Looks like schedd is not starting.
Have you checked the master log on the submit node to see what's going on?
condor_status works, but condor_q fails on all 3 nodes, but with different outputs:
`condor_q` isn't expected to work on anywhere other than the submit node without further configuration.
Since the submit node is definitionally where the schedd is running, it's more than a little alarming that `condor_q` is trying to find a schedd running on the central manager instead. What does `condor_status -any` report?
- ToddM