CONDA style MOOSE installation issue in HPC

[mdh23rct]

Dear all,
I have installed Conda MOOSE Environment in our HPC cluster. However, when I run the slurm script, it gives the error "error while loading shared libraries: libwaspcore.so.04: cannot open shared object file: No such file or directory"

I looked at the projects/moose/conda/wasp/ directory. I am not seeing any libwaspcore.so.04 file in the wasp folder. What I can do now?

During cloning and making the moose folder I did not see any error, though.

I would appreciate if you could manage time to reply.

Thanks.

[milljm]

While working within an HPC enclave, it is best to use that clusters vetted compiler stack, rather than install ours via Conda.

You should instead be following these instructions: https://mooseframework.inl.gov/getting_started/installation/hpc_install_moose.html

The error you are receiving is likely a result of building with one set of compilers while attempting to use another (HPC modules verses our Conda packages).

The location you are describing moose/conda/wasp/ is the location containing the recipes necessary to build the Conda moose-wasp package used for redistribution on our https://conda.software.inl.gov/public channel. Basically, nothing you need concern yourself with.

[mdh23rct]

@milljm I ran the slurm script you have suggested. Please have the slurm output showing a lot of information with the same error
"./recryss-opt: error while loading shared libraries: libwaspcore.so.04: cannot open shared object file: No such file or directory"

slurm-output.txt

Please help. Thanks.

[GiudGiud]

can you report the output of ldd recryss-opt

[milljm]

I would be interested in seeing ldd recryss-opt as well.

One glaring concern of mine, is that recryss-opt was built using Conda compilers, yet Conda is no where to be found in your SLURM batch environment right before the binary is run.

[GiudGiud]

One glaring concern of mine, is that recryss-opt was built using Conda compilers, yet Conda is no where to be found in your SLURM batch environment right before the binary is run.

yeah I agree.
Let's conda activate moose right before the srun comand if you are going to be using a conda-build on this cluster (not usually recommended)

[milljm]

I don't see conda anywhere in the output. Not in the env | sort, and not in PATH... I am not so sure a conda activate will work in our case now that I think about it.

Strangely, the only mentioning of Conda, is our public channel. Hmm... actually, I think that is being set by our diagnostic script.

@mdh23rct, if you can re-run the ./diagnostic.sh script manually, and report those results again, that may prove helpful. By manually, I mean, not via a SLURM job submission.

[mdh23rct]

@GiudGiud @milljm Hi both, finally, I have been able to build MOOSE in our cluster with some failed tests. Since the XDR output is not supported in our cluster, it has been removed from the libmesh.sh script. I am not familiar with other failed tests. I have attached the slurm output file here. Would you be kind have a look into it and let me know if those failed tests will affect my simulations using the phase-field module?
slurm-output.txt

Thanks.

[mdh23rct]

@GiudGiud as I have mentioned,
I have the error:
make: [/mnt/lustre/a2fs-work4/work/e882/e882/mdh23rct/projects/moose/framework/moose.mk:399: prebuild] Error 1 (ignored)
after running make -j 6.
The question is how to solve this error? Jinja2 is a minor warning.

[GiudGiud]

The question is how to solve this error? Jinja2 is a minor warning.

maybe. But it all appears in one block so I d make sure that it's not that first. It's a ModuleNotFoundError not a ModuleNotFoundWarning

Traceback (most recent call last):
File "/mnt/lustre/a2fs-work4/work/e882/e882/mdh23rct/projects/moose/framework/../scripts/premake.py", line 20, in init
from versioner import Versioner
File "/mnt/lustre/a2fs-work4/work/e882/e882/mdh23rct/projects/moose/scripts/versioner.py", line 26, in
import jinja2
ModuleNotFoundError: No module named 'jinja2'

Traceback (most recent call last):
File "/mnt/lustre/a2fs-work4/work/e882/e882/mdh23rct/projects/moose/framework/../scripts/premake.py", line 261, in
PreMake().check()
File "/mnt/lustre/a2fs-work4/work/e882/e882/mdh23rct/projects/moose/framework/../scripts/premake.py", line 27, in init
self.versioner_meta = Versioner().version_meta()
UnboundLocalError: local variable 'Versioner' referenced before assignment
make: [/mnt/lustre/a2fs-work4/work/e882/e882/mdh23rct/projects/moose/framework/moose.mk:399: prebuild] Error 1 (ignored)

[mdh23rct]

@GiudGiud please find attached the output of ./run_tests -j 6 from directory /moose/test/.
slurm-output-2.txt
Do you think everything is fine inside?

[mdh23rct]

@GiudGiud when I run ./run_tests -j 6 from the application folder, I am getting this error:
'''ERROR: Your Plugin Tester "RunHPC" failed to import. (skipping)

No module named 'paramiko'

ERROR: Your Plugin Tester "RunSlurm" failed to import. (skipping)

No module named 'paramiko'

ERROR: Your Plugin Tester "RunPBS" failed to import. (skipping)

No module named 'paramiko'
test:kernels/simple_diffusion.test ............................................................................................................................ OK

Ran 1 tests in 1.9 seconds. Average test time 1.6 seconds, maximum test time 1.6 seconds.
1 passed, 0 skipped, 0 failed
'''
Last time, I did not have these error while building an application. Any reason why this is happening?

[GiudGiud]

This is a new tester we added. Did you update the moose subdmodule since last time?

you can install this python package to get this tester to work