
> that directory, a subset of the following files:
> 
> etab export-lock rmtab statd/ state xtab
> 
> But which of these files (statd is a directory) do we need to keep NFS
> state across fail-over? All of them?
> 
> Regards,
> Ingvar Hagelund
> 
-- 
Greg Dickie
just a guy
Maximum Throughput