Problem

NFS V4.1

NFS V4.x

CFT HA

NFS SHARE

Resolution

About NFS 4.1 (or even higher level), the new features are supposed to get it more reliable and address some performance issues with parallel accesses (pNFS extension)

The recommendation would always be to use the latest version of NFS which is supposed to be more reliable and more performant.

At time of writing, we would even recommend NFS 4.2 if it is correctly supported by NFS server and kernel.

Yet, most often customers use quite old kernels so they have to ensure there is no known issue about the NFS client driver with their kernel version.

Also, we have some requirements on the mount on client side which are documented, but we don't have any recommendation for the configuration of the NFS server, because there are too many solution providers (linux NFS daemon nfsd, NetApp, Huawei and their in-house implementation, ...).
The documentation of the NFS server provider and its recommendations should be followed.

About the NFS architecture, CFT is not really aware of working with a network shared file system (it could be a local disk for multi-nodes on a single host for instance, a GPFS mount, etc...), as there are several nodes it knows it has to protect the file accesses with file locks, it also uses extensively file locks to know whether a node is up and running.

Only the kernel knows about the IP of the NFS server; whether there are different or not for the different hosts is only an issue for the NFS server which has to handle each client correctly knowing the different clients which access the server from different domain will access the same group of files.