python - Pandas import sometimes fails

Question

I'm running hundreds of parallel Python jobs on several Windows servers (using HTCondor for job distribution). I have set up a Conda environment on a network path - this environment has pandas installed. I have confirmed that the pandas install is working as expected, by activating the environment and testing that the import works.

However, out of several hundred jobs being launched, a handful of them fail with strange import errors. For example:

ModuleNotFoundError: No module named 'pandas.core.arrays.datetimes'

Or

ModuleNotFoundError: No module named 'pandas.core.dtypes.dtypes'

Am I dealing with bizarre network behaviour here, or is there something wrong with my environment?

Answer 1

等待大神答复