set default singularity cache directory to $WORKDIR/.singularity

[kelly-sovacool]

similar to RENEE's behavior.
Otherwise, users will run out of space if their SINGULARITY_CACHEDIR env var is set to the default in ~/.singularity

[kopardev]

Fixed on the feat_singularity_cache_fix branch. Alexei testing it out.

[kopardev]

did not fix the issue according to Alexei... I am not sure whye.. asking Kelly to test.

@kelly-sovacool can you run these

module load ccbrpipeliner
/gpfs/gsfs10/users/CCBR_Pipeliner/Pipelines/ASPEN/feat_singularity_cache_fix/bin/aspen --help
/gpfs/gsfs10/users/CCBR_Pipeliner/Pipelines/ASPEN/feat_singularity_cache_fix/bin/aspen -m=init -w=/data/CCBR/some/path
/gpfs/gsfs10/users/CCBR_Pipeliner/Pipelines/ASPEN/feat_singularity_cache_fix/bin/aspen -m=dryrun -w=/data/CCBR/some/path
/gpfs/gsfs10/users/CCBR_Pipeliner/Pipelines/ASPEN/feat_singularity_cache_fix/bin/aspen -m=run -w=/data/CCBR/some/path

This should pull the dockers in /data/${USER}/.singularity folder and not /home/${USER}/.singularity folder ... Can you confirm this expected behavior?

[kelly-sovacool]

/gpfs/gsfs10/users/CCBR_Pipeliner/Pipelines/ASPEN/feat_singularity_cache_fix/bin/aspen -m=run -
w=/data/$USER/aspen_test_iss-33

The message says it is setting the path correctly:

aspen
Working Dir       : /gpfs/gsfs12/users/sovacoolkl/aspen_test_iss-33
Samples Manifest  : /gpfs/gsfs10/users/CCBR_Pipeliner/Pipelines/ASPEN/feat_singularity_cache_fix/config/sam
ples.tsv
Singularity cache dir is NOT provided and NOT set!
Setting it to "/data/sovacoolkl/.singularity"
python is /usr/local/bin/python

However, it did not change my default singularity cache dir in the shell (which I have set in my bashrc):

echo $SINGULARITY_CACHEDIR

/data/CCBR_Pipeliner/SIFS

I also noticed there isn't a statement to export the cache dir in the bash submission script submit_script.sbatch.
Would it be better to set the cache dir in the bash submission script rather than modifying this variable in the user's shell, in case this would have unintended consequences for other tools?

[kelly-sovacool]

Trying to implement this sort of complex feature makes me want to rewrite the CLI in Python so we can begin writing unit tests, like we have for RENEE.