NFS Troubleshooting

While in theory, NFS is supposed to be a great technology, my experience with it has been that it works well when the demand on it is not great. I don't like it for mission-critical purposes, in implementations that have long-standing NFS mounts, or those that have high throughput.

Too often NFS gets "stuck" and really messes up a server. Despite soft mounts claiming to take this away, we have seen time and time again NFS cause us pain. Df commands don't return properly, and the system behaves very strangely.

Trying to connect a UNIX server to the Microsoft version of NFS - "UNIX Services for Windows" is an utter nightmare.

So rule #1: don't use NFS if you don't have to. I'll take FTP transactions over NFS any day.

If you do get into NFS issues, here are a few commands I've kept in my archives on troubleshooting NFS (particularly on Solaris servers):

• Run "ps -ef | grep nfsd" a few times & see if cpu time is accumulating
• Run "ps -ef | grep mountd". If it is not running maybe run /usr/lib/nfs/mountd
• From a NFS client, run "showmount -e server" or "dfshares server" to see what shares look like from the client perspective
• From the client: "/usr/sbin/rpcinfo -u server nfs". This should give info on what versions of NFS are configured and waiting for connections. Also run "/usr/sbin/rpcinfo -u server mountd" for details
• Check "/etc/dfs/dfstab" (or equivalent such as "/etc/exports") to make sure they are configured properly

Last resort- reboot the server; sometimes this can clean things up.

- John