[tlug] NFS export issue

[Josh Glover <jmglov@example.com>]

This may well be the weirdest thing I've seen today:

I am NFS mounting /home/ on one box from a Solaris 8 box running NFS 
server. For some users, things are swell. For others, they do not have 
write perms to their own homespace. The problem stems from ownership 
problems, but I have been unable to solve them. Here is a sample of /home:

drwxrwxr-x    2 141      bin           512 Mar 16 10:35 jmglov
drwxrwxr-x   40 john     bin          3072 Mar 13 15:36 john
drwxrwxr-x    5 137      bin           512 Feb 21 08:38 judith
drwxrwxr-x    7 kellyw   bin           512 Dec 13 09:21 kellyw

Two of these users, john and kellyw, can happily use their homespace. 
The other two, jmglov and judith, cannot, as they do not own their 
homespace. The question is, why not? and where does the 141 come from? 
On the server, /home looks like:

drwxrwxr-x   2 jmglov   users        512 Mar 16 10:35 jmglov
drwxrwxr-x  40 john     users       3072 Mar 16 11:10 john
drwxrwxr-x   5 judith   users        512 Feb 21 08:38 judith
drwxrwxr-x   7 kellyw   users        512 Mar 16 11:11 kellyw

As one might expect, jmglov's homespace works fine on the server. I have 
tried restarting the NFS server, but that did nothing.

Back on the client, mount reports that /home is being mounted rw:

jabba:/home on /home type nfs (rw,hard,intr,addr=XXX.XXX.XXX.XXX)

(IP address censored to protect the guilty)

I cannot umount /home, as umount always reports that /home is busy, even 
when I stop the NFS server. Strange. I cannot force it, but I can umount 
-r and have it mounted readonly. Which buys me nothing.

So I am out of ideas, for now. What I have not done is restarted the 
client. I would rather not, as it is a high availability machine. And I 
should not have to.

Any suggestions?


-- 
Josh Glover <jmglov@example.com>

Computer Science Undergrad
"No segfault, no problem!"