NFS from HP-UX to Linux

[Tonatiuh]

Hello Steven,

Maybe I described bad the role of each server.

HP-UX has the directory to be mounted.

Linux just mount the directory which exists in HP-UX.

The HP-UX /etc/exports is shown in the first message of this thread.

I have ran this in HP-UX but it does not shows anything when I try to write the NFS from Linux.

tail -f /var/adm/syslog/syslog.log

[Dave Olker]

Hello Tonatiuh,

So just to clarify - the 11.11 system is the NFS server. The Linux box is the NFS client. The client is able to successfully mount the filesystem. Users on the client can successfully READ from the filesystem. The users on the client cannot WRITE to the mounted filesystem.

Is this accurate?

If so can you please collect a network trace of the failing write request using either nettl on the 11.11 system or ethereal on the Linux system. Please post the raw trace file so that we can look at the trace to see if it shows any evidence of why the write is failing.

Regards,

Dave

[Tonatiuh]

Hello Dave,

>So just to clarify...

The scenario explained is exactly the one I have.

Soory for my ignorance. I am not sure about what must I do to collect this trace.

I have ran the nettl -start (in HP-UX), tried to create a file from Linux, then issued nettl -stop (in HP-UX). The file generated /var/adm/nettl.LOG000 is not readable.

[Dave Olker]

Hi Tonatiuh,

On the HP-UX server:

# /etc/nettl -tn pduin pduout loopback -e ns_ls_ip -f

This will start the nettl trace and log the packets to the file you specified with the -f option.

While the trace is running, reproduce the problem on the Linux client.

Once the problem has been reproduced, stop the nettl trace on the HP-UX server:

# /etc/nettl -tf -e all

Please try to reproduce the problem as quickly as possible and have the trace collecting as little time as possible as the trace files can very quickly grow large.

Once the trace is stopped, the trace file will be whatever name you assigned on the first nettl command-line above. It will have some kind of extention like .TRC00 or something similar. This is the raw/binary trace file. It is not ASCII text so it will not be readable.

Regards,

Dave

[Tonatiuh]

# /etc/nettl -tn pduin pduout loopback -e ns_ls_ip -f /netlog
nettl : The tracing and logging facility has not yet been started
To start the facility, execute the following command:
/usr/sbin/nettl -start
If the problem persists, contact HP representative with the
following information:
Shared memory tables removed.
(shmget, errno 2: No such file or directory).

[Dave Olker]

Bummer.

That's a known problem with kernel parameters being sized too small. Are you in a position to reboot your 11.11 system? If so I'll look up the appropriate kernel variables that you need to increase. You'd have to rebuild your kernel and reboot.

Is this something you can do at this point or can the box not be rebooted?

Dave

[Tonatiuh]

Dave,

Are you talking about my 'nettl' problem? If so...

I cannot reboot this server right now, but I can do it during the weekend.

Please help me about what kernel parameter I should increase.

[Dave Olker]

Wait a second, I just realized you issued the "nettl -stop" command previously when you were trying to collect a trace. I think you may have inadvertantly caused this problem by shutting down the nettl subsystem.

Try this:

# /sbin/init.d/nettl start

Once the nettl subsystem gets started again try starting a nettl TRACE using the commands in my previous post and see if they work. I think you may have just shut down the entire nettl subsystem, which needs to be running to collect a trace.

Regards,

Dave

[Tonatiuh]

I got it, I guess.

# /sbin/init.d/nettl start
Initializing Network Tracing and Logging...
Done.
# /etc/nettl -tn pduin pduout loopback -e ns_ls_ip -f /netlog
# etc/nettl -tf -e all
# /sbin/init.d/nettl stop
nettl stopped

I transfered the file from HP-UX to my Win PC using bin mode.

[Dave Olker]

Ok, I looked at the trace and the CREATE call is failing with an error of: NFS3ERR_ROFS.

This means the filesystem is mounted read-only. Either the local filesystem on the server is mounted read-only, the NFS filesystem on the Linux system is mounted read-only, or the filesystem is being exported on the server as read-only.

Please double-check how the filesystem is mounted both on the 11.11 server and on the Linux client. Also, check how the filesystem is being exported on the server. Make sure that it is not being mounted or exported as read-only.

Regards,

Dave