case.setup ERROR. What is "modulecmd"?

[bidyut]

Dear All,
I am trying to port CESM to a new machine. Here is a snippet of my config_machine.xml

<module_system type="module">
      <init_path lang="perl">/mnt/nfs/clustersw/Debian/bullseye/lmod/lmod/init/perl</init_path>
      <init_path lang="python">/mnt/nfs/clustersw/Debian/bullseye/lmod/lmod/init/env_modules_python.py</init_path>
      <init_path lang="csh">/mnt/nfs/clustersw/Debian/bullseye/lmod/lmod/init/csh</init_path>
      <init_path lang="sh">/mnt/nfs/clustersw/Debian/bullseye/lmod/lmod/init/sh</init_path>
      <cmd_path lang="perl">/mnt/nfs/clustersw/Debian/bullseye/perl/5.34.0/bin perl</cmd_path>          <--- I should have used MODULECMD here
      <cmd_path lang="python">/mnt/nfs/clustersw/Debian/bullseye/python/3.9.9/bin python</cmd_path>     <--- I should have used MODULECMD here
      <cmd_path lang="sh">module</cmd_path>
      <cmd_path lang="csh">module</cmd_path>
      <modules>
        <command name="purge"/>
      </modules>

I could not find "modulecmd" paths in my machine that are required in the config_machine.xml file (as shown above).

When I run the case.setup it gives me the following ERROR

./case.setup
ERROR: module command None purge  failed with message:
/bin/sh: 1: None: not found

My question is: Am I getting this ERROR because I did not use the "modulecmd" path? what is "modulecmd"? How can I fix this issue? Am I supposed to "install environment-modules"? [FYI: the machine does use module]

Thank you in advance.
Bidyut

 

[dbailey]

Moving this to Infrastructure and porting.

 

[jedwards]

Depending on what flavor of modules you have you may need something like
<cmd_path lang="python">/apps/lmod/lmod/libexec/lmod python</cmd_path>
<cmd_path lang="perl">/opt/apps/lmod/lmod/libexec/lmod perl</cmd_path>

or something like:
<cmd_path lang="perl">/opt/modules/default/bin/modulecmd perl</cmd_path>
<cmd_path lang="python">/opt/modules/default/bin/modulecmd python</cmd_path>

This is the tool used to run module commands from a given shell (sh, csh, perl, python)
You may need to check with your system administrator.

 

[bidyut]

Thank you @jedwards . My system admins helped me to resolve the issue. I am using module system "None" in the config_machine and instead loaded the modules manually. It is working at the moment although not the best solution probably.

 

[jedwards]

You should not need to do that - what version of modules do you have that doesn't have one of the commands above available?

 

[bidyut]

Hi @jedwards ... sorry for the prolonged silence. I tried the way you suggested and could get it going. Thank you.

But, now it is failing while building csm_share. I did see some "shr_infnan_isnan_real" issue in the csm_share.bldlog and tried the fix that you had suggested in CESM2.1.0 case.build error with CSM_share shr_infnan_mod.F90, but it did not work.

Attaching the config_machine.xml, config_compilers.xml (config_compilers.xml is essentially the downloaded copy, as is) and the csm_share.bldlog herewith.
With a modified config_compilers.xml file (suggested by the system admins; attached as the config_compilers.xml.modified), I get the same error.

Can you please suggest.
Thanks a lot.

 

[jedwards]

No files attached.

 

[bidyut]

Oh ... my bad. Did not notice that the files that I had attached did not have any of the allowed extensions and hence were declined. Sorry about that. Attached the files now as .txt files. Thank you.

 

[jedwards]

I think that this error indicates that your gnu compiler is too old. But you didn't tell me what version you are using.
Now that you have modules working maybe you should try intel again.

 

[bidyut]

gcc --version
gcc (Debian 10.2.1-6) 10.2.1 20210110

Does it help?

 

[bidyut]

Hi @jedwards , I tried to install cesm.2.2.0. The same scripts/modules that were giving error (as mentioned above) with CESM.2.1.3 could build CESM2.2.0. However, the run fails. On a closer look, I realized I actually get the error during ./case.setup [Porting error: source not found]. Despite the error during ./case.setup, ./case.build works but ./case.submit fails.

So, now I am following the discussions in Porting error: source not found and waiting for your reply there.

Thanks.

 

[bidyut]

Nonetheless, I am still curious, why I could build CESM2.2.0 but failed to build CESM2.1.3. For what its worth, the error reported in Porting error: source not found occurs while doing ./case.setup for both CESM2.1.3 and CESM2.2.0 in my case.
Thanks.