This may be old news, but I discovered it while playing around with a F630 running 5.0.2.

The netapp filer itself follows symlinks. I guess that shouldn't be a surprise, but, technically, I don't think that a filer has to follow symlinks. In NFS, the CLIENT follows the symlink, not the server. The server simply passes to the client the contents of the symlink. So one could design a filer that does not itself follow symlinks and it would still be a correct NFS server. (CIFS is another story.)

Our F630 has two volumes: /vol/vol0 (the root volume) and /vol/vol1, which NFS clients mount on /h2.

Just for fun, I created this symlink in /vol/vol0

/vol/vol0/h2 -> /vol/vol1

So now on the netapp, /h2 -> /vol/vol1

I discovered I can now use /h2 in config files wherever I would ordinarily use /vol/vol1.

In /etc/exports I have successfully exported /vol/vol1 like this:

/h2 access=foo.bar.EDU

And this command works on NFS clients for mounting:

mount home2.Virginia.EDU:/h2 /h2

And it also works in the quotas file.

/h2/users/loginid user@/h2 100M 100K

Note that this trick only works where /vol/vol1 would be appropriate. For example I still have to use "quota on vol1", but I can change the volume name from vol1 to h2, to fix that up. (and then change my symlink to /vol/vol0/h2 -> /vol/h2)

An advantage to using this trick occurs to me. If you change a volume name, you may be forced to change quite a few config files. If you use a symlink in the config files, you can rename a volume and then just change the link.

Also, pathnames on clients, such as /h2/foo/bar will also work on the server. So then you don't have to always translate paths in your head from "client-style" to "server-style" when running a command on the server, such as "quota report /h2/foo/bar"

Disclaimer: I don't know if Network Appliance will guarantee this behavior in future releases.

Steve Losen scl@virginia.edu phone: 804-924-0640

University of Virginia ITC Unix Support