Cheyenne: CLM-FATES case build failed

[xiulingao]

Hi there,

I'm running CLM-FATES on cheyenne using NUOPC coupler. I was running the case after initiated a conda environment as there are some specific packages I need to use for generating parameter files for ensemble members. It failed at case build stage with error messages below:


/glade/work/user/conda-envs/py3.9/bin/x86_64-conda-linux-gnu-ar: two different operation options specified


gmake[1]: *** [libmpeu.a] Error 1


gmake: *** [subdirs] Error 2

I also attached the log file and software_environment file for more information. Is this due to the conda environment?

Thanks!

Xiulin

 

[jedwards]

Yes it looks like maybe there is a problem with the conda environment.
These flags
ifort: command line warning #10006: ignoring unknown option '-fallow-argument-mismatch'
ifort: command line warning #10006: ignoring unknown option '-mismatch_all'
ifort: command line warning #10121: overriding '-xCORE_AVX2' with '-march=nocona'
ifort: command line warning #10006: ignoring unknown option '-fno-plt'
ifort: command line warning #10006: ignoring unknown option '-ffunction-sections'
ifort: command line warning #10006: ignoring unknown option '-pipe'

Are not valid for the intel compiler and not set by cesm. Do you need conda to build or only to run?

 

[xiulingao]

Hi Jim,

Thank you for the response. The bash scrtipt I'm running to create, setup, build cases also calls a local R script to generate a bunch of parameter files by sampling acorss trait space. There are some specific R packages I need to install for doing this, which are not in the R module currently available on Cheyenne, and that's why I created a conda environment where I installed all the packages myself. So I think I don't even need the conda environment once all parameter files are generated as cesm does load necessary module during case build.

Is there a way to work around?

Thank you!

Best,
Xiulin

 

[jedwards]

If you run your script outside of the conda environment (with the R script commented out) and everything works, then
that points strongly to an issue in the environment. Can you enter the conda environment at the point where you call the
R script and then exit again after that step is complete?

 

[xiulingao]

Thank you Jim. I solved this by specifying the path of Rscript and run it without activating the conda environment. Now it works. And it's definitely the conda environment causing the issue.

Thank you again!