CodeSaturne 3.0.3 issue on a PBS cluster

All questions about installation
Forum rules
Please read the forum usage recommendations before posting.
Post Reply
Bobopart
Posts: 8
Joined: Fri Feb 15, 2013 12:13 pm

CodeSaturne 3.0.3 issue on a PBS cluster

Post by Bobopart »

Hello everybody,

I am trying to use Code_Saturne 3.0.3 on a PBS cluster which use the Environment Modules package. When I submit a job, it instantaneously stops and no error files are generated.

I therefore try to just initialize the computation (in interactive mode using the options --initialize), and I got the following error :
code_saturne run --initialize --param CM_CS_V3_Cluster.xml
Traceback (most recent call last):
File "/cm/shared/apps/saturne/3.0.3/code_saturne-3.0.3/bin/code_saturne", line 77, in <module>
retcode = cs.execute()
File "/cm/shared/apps/saturne/3.0.3/code_saturne-3.0.3/lib/python2.6/site-packages/code_saturne/cs_script.py", line 73, in execute
return self.commands[command](options)
File "/cm/shared/apps/saturne/3.0.3/code_saturne-3.0.3/lib/python2.6/site-packages/code_saturne/cs_script.py", line 125, in run
return cs_run.main(options, self.package)
File "/cm/shared/apps/saturne/3.0.3/code_saturne-3.0.3/lib/python2.6/site-packages/code_saturne/cs_run.py", line 201, in main
domains=d)
File "/cm/shared/apps/saturne/3.0.3/code_saturne-3.0.3/lib/python2.6/site-packages/code_saturne/cs_case.py", line 95, in __init__
cs_exec_environment.set_modules(self.package_compute)
File "/cm/shared/apps/saturne/3.0.3/code_saturne-3.0.3/lib/python2.6/site-packages/code_saturne/cs_exec_environment.py", line 249, in set_modules
stdout=subprocess.PIPE).communicate()
File "/usr/lib64/python2.6/subprocess.py", line 642, in __init__
errread, errwrite)
File "/usr/lib64/python2.6/subprocess.py", line 1234, in _execute_child
raise child_exception
OSError: [Errno 2] No such file or directory
I therefore investigate a little bit, but it is not easy since I don't have admin privileges. The problem seems to come from the function set_modules in the script cs_exec_environment.py. Then, using the previous traceback, I went back to cs_script.py, and I noticed the line:
self.env_modulecmd = "/cm/local/apps/environment-modules/Modules/3.2.10/bin/modulecmd"
The problem is that this path is not correct, the installed version of Environment Modules is 3.2.6 on the cluster. Maybe the version has been changed after the installation of Code_Saturne.

So, do you think that it is the reason of my problem or am i completely wrong ? It is the first time I use Code_Saturne on a cluster, and I am not familiar with such computations.

Thanks a lot

Boris
Yvan Fournier
Posts: 4070
Joined: Mon Feb 20, 2012 3:25 pm

Re: CodeSaturne 3.0.3 issue on a PBS cluster

Post by Yvan Fournier »

Hello,

Yes, your analysis seems good.

There is an install option for Code_Saturne to ignore modules (i.e. not try to recored loeded modules and reload those when running) at configure time (--with-modules=no).

Unfortunately, we did not plan an option to bypass module handling (to handle you own) for an installed verion, assuming that if the module command itself was updated, modules used to compile the code would change also (which might or might not be the case).

A deeper implementation issue is that environment modules are usually handled as an alias, and commands in specific directories, such as "/cm/local/apps/environment-modules/Modules/3.2.10/bin/modulecmd", but not executable in the main path which might be "stable" across upgrades, making automation more diffficult.

So ideally, if the environment changed, Code_Saturne should be reinstalled (in which case I recommend upgrading to the latest bug-fix release, 3.0.9).

Is it realistic to ask the admins to handle this ? If not, I can provide suggestions on installing the code on your own account (Code_Saturne does not require admin priviledges).

Also note that version 3.0 is nearing end of life: when we release 5.0 late April, 3.0 will become "obsolete" and reach the end of its 4 year maintenace lifetime (though this is a separate issue).

Best regards,

Yvan
Bobopart
Posts: 8
Joined: Fri Feb 15, 2013 12:13 pm

Re: CodeSaturne 3.0.3 issue on a PBS cluster

Post by Bobopart »

Hello Yvan,

I do understand. I am trying to install Code_Saturne V3 on my own account, up to now, it goes smoothly. If it fails, I will ask the administrator of the cluster to reinstall Code_Saturne with the option --with-modules=no.

For your advice, I know that V3 will become obsolete soon but we developed specific models for the lagrangian modules. I am quite afraid to integrate them in the new stable versions since we modified some core subroutines of the lagrangian module (lagune, lages1, etc.). I didn't have a look at the new version yet, but It seems that the lagrangian module has been quite modified since the version of Code_Saturne 3.0. Therefore, I guess it will take me a lot time to integrate our models in a newer version.
I have no solution to that problem yet, I would like to use the new stable version but I cannot spend too much time in integrating our models in this version.

Best regards,

Boris
Yvan Fournier
Posts: 4070
Joined: Mon Feb 20, 2012 3:25 pm

Re: CodeSaturne 3.0.3 issue on a PBS cluster

Post by Yvan Fournier »

Hello,

Yes, there were massive changes in the Lagrangian model last year. We did notice some bugs in the previous implementation of the model (especially, 2nd order was not working), so moving to the new version in the following months would be recommended (directly on version 5.0, bypassing version 4.0).

All that is statistics related should be simpler in the new version (I still have a few changes pending for the boundary statistics).

Best regards,

Yvan
Bobopart
Posts: 8
Joined: Fri Feb 15, 2013 12:13 pm

Re: CodeSaturne 3.0.3 issue on a PBS cluster

Post by Bobopart »

Hello,

That's the massive changes I am afraid of, it does sound as if it will not be easy to integrate our models in these new versions. But I know that I will have to do it in the near future.

For my problem, I installed Code_Saturne 3.0.7 in my home directory this morning, and I am currently doing some tests. Up to now, all is going smoothly.

Thanks for your advices !

Best regards

Boris
Post Reply