That said, this is what the line should look like:
## Package Plan ##
environment location: /mnt/anw-gold/NP/yvdwerf/data_ENIGMA_PD/projects/sct/sct_5.7/python/envs/venv_sct
In other words, it seems like the conda environment gets created in the wrong location due to the fact that there is an extra bit of “$CONDA_WORKING_DIR” text instead of the expected “python/” text, which then later on causes the “pyvenv.cfg” failure.
Because of this, my hope is that by temporarily setting the “$CONDA_WORKING_DIR” variable to be “python/”, then the variable will evaluate to the correct path.
Thanks so much for your suggestion and elaborate explanation. The $CONDA_WORKING_DIR is indeed already located somewhere else on the server (in the home dir (/mnt/home2/yvdwerf)). Your suggestion seems to solve that particular issue, but later on in the script apparently still wants to do stuff in the home dir, which has very limited space. That may explain the following error (see also log attached):
Building wheels for collected packages: docopt, pathtools, promise, transforms3d
** WARNING: Building wheel for docopt failed: [Errno 28] No space left on device: ‘/mnt/home2/yvdwerf/.cache/pip /wheels/72’**
** WARNING: Building wheel for pathtools failed: [Errno 28] No space left on device: ‘/mnt/home2/yvdwerf/.cache/ pip/wheels/3e’**
** WARNING: Building wheel for promise failed: [Errno 28] No space left on device: ‘/mnt/home2/yvdwerf/.cache/pi p/wheels/29’**
** WARNING: Building wheel for transforms3d failed: [Errno 28] No space left on device: ‘/mnt/home2/yvdwerf/.cac he/pip/wheels/b5’**
Failed to build docopt pathtools promise transforms3d
It installed succesfully, thanks a lot to you both for your help. Despite the extra flag, it kept storing data (~545MB) in the cache folder of my home dir. Luckily, for this run I cleaned up the cache beforehand, leaving enough space for the complete installation.
Despite the extra flag, it kept storing data (~545MB) in the cache folder of my home dir.
Ah okay, ~/.cache/pip is something we hadn’t considered much before. I think we should talk about how to handle this internally, because surely you’re not the only person running SCT in constrained spaces.
I’m glad it’s working for you now. Have a good day!