Svilen wrote:
> I'm running a small home network (Fedora core 3/4) with a shared /home
> disk. NFS seems to work happily untill the moment I shut down one of the
> client machines. It complains about (RPC error 101) not being able to
> shut down the host process. In result an empty file named on the machine
> IP remains in /var/lib/nfs/sm
>
> This is still not a big issue, the real trouble comes when I reboot the
> server. The file I mention above is moved to /var/lib/nfs/sm.bck
> directory during the shutdown (a message appears telling me that not all
> clients has been cleanly unmounted). During the following boot-up
> nfslock daemon crashes on the server - in result when I try to launch
> KDE for example from the client machine, it hangs-up and I can see on
> the console some messages like
> "lockd:server 192.168.0.3 not responding..."

This sounds not too different from a problem I had about 10 years
ago. File locking doesn't work all that well with unix, and
even less well with NFS.

NFS is supposed to be stateless, but file locking can't be
stateless, so a stateful protocol was added. /var/lib/nfs/sm
is the memory of files locked through a reboot.

As far as I know, the solution is to remove sm and sm.bck,
and any other files related to file locking, restart the
server and the clients.

Then again, things might have changed in 10 years.

-- glen

glen herrmannsfeldt wrote:
I've had lost hope that somebody will help with this problem - so
thanks for the post.
Yes, that is resolving the trouble, but just untill the next server
reboot. I was looking for a permanent solution. Besides it doesn't feel
very comfortable to know that something in your home directories is dodgy.
The problem is still there and any help will be really appreciated

Svilen

> Svilen wrote:
>
>> I'm running a small home network (Fedora core 3/4) with a shared /home
>> disk. NFS seems to work happily untill the moment I shut down one of
>> the client machines. It complains about (RPC error 101) not being able
>> to shut down the host process. In result an empty file named on the
>> machine IP remains in /var/lib/nfs/sm
>>
>> This is still not a big issue, the real trouble comes when I reboot
>> the server. The file I mention above is moved to /var/lib/nfs/sm.bck
>> directory during the shutdown (a message appears telling me that not
>> all clients has been cleanly unmounted). During the following boot-up
>> nfslock daemon crashes on the server - in result when I try to launch
>> KDE for example from the client machine, it hangs-up and I can see on
>> the console some messages like
>> "lockd:server 192.168.0.3 not responding..."
>
>
> This sounds not too different from a problem I had about 10 years
> ago. File locking doesn't work all that well with unix, and
> even less well with NFS.
>
> NFS is supposed to be stateless, but file locking can't be
> stateless, so a stateful protocol was added. /var/lib/nfs/sm
> is the memory of files locked through a reboot.
>
> As far as I know, the solution is to remove sm and sm.bck,
> and any other files related to file locking, restart the
> server and the clients.
>
> Then again, things might have changed in 10 years.
>
> -- glen
>