Closed wwieder closed 3 days ago
for anyone needing a functional ctsm_pylib environment on NCAR systems, you can use this YML file until we have a fix. /glade/u/home/wwieder/ctsm_pylib_ww.yml
@samsrabin or @slevis-lmwg is this something you can solve easily?
Hmm. the error in import requests
sounds like what @ekluzek and @glemieux are experiencing on Izumi, but it looks like this is happening for Katie on Derecho...
Generally nothing having to do with Python environments is easy, but some more diagnostic info might be helpful. @katierocci, could you point us to the directory from which you set up the Python environment? And do you know the exact command you used?
@samsrabin I set up the python environment in /glade/work/krocci/CTSM and ran the command ./py_env_create
You can see the diffs between my older ctsm_pylib environment that Katie was able to run with and the one that's created currently on derecho by diff-ing the yml files in my home directory
diff /glade/u/home/wwieder/ctsm_pylib_June24.yml /glade/u/home/wwieder/ctsm_pylib_ww.yml
Note, Katie's current env was created from ctsm_pylib_ww.yml. My current env is reflected in the *June24.yml file
@samsrabin wondered if we could move beyond python 3.7, but this isn't really possible on izzumi. With Izzumi going away maybe we have an opportunity to migrate to a newer python version.
@samsrabin will try pinning all packages in the conda_env_ctsm_py.txt file and get rid of the conda_env_ctsm_py_latest.txt file.
I was unable to reproduce this, so I asked Katie to run py_env_create
again with the created environment's name changed to ctsm_pylib_test
. That ended up working, so I guess something weird must have gone wrong in her original attempt. I'll close this issue.
However, in my testing, I did run into #2563. I'll take on that PR (#2607) because I'm personally annoyed by it now!
Brief summary of bug
run_neon won't work with newly created ctsm_pylib environments.
General bug information
CTSM version you are using: [output of
git describe
] master, ctsm5.2.007Does this bug cause significantly incorrect results in the model's science? [Yes / No] no, but it renders run_neon useless for new users
Configurations affected: [Fill this in if known.] run_neon, but possible other python tools that depend on ctsm_pylib
Details of bug
Out of the box run_neon fails with the following error. This was using a ctsm_pylib environment @katierocci created yesterday. The case ran fine when we used an older version of ctsm_pylib that I had instead.
We also were able to run with for another user who created the python environment a few weeks ago, using the same CTSM tag. This me think it's a recent update to derecho that's causing the incompatibility?
Important details of your setup / configuration so we can reproduce the bug