Setup global rc files for users

[ickc]

• which includes env var, e.g. SCRATCH pointing to pre-production CephFS and later changed to production CephFS path.
• document this
• decide supported shells (bash, zsh?)

[ickc]

We need to think about this more carefully. Summarizing the situation:

• Worker nodes are "blank state", they do not see the same FS nor environment nor even has a HOME (#4)
• Global rc files can either be setup to our submit node vm77, or facility-wise. I doubt the later would happen as it is not for our experiment only.
• If we change global rc files on the submit node vm77 only, HTCondor provides a way to propagate env var, at submission time, to the worker nodes when the job runs.
  • This however to me is a bad practice where the worker nodes are "blank state". If they do not see the same environment, but we propagate env var there, then it is a broken environment making things confusing to the end users, and hard to debug by us.