TLDR
- for a specific set of modules, not all (e.g
scipy-stack, python, arrow, opencv)
- warn when user load
scipy-stack without pinning a version
Goals:
- Decrease the number of users, job crashes and tickets related to updated modules where one did not pin a version and got the latest by default
- Increase reproducibility of job scripts
When we add a new version of a module like scipy-stack, it can create issues for users who do not specify a version for the module they use. Hence, some users get the latest module without realizing it and their code may break resulting in a job crash and tickets. This also hinders reproducibility of job scripts as the modules loaded may have changed.
I suggest to add a warning to a specific set of modules : scipy-stack, arrow, python, opencv, and others we judge useful, when no version is specified . Something like
$ module load scipy-stack
Please specify a version instead of relying on the default one! -> module load scipy-stack/2023a.
TLDR
scipy-stack,python,arrow,opencv)scipy-stackwithout pinning a versionGoals:
When we add a new version of a module like
scipy-stack, it can create issues for users who do not specify a version for the module they use. Hence, some users get the latest module without realizing it and their code may break resulting in a job crash and tickets. This also hinders reproducibility of job scripts as the modules loaded may have changed.I suggest to add a warning to a specific set of modules :
scipy-stack,arrow,python,opencv, and others we judge useful, when no version is specified . Something like