On Fri, 19 May 2006, Daniel Qarras wrote:
> However, when I go out to a different network where those directories
> are not available, I would like autofs to figure that out and not to
> try mount them and cause jams. What should the script output if those
> mounts are available?
If I remember correctly, if the programmatic map prints nothing and exits
with a nonzero exit code, autofs will realize it has failed and give up. So
now, how does the script realize that the 10.0.0.x IP address assigned to
eth0 is not *your* 10.0.0.x, but a customer's net or an Internet cafe
somewhere? Perhaps a less obvious choice of subnet could be helpful; this
is what I do on my own net.
But I'm wondering, what's in server:/home or server:/work? The ideal is to
set up the laptop so the NFS mounted directories are not touched without
voluntary user action, upon which autofs swings into action, and if the
user forgets that he's not at headquarters, he deserves the 30 second
timeout (unless the programmatic map script heads off NFS mounting). It
isn't going to fly, if the user's home directory is in server:/home/$USER.
On my own laptop the homedir is local, with backups to the server (and from
there to removeable media), initiated manually.
If ever you might expand to have two or more NFS servers, it's important to
use separate subtrees for each, e.g.
/var/autofs/server1/home
/var/autofs/server1/work
/var/autofs/server2/home
/var/autofs/server2/work
The reason is like this: Suppose the mount points were all jumbled
together, e.g.
/var/autofs/home1
/var/autofs/home2
/var/autofs/work1
/var/autofs/work2
(UCLA-Mathnet actually made this mistake in the distant past.) Suppose one
server is down and the other isn't, and a user hosted on the working server
tries to log in. Let's say his homedir is /var/autofs/home2/luser.
Several "pwd" operations can't be avoided, and it needs to stat all
memebers of /var/autofs to find out which name goes with the inode
which is the parent of the homedir. Frequently it will stat a mount point
from the dead server before getting the right one, and suffer a 30
second timeout, several times. Hiss, boo.
James F. Carter Voice 310 825 2897 FAX 310 206 6673
UCLA-Mathnet; 6115 MSA; 405 Hilgard Ave.; Los Angeles, CA, USA 90095-1555
Email: [EMAIL PROTECTED] http://www.math.ucla.edu/~jimc (q.v. for PGP key)
_______________________________________________
autofs mailing list
[email protected]
http://linux.kernel.org/mailman/listinfo/autofs
