I was under the impression that then when a condor job is submitted and executed on another node, condor will spoof things such that the job still sees the submit machine's file directory.
See the following page for a discussion of how HTCondor manages files in the absence of a shared filesystem:
https://htcondor.readthedocs.io/en/latest/users-manual/file-transfer.html
The other issue I'm seeing is that when I run condor_reconfig -al or condor_restart -all, I get this:
You probably need to set ALLOW_ADMINISTRATOR, probably to condor_pool@*/*. (Be sure to run those commands as root, since the tools can't otherwise read the pool password.)
- ToddM