NFS problems

[Pablo Rudaeff]

I installed SFU 3.5 on W2k03 for using NFS Gateway to NFS Server installed on HPUX ver 11.0.
The file system on HPUX Server is defined as automounted
At the Windows side, there is no problems at all. The problem is in the HPUX side:
- there is a windows apps that copies files (30-40 ascii files every 15 minutes) to an nfs share in HPUX, and another unix apps that reads this files (from the same nfs).
Sometimes when the unix app tries to read these files (the unix app runs on the same HPUX nfs server), the process hangs-> I cannot kill the process with this error:
NFS Server (pid xxxxxx@/home/dir not responding still trying (pid xxxx is pointing the automounter process)

At the same time, all other nfs clients (windows, unix etc) are working fine.

To resolve the problem and I need to reboot the Unix Server.

The problem is in our SAP Production Server, ---> Please HELP!!!!!!!

[RAC_1 · Accepted solution]

It may happen that the windows puts the files and it has not finished the work. at the same time if application tries to read files. (File may be in use by windows)

You may also want to send -USR2 signal to autmount/automountd process. this will put it in debug mode and logging will start. Then check log file in /var/adm/syslog. Tis may give you some idea on what exactly is happeneing.

Anil

[Jan Sladky]

Hi Pablo,

first of all you should be sure that windows finished copying before unix apps read these files.
Are you sure ? Have you synchronised these tasks ?
Can you prolong the cron period ?

rgds Jan

[Pablo Rudaeff]

Hi Anil,
Yes, we thought that was the problem, and we test int in debug mode.
Thanks!!!!!
Pablo

[Eric Antunes]

Hi Pablo,

Do you have your HP-UX patches up to date? Check at the "Find Patches for my HP-UX system" section to see you you have all the NFS related patches installed:

http://www4.itrc.hp.com/service/patch/searchBySystem.do?BC=patch.breadcrumb.main|

Here are some important ones:

PHKL_22170, PHKL_24082, PHNE_20371, PHNE_27902, PHNE_29231, etc...

Best Regards,

Eric Antunes

[Pablo Rudaeff]

Hi Eric,
Allready done, didn't help
Thanks a lot,
Pablo

[Todd Whitcher]

Hi Pablo,

Just to get a clear picture of what your set up is like you have your HPUX 11.0 server configured as a NFS server sharing out files to Unix and Windows systems running SFU and the same HPUX server is using Automounter to access these same NFS files? So the HPUX is both a NFS server and a NFS client to itself via Automounter? Correct?

Are you in a ServiceGuard environment?

So if the set up is like I described the problem is local to the HPUX server since its a NFS server and cleint to itself.
Or is the /home/dir mounted from another NFS server?

Run a exportfs -v on your NFS server and post that output.

Also post the contents of nfsstat -m


With the Legacy Automounter it simulates a NFS server. So you see NFS server pid XXXX@/home/dir not responding. Debug should help you clarify which NFS server is not responding, your local system or a remote system.

The error message is the same as "NFS server not responding"


Automounter Debug:
Enabling debug to the Automounter via the Kill -17 (SIGUSR2) as suggested earlier will
TOGGLE debug on/off to /var/adm/automount.log (not syslog.log)

This should show you exactly what automounter is complaining about.

Test w/ Static NFS instead of Automounter:

Set up a static NFS mount point when the Automounter problem occurs, see if you can access this mount point outside of the Automounter client application to narrow down your issue. Is it a problem w/ NFS in general or just Automounter etc.

Create a test directory

ex. /nfstest

mount nfs_server:/exported_file /nfstest

See if you can access it.

Use RPCINFO to the NFS server to test the responsiveness of your nfs application.

rpcinfo -p nfs_server_name

That prints a list of registered programs

rpcinfo -u nfs_server_name nfs

That sends a null call to the nfs server to see if its responding, it should be.

There isnt enough data to t-shoot this via the itrc but testing the above should help you out. Basically there is a NFS server that is not responding and you need to figure out why.

There is a white paper available here that gives a lot of great information on t-shooting NFS performance issues.

http://www.docs.hp.com/en/netcom.html#NFS%20Services

Hope this helps

Todd

[Pablo Rudaeff]

Hi Jan,

You are right, it may happen the writting in Windows and the reading in HP-UX occur at the same time, which one has its own scheduling.
We are planning a redesign of the application with synchronization of the activities, first the writting of the files
by the Windows and then the reading by the application in HP-UX.
We will use a product of BMC called ControlM
for scheduling of both Windows and HP-UX programs.
Thanks,
Pablo

Hi Todd,

The picture is as you describe it, the same host is playing the role of NFS Server and NFS client.
We are in a Service Guard environment too.
After the third "crash" with the automounter hanging and need of reboot of the server, we
had to find an immediate solution.
This was the remove of the new filesystem from the automounter definition and add to
the static NFS.
The application is running now about 24 hours
w/o problems.
The second step is the redesign of the application, so that the writting and the reading will be synchronized.
The third step will be the redefine of the filesystem to the automounter.

Thank you a lot for the help,

Pablo