Redhat:NFS Limits

Statelessness

There is no state associated with NFS which means there's no need to log in separately to the service.

When an NFS share is mounted, the request is sent to the server and if the request is valid and handle is sent back.

Symbolic Links

Any symbolic links in NFS shares can cause problems. Symbolic links are interpreted as if they are on the local system. If the mount point does not exist on the client system then it will fail.

Root Squash

Root squash prevent the root user on the local system having root privileges in the NFS shares. This can be disabled with teh 'no_root_squash' option.

NFS Hangs

Due to the statelessness of NFs, clients will wait for the server, even if the server has gone offline. This causes any processes using NFS shares to hang. At this point is is almost impossible to unmount the NFS share as it will be busy - a reboot of the system may be the only way to fix it.

There are several things that can reduce the problem:

Ensure reliability of NFS servers
Avoid mounting multiple NFS servers at once
Only mount required NFS shares
Use the sync options to limit file loss
Avoid using a mission critical system as an NFS client
Keep NFS shares out of search paths
Keep NFS shares out or / directory

Inverse DNS Pointers

The NFS server uses DNS lookup to get the IP of the NFS server. if it can't find the server IP it will deny access. It will also add a log message.

File Locking

It is possible to mount the same NFS share multiple times on the same client. This can lead to file locking if two users are accessing the same file at the same time. This is addressed by the file-locking daemon service.