NFS problems

[Ben Wilson_4]

Hi,
I have half a dozen HP-UX servers that are using NFS without any problems. I wanted to seperate the NFS traffic to a seperate subnet, so I turned up one of the other interfaces on the hosts (each host has at least 4 interfaces, 3 where not being used!), added host names for those interfaces, and updated the host tables on each machine. I can ping through my new subnet with no problem now.

BUT, when I tried to remount the NFS shares with the new hostnames, I am getting a PERMISSION DENIED error (the umounts occured without incident).

Any ideas? My guess would be that NFS isn't "allowed" through the new interface, but I'm not sure where to look for this type of configuration, unfortunatley I'm not terribly familar with HP-UX.

Thanks!
Ben

[Sridhar Bhaskarla · Accepted solution]

Hi,

On all the NFS servers (where you exported your directories), make sure /etc/exports file has been updated with the "new" hostnames. Use exportfs -au and exportfs -a to re-export the directories|file systems.

-Sri

[Jeff Schussele]

Hi Ben,

Few questions:

1) What search order does your /etc/nsswitch.conf use?

2) If dns THEN files, are these new hostnames in DNS?

3) Any hostnames designated for access in the /etc/exports file that need to be changed?

4) Any restrictions in /var/adm/inetd.sec that would disallow these new server names or subnets?

HTH,
Jeff

[Todd Lehr]

You probably need to update your /etc/exports file on the server. When the mount connection is made the server does an IP address to name conversion and then compares that to the contents of the /etc/exports.

Once the /etc/exports files is updated do a exportfs -a to activate it

[Ben Wilson_4]

Doh! I didn't even have to finish reading that to realize I forgot to update the exports...

Fixed that, now I'm getting a timeout:

nfs mount: getaddr_nfs: hbgint01_internal: NFS service not responding
nfs mount: retry: retrying(1)
nfs mount: retry: giving up

hbgint01=public, old nfs mount hostname
hbgint01_internal= private(not routable) new nfs hostname

nsswitch.conf is files first, and the names are resolvable with icmp, ftp, telnet, etc.

I'm probobly missing something easy again, you guys are just too fast!

So how do I do these points anyway? Can I split them up, there's a total of 10 per question?

Thanks,
Ben

[Jeff Schussele]

Hi (again) Ben,

Did you do the
exportfs -au
exportfs -a
after the edit?

Else just bounce the NFS server.

/sbin/init.d/nfs.server stop
/sbin/init.d/nfs.server start

AND you can assign up to 10 pts for EACH answer if you so desire.

Rgds,
Jeff

[Jeff Schussele]

I'm a dummy...

I should explain that the exportfs & /sbin/init.d/nfs.server stop/start commands are for the exporting server(s) only.

How are you mounting these on the NFS clients?

If you do from command line make sure you use the new hostname there.

If you added these to the /etc/fstab files such that they'll mount at boot, then you'll need to edit the /etc/fstab file for the proper hostname(s) and run
mount -a
to remount these. They'll also mount at boot then.

Entries to /etc/fstab look like as follows:

srvr_name:/path/to/export_dir /mount/point nfs rw 0 2

HTH,
Jeff

[Todd Lehr]

You mention that you can ping on the new subnet, but does that mean you can ping from the NFS server to client and vice-versa.

Also do a

traceroute client-name from the server
traceroute server-name from the client

Do you have 2 interfaces on the NFS server or one, and if only one how is the traffic being routed between the new subnets

[Ben Wilson_4]

The mystery deepens....
I couldn't see anything else wrong with what I am trying to do, so I went on to the other hosts, and they worked fine! The only problem is with the first host, he can mount everyone elses shares, but no one can mount him.

I decided to restart the nfs.core, nfs.server, and nfs.client. Well, that made it worse. I can't mount on either interace/hostname now:

nfs mount: get_fh: hbgint01:: RPC: Program not registered

is my error now, and it looks worse.

Is there a specific order the services should be started? I did cor, server, client.

Thanks for all the help,
Ben

[Shannon Petry]

The timeouts are caused by name-number translations 90% of the time.

Ensure that the new hostnames and ip's are in all the systems host files, NIS, DNS, NIS+, or whatever is relevant.

Also, make sure that noone put a mountd entry into /var/adm/inetd.sec which would not allow the new hostnames to connect.

Regards,
Shannon

[Ben Wilson_4]

rebootnig the offending node solved the last of my problems, thanks for the help!

Ben