Page 1 of 1

Code_Saturne V6 installation on Cluster with SLURM using openmpi/3.0.0

Posted: Thu Jul 25, 2019 11:28 am
by samir_laurent
Dear Code_Saturne Dev Team, Dear Yvan,

I tried to install a V6 CS version on our cluster.
1) git clone -b V6.0
2) config.log in attachments
3) No problem with the install
4) The case is running on the master node
5) "Impossible to get a valid username to create temporary files"
Open MPI was unable to obtain the username in order to create a path
for its required temporary directories. This type of error is usually
caused by a transient failure of network-based authentication services
(e.g., LDAP or NIS failure due to network congestion), but can also be
an indication of system misconfiguration.

Please consult your system administrator about these issues and try
again.


6) Nothing is written in the Result folder (post-processing, listing is pointing to run_solver.log which is empty)
7) [node202:14152] 5 more processes have sent help message help-orte-runtime.txt / orte:session:dir:nopwname
[node202:14152] Set MCA parameter "orte_base_help_aggregate" to 0 to see all help / error messages
solver script exited with status 64.

Error running the calculation.

Check Code_Saturne log (listing) and error* files for details.

Error in calculation stage.


This message is the output of Slurm error file.

Thank you for your patience :)

Laurent

Re: Code_Saturne V6 installation on Cluster with SLURM using openmpi/3.0.0

Posted: Thu Jul 25, 2019 12:38 pm
by Yvan Fournier
Hello,

Could you post your "post-install" code_saturne.cfg file ?

If you have multiple MPI libraries installed, uncommenting and editing the mpiexec entry in that file may be necessary to ensure the command matches that of the library you used and is associated with the batch system.
This has not changed significantly since at least v4.0.

Regards,

Yvan

Re: Code_Saturne V6 installation on Cluster with SLURM using openmpi/3.0.0

Posted: Thu Jul 25, 2019 5:00 pm
by samir_laurent
Thank you for your very quick reponse!

Where can I found the code_saturne.cfg file? I got only code_saturne.cfg_template in the install repository.

SRC_REP : /src/code_saturne-6.0/
BUILD_REP: /build/code_saturne-6.0/
INSTALL_REP: /opt/code_saturne-6.0/

I got only /opt/code_saturne-6.0/etc/code_saturne.cfg.template.

Please do not hesitate to ask me more precisely information if needed!

Regards,
Laurent

Re: Code_Saturne V6 installation on Cluster with SLURM using openmpi/3.0.0

Posted: Thu Jul 25, 2019 6:15 pm
by Yvan Fournier
Hello,

Search for "post install" in the installation guide : https://www.code-saturne.org/cms/docume ... stallation (installation for v6.0 has not changed relative to v5.3).

If you have not done the post-install on a cluster, you can not directly use the batch system from the GUI, and chances are you will point to the wrong MPI, as auto-detection is tricky.

Regards,

Yvan

Re: Code_Saturne V6 installation on Cluster with SLURM using openmpi/3.0.0

Posted: Thu Jul 25, 2019 7:55 pm
by samir_laurent
Hi,

I just copy the code_saturne.cfg_template file into code_saturne.cfg.
I do not perform any changes in the file. All the lines are commented.
And it works with an sbatch runcase command...
I forgot to give you this detail. I did not use the GUI to launch simulation..

When I try to fill in some field, I got a python parser error.

Any idea to explain that?
1) the fact that it works now with a commented .cfg file
2) my misunderstanding of the way to fill in the file. (I've tried with or without ' ' I saw the post of Quentin..)

Please find in attchments the code_saturne.cfg file that I am trying to fill in. (I changed the extension to attach the file)

What I understand is that code_saturne.cfg replace the old part of runcase in 2.X versions right? Using a .cfg file enable us to schedule Code_Saturne job directly through the GUI?

Regards,
Laurent.

Re: Code_Saturne V6 installation on Cluster with SLURM using openmpi/3.0.0

Posted: Fri Jul 26, 2019 10:19 am
by Yvan Fournier
Hello,

No, this is not how it works. You must copy the template file into the .cfg file, but only uncommented / edited entries will modify the script's behavior. You still need to use the runcase (but if you have uncommented batch = SLURM or replaced it with a batch template tailored for your cluster, you will have a batch directives at the top of newly created runcase files).

I also recommend using "code_saturne submit" instead of sbatch once the batch system is configured (as it compiles user files interactively and then submits the prepared run).

In any case, you seem to need to edit the mpiexec entry so as to point the the absolute path of the mpiexec command matching your install, to avoid confusion with other similar commands in your path.

Regards,

Yvan