You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Only run test_derecho_mpirun Python test on Derecho #2987
changed the title [-]Only run test_derecho_mpirun unit test on Derecho[/-][+]Only run test_derecho_mpirun Python unit test on Derecho[/+]on Mar 6, 2025
Or maybe this should be moved out of our Python testing entirely—maybe even out of our repo? @ekluzek would it make sense for this to be a CIME test instead?
changed the title [-]Only run test_derecho_mpirun Python unit test on Derecho[/-][+]Only run test_derecho_mpirun Python test on Derecho[/+]on Jun 6, 2025
@ekluzek This came across my radar again. Why does this need to be one of our Python system tests? Could it be a SystemTest that's part of aux_clm (or its own special suite) instead?
Actually this test doesn't actually require running on Derecho. It doesn't actually run mpi, it just makes sure that the mpirun command is as expected. So it can be run anywhere.
It would be good to run all the python tests in a github runner. But, this one isn't any different than the rest.
So I'm closing, but feel free to reopen if you think it shouldn't be.
Or maybe this should be moved out of our Python testing entirely—maybe even out of our repo? @ekluzek would it make sense for this to be a CIME test instead?
This is a good point. That might make sense. The reason I have it here is to make sure that mpi is going to work correctly for mksurfdata_esmf. So in a way it's particular to a CTSM tool.
Or maybe this should be moved out of our Python testing entirely—maybe even out of our repo? @ekluzek would it make sense for this to be a CIME test instead?
This is a good point. That might make sense. The reason I have it here is to make sure that mpi is going to work correctly for mksurfdata_esmf. So in a way it's particular to a CTSM tool.
Why have this separate test in our Python suite—does the mksurfdata_esmf testing not effectively check this?
The mksurfdata test is more complicated. And when it was failing due to a change in cime/ccs_config, it wasn't clear what the problem was. This test was able to clearly show the problem in that case.
So this is here in case changes in cime or ccs_config break mksurfdata again. It should help us to figure it out quickly.
That's not what I meant, but I don't like what I meant anymore anyway! Given Erik's use case, where this is only really useful during mksurfdata_esmf debugging, I think it's fine to skip during e.g. GitHub Workflow runs. With #3761 it will be easy to make that happen.
Activity
[-]Only run test_derecho_mpirun unit test on Derecho[/-][+]Only run test_derecho_mpirun Python unit test on Derecho[/+]samsrabin commentedon Jun 6, 2025
Or maybe this should be moved out of our Python testing entirely—maybe even out of our repo? @ekluzek would it make sense for this to be a CIME test instead?
[-]Only run test_derecho_mpirun Python unit test on Derecho[/-][+]Only run test_derecho_mpirun Python test on Derecho[/+]samsrabin commentedon Feb 11, 2026
@ekluzek This came across my radar again. Why does this need to be one of our Python system tests? Could it be a SystemTest that's part of
aux_clm(or its own special suite) instead?1 remaining item
ekluzek commentedon Feb 12, 2026
Actually this test doesn't actually require running on Derecho. It doesn't actually run mpi, it just makes sure that the mpirun command is as expected. So it can be run anywhere.
It would be good to run all the python tests in a github runner. But, this one isn't any different than the rest.
So I'm closing, but feel free to reopen if you think it shouldn't be.
ekluzek commentedon Feb 12, 2026
This is a good point. That might make sense. The reason I have it here is to make sure that mpi is going to work correctly for mksurfdata_esmf. So in a way it's particular to a CTSM tool.
samsrabin commentedon Feb 12, 2026
@ekluzek I'm reopening because this is still an issue, but it's a different issue. Will work on it eventually.
samsrabin commentedon Feb 12, 2026
Why have this separate test in our Python suite—does the
mksurfdata_esmftesting not effectively check this?ekluzek commentedon Feb 12, 2026
The mksurfdata test is more complicated. And when it was failing due to a change in cime/ccs_config, it wasn't clear what the problem was. This test was able to clearly show the problem in that case.
So this is here in case changes in cime or ccs_config break mksurfdata again. It should help us to figure it out quickly.
samsrabin commentedon Feb 12, 2026
So would it be okay to move this test out of our Python testing dir and into
tools/mksurfdata_esmf/?slevis-lmwg commentedon Feb 14, 2026
By that do you mean "out of python testing and into clm_pymods" or something along those lines? I would consider that.
samsrabin commentedon Feb 17, 2026
That's not what I meant, but I don't like what I meant anymore anyway! Given Erik's use case, where this is only really useful during
mksurfdata_esmfdebugging, I think it's fine to skip during e.g. GitHub Workflow runs. With #3761 it will be easy to make that happen.