[Linux-cluster] Purpose of fencing devices

["Sævaldur Arnar Gunnarsson [Hugsmiðjan]"]

Could someone explain to me the purpose of the fencing hardware in a 
cluster with a shared storage resource.

When one of the cluster member goes down all access to the shared volume 
(GFS) is closed off.
No other cluster member can read or write to the volume until the failed 
node comes back up.

Are fencing devices used to close off the access the dead node has on 
the filesystem so the other nodes can access (read/write) the fileystem 
as usual ?
-- 
Sævaldur Gunnarsson :: Hugsmiðjan