homedirs and NFS

[mgais]

Hello,

we have several HP-UX 11.00 servers, where users have their homedirs mounted via NFS
(it's a Sun solaris 8 server) in this way:

daisy:/users /users nfs rw,bg,hard,intr,nosuid,nodevs 0 0

Now the problem is that sometimes when there is a network problem and the link goes down (and up again), the syslog reports something like this:

automountd[541]: mount_get_devid: server daisy not responding for /users
automountd[541]: server daisy:/users is down and can not get the device id! errno= 238

If a new user is logging in, it happens that the login process hangs at last line of the profile script, just before the usual "trap 123", but with a couple of Ctrl-C she can get your shell (after apparently the whole environment has been set correctly), and she can also access her homedir.

I suppose that the status of the NFS link is not fully ok, but I don't know where to investigate.
Can you explain me WHAT is going wrong during the login process and the possible reasons why it hangs? If you have any suggestions on how the filesystem should be mounted (or exported from Sun side), I would appreciate it as well.

Thanks

[Robert-Jan Goossens]

Hi Massimo,

Could you give the full command on Solaris

# share ?

Robert-Jan.

[Robert-Jan Goossens]

Just checked sun solve, you can use the soft option when mounting the nfs mount, but you have to export (share) the mount with the same option share -F nfs -o bg,soft etc

could you check your solaris server /etc/dfstab what sort of options you have configured.

Regards,

Robert-Jan.

Ps. there are same known problems using the hard option.

[mgais]

On the Sun fileserver there is no special option for sharing, so the /etc/dfs/dfstab is simply:

share -F nfs -s "Home Dirs" -o rw=host1:host2:host3

where hostN is the FQDN for each server using that filesystem as homedir (there are about 20 hosts listed).

[mgais]

Whoops! Sorry, I missed last arg in the dfstab:

share -F nfs -d "Home Dirs" -o rw=host1:host2:host3 /users

[monasingh_1]

that last option should not be required one for the problem you described.

it is the delay / clog in the netowrk as you mentioned...

[Russell Campbell]

Are you NFS mounting /var/mail? Check the users HISTFILE in there .profile. Some times it will hang on that. Try Changing it to

export HISTFILE=/tmp/${LOGNAME}.sh_history

[mgais]

Nope, I do not NFS mount /var for users, and they are not using .profile, just the default /etc/profile.

Anyway, I'd be interested in reading what are the problems with the "hard" option for mount. Thanks

[Robert-Jan Goossens]

Hi Massimo,

well all problem with hard mount should have solved with solaris 8.

This morning i've created some hard nfs mount from my workstation to a dev hpux box -> pulled the networkplug for a minute pluged it in again and had no problems.

used next commands
db001:/# mount -o rw,hard,bg jr0001:/opt /mnt

the /etc/mnttab look like this
jr0001:/opt /mnt nfs rsize=32768,wsize=32768,NFSv3 0 0 1045843817

Hope it helps.

Robert-Jan.

Ps. if you like I could take a look at my patches (ws sol 8 and hpux 11.0 )

[Larry Smith_12]

Hi Massimo,

I stumbled upon your posting while googling for more information on this problem. It just so happens that I am running across the same issue that you have described here. In my case, I have a single HPUX 11.11 client and a bunch of Solaris clients. My NFS server is also running Solaris 8. The odd thing is that I do not have any problems with my Solaris clients. However, the HPUX client is exibiting the exact same behavior you described. I am also getting these same syslog messages. I was wondering if you were able to figure out what was the source of this problem in your case. Any help will be greatly appreciated.

Thanks in advance